This thesis is dedicated to my father late Sidramayya Basayy Math (S.B.

Math)

& To my loving daughter Mansvi

Math(Gubbi)

For their endless love

ACKNOWLEDGEMENT
I consider, this is a great privilege to express my deep gratitude to many respected personalities who guided, inspired and helped me in the successful completion of my research. I thank the Lord almighty for his blessing throughout the period of the research without which I could not have completed this work. First and foremost, I would like to thank to the Indian Institute of Information Technology Allahabad for selecting me as a research scholar amongst many of applicants, and for providing all the facilities required to complete my work- particularly library, patent referral center lab, 24X7 high bandwidth internet and health centre, as well as the married research scholars apartment for my residence. I am also thankful to IIIT Allahabad for providing me the research fellowship during the entire period of research over there. I also express my deep sense of gratitude and respect to my supervisor, Prof. R.C. Tripathi, for his dedication, encouragement, motivation and his inspiring guidance throughout my thesis work. I am extremely thankful to Dr. M.D. Tiwari, Honble Director, Indian Institute of Information Technology-Allahabad for his constant inspiration, unreserved support and kindness to help me in every possible way to sustain my research work. I would also like to thank all the faculty, officers and supporting staff of Indian Institute of information Technology, Allahabad, and my inspiring colleagues and friends.

My greatest gratitude is extended to my family. Their unreserved support and motivation at every step of research is inexpressible. Without their continuous support and encouragement, I would not have completed my thesis.

Shrishail Math Date: Place: Allahabad RS-39

ii

ABSTRACT
We present new image forgery techniques that are capable of detecting traces of tampering in digital images which dont use any watermarks or specialized hardware etc. these are known as passive approaches. Such techniques work under the assumption that images contain natural properties from a variety of sources, including the scene view (world view), device view (lens and sensor) and local view. Basic assumption is that these properties may be disturbed by digital tampering and by measuring them, one can expose the forgery. In this context, we present the following techniques: (1) Exploiting image invariant features for detecting the copy paste forgery. (2) Supervised approach based on image quality features and moments for detecting the image compositing forgery. (3) The statistical methods based on basic image processing operation such as luminance, color space properties using HSV color space model and designing customized masks for filtering operations. The common theme of these techniques is that they exploit image features for specific type of cases. Although each technique is applicable to typical category of cases, they add to a growing set of image forensic techniques that together will compliment the

process of detecting digital image forgery to finally develop a unified tool.

iii

Table of Contents
Acknowledgement
..... .i ..iii

Abstract..... 1. Introduction.....................................................................................................................2 1.1Introduction..............................................................................................................2 1.2Motivations..............................................................................................................2 1.3Problem Formulation and scope..............................................................................5 Approach.....................................................................................................................7 1.3.1 Image Authenticity.....................................................................................7 1.3.2Methods.......................................................................................................7 1.4Thesis organization and Contribution......................................................................9 1.4.1Thesis Organization....................................................................................9 1.4.2Overview of the Thesis ...........................................................................10 1.4.3 Contributions............................................................................................11 2. Theoretical Background ...............................................................................................13 1.5 Introduction...........................................................................................................13 1.6 Forgeries...............................................................................................................13 1.6.1 History of Image Forgery.........................................................................13 1.6.2 Image Forgery in the Digital Era.............................................................15 1.6.3 Image Forgery Life Cycle........................................................................18 1.6.4 Image Forgery Types...............................................................................20 1.7Forgery Detection Approaches..............................................................................30 1.7.1Active Methods.........................................................................................31 1.7.2Passive Methods........................................................................................32

iv

1.8Image Formation....................................................................................................33 1.8.1Lighting.....................................................................................................35 1.8.2Reflectance and Shading...........................................................................35 1.8.3Optics........................................................................................................36 1.9 Camera Parameters...............................................................................................38 1.9.1Camera Settings........................................................................................39 1.9.2Sampling and aliasing: .............................................................................39 1.10Discussion ...........................................................................................................40 3. Literature Review.........................................................................................................41 1.11Introduction to digital image forensics................................................................41 1.12Active Forgery detection Methods......................................................................42 1.13Passive Blind Forgery detection Methods...........................................................43 1.13.1Copy paste Forgery detection.................................................................44 1.13.2Image composite forgery detection.........................................................49 4. Copy Paste Forgery Detection......................................................................................51 1.14Introduction..........................................................................................................51 1.15Related Literature ...............................................................................................51 1.16SIFT Features.......................................................................................................52 1.17SIFT Features based forgery detection system....................................................52 1.17.1The Proposed method..............................................................................52 1.17.2SIFT Feature Extraction..........................................................................53 1.17.3 Clustering...............................................................................................54 1.17.4Forgery Decision ....................................................................................55 1.18 Experiment and Results......................................................................................55 5. Image composite Detection using supervised approach...............................................58 1.19 Introduction.........................................................................................................58 1.20 Proposed Method ...............................................................................................58 v

1.21Experiment and Results.......................................................................................63 1.22 Algorithm ...........................................................................................................67 Conclusions and Discussions......................................................................................68 6. Image composite detection using Image Processing Methods.....................................69 1.23Introduction..........................................................................................................69 1.24Related Works.....................................................................................................69 1.25Forgery Detection Based on Image processing Methods....................................69 1.25.1Analysis of Luminance levels.................................................................69 1.25.2Analysis using HSV color space model..................................................70 1.25.3Detection of tampering based on customized filtering masks................71 1.26Methods based on Luminance level....................................................................73 1.27Methods based on HSV color space model.........................................................75 1.28Methods based on customized masks..................................................................76 1.29Methods based on customized filter masks.........................................................77 1.30Experiments and Results......................................................................................79 1.30.1Result analysis of luminance level test...................................................81 1.30.2Result Analysis of HSV test....................................................................81 1.30.3Result Analysis of Customized Filter.....................................................82 1.31Results and Discussions.......................................................................................83 7. Conclusion and Future Work........................................................................................85 1.32 Summary ............................................................................................................85 1.33Conclusions of Research......................................................................................85 1.34Recommendations for Future Research...............................................................86 1.34.1Multimodal forensics..............................................................................86 1.34.2Color Theory...........................................................................................86 1.34.33D scene Consistency.............................................................................87 1.34.4Fusion......................................................................................................87 vi

1.34.5Evolution.................................................................................................87 Morphing .........................................................................................................87 S.Math,R.C.Tripathi, Copy paste forgery detection using supervised approach, 2011 IEEE Conference on Computer Applications & Industrial Electronics, Shah Alam, Malaysia......................................................................................108

List of Figures
Figure 1: Popular news media reports on fake pictures......................................................3 Figure 2: Some fake pictures which appeared in popular media including Science Journal and National Geography...............................................................................4 vii

Figure 3: Two popular forgery creation techniques...........................................................6 Figure 4: Thesis Structure...................................................................................................9 Figure 5: First officially reported image forgery................................14 Figure 6: Most popularly circulated fake images in the last decade...........15 Figure 7: Most popularly circulated fake images in the last decade................................16 Figure 8: Most popularly circulated fake images in the last decade....17 Figure 9: Image forgery cycle....................................19 Figure 10: Image forgeries in history...............................................................................21 Figure 11: Image Composite..............................................................................23 Figure 12: Another example for image compositing........................................................24 Figure 13: Example for Copy paste forgery............................................25 Figure 14: Can you distinguish real images from computer generated images?...........26 Figure 15: Example for retouching...................................................................................27 Figure 16: Example for image Enhancing....................................28 Figure 17: Example for morphing....................................................................................29 Figure 18: Image forgery detection methods...........................31 Figure 19: Parameters influencing on image formation.....................32 Figure 20: Components of image formation.......................................34 Figure 21: simple image formation model...............................................................35 Figure 22: Snell's Law....................................................................36 Figure 23: Simple pinhole camera model.........................................................................37 Figure 24: Thin lens camera model ..........................................37 Figure 25: camera sensing pipeline................................................38 Figure 26: steps in active image forgery detection......................42 Figure 27: Scene authenticity parameters..............................43 Figure 28: Copy paste forgery creation model.........................45 Figure 29: Copy paste forgery detection frameworks.................................46 viii

Figure 30: Composite image creation model................................49 Figure 31: Original image Figure 32: Copy paste forged image...........................................................................................................................51 Figure 31: Original image Figure 32: Copy paste forged image...........................................................................................................................51 Figure 33: Schematic Diagram of proposed model..................53 Figure 34: Results of SIFT............................................................................................57 Figure 35: Proposed model based on image features...................................................58 Figure 36: Moment Extraction Procedure..............................................60 Figure 37 Prediction Context........................................61 Figure 38: Correctly judged as authentic images...............................65 Figure 39: Correctly judged as composite images..................66 Figure 40: misjudged as spliced images but these are authentic images ....................................................................................................................................66 Figure 41: misjudged as real images but these are composite images..........67 Figure 42: Example of changes in luminance levels........................................................70 Figure 43: a) original in RGB format b) HSV..................................................................71 Figure 44: Test image................................................................................74 Figure 45: luminance level at threshold 0.3..................................................75 Figure 46: Inverted result of figure 4 with threshold at 0.6..............................................75 Figure 47: Test image.......................................................................................................76 Figure 48: Forgery detected in HSV.................................................................................76 Figure 49: Test image.......................................................................................................78 Figure 50: Result of customized filter mask.....................................................................78 Figure 51: inverted result of customized mask result.......................................................79 Figure 52: Original test images...........................................................................................I Figure 53 original images .................................................................................................II Figure 54: forged images..................................................................................................III

ix

Figure 55: Result from HSV model..................................................................................IV Figure 56 Results from customized mask..........................................................................V Figure 57 Results from luminance test.............................................................................VI

List of Tables
Table 1: Results of SIFT method..55 Table 2: Results..(Chapter 5 Supervised approach)....63 Table 3: Image processing based methods for forgery detection..79 Table 4: summary of forgery detection methods and image formats used79 Table 5: Summary of Luminance based test results80 Table 6: Summary of HSV based test results...81 Table 7: Summary of Customized filter results...82 Table 8: Summary of All three methods..82

xi

1. Introduction 1.1 Introduction

Widespread easy and low cost availability of digital cameras and the prevalence of photo sharing and management websites such as flicker, Picasa and other popular websites provide the photo sharing and management applications in one or other forms. Digital images are playing every important role in our daily life; the digital images are omnipresent right from the cover pages of journals, newspapers, magazines etc. to evidences in court rooms, teaching aids etc. Images are used everywhere either as a personal memory evidences or for official purposes. The digital images can be forged; their authenticity cant be taken for a granted. The quote seeing is no longer believing is also true for images. The digital image forgery creation consumes lot of time and also requires high level skills. However with easy availability of photo editing and computer graphics software, the common persons may also edit or forge the digital images for various purposes. It is therefore essential and necessary to develop techniques and tools to detect the digital image forgeries.

1.2 Motivations
In recent years, many tampered images made international news headlines. The figure1 shows such news headlines and figure 2 show some fake photos which appeared in a popular media including Science Journal and National Geography. Fake photo won the prestigious National geography award later detected, withdrawn and apologized by National Geography. The hoax photos of British soldiers abusing Iraqi prisoners were also reported lately in the news papers, All this clearly indicates the entry of digitally manipulated forged images into our daily life. Such fake images are affecting the science, business, law, politics, media.etc, tarnishing the common person's perceptions about images. As result the common man's belief in the images is changing.

Figure 1: Popular news media reports on fake pictures

Figure 2: Some fake pictures which appeared in popular media including National Geography

Science Journal and

The politicians have been noted to create a false history using forged images; Soviet Union, Stalin and Lenin had the enemies of the people removed from historic records, Todays digital era has raised serious questions over trustworthiness of photographs; though, the trustworthy photography plays a major role in news reporting, national intelligence services, criminal investigations, pictures for online business etc. The advent of low cost, high performance computing devices, sophisticated graphic and photo editing softwares allow common people to do complex image manipulations in a relatively easy way. It is essential to develop credible methods for detecting the digital image forgery. It is the theme of our work. The scientific and automatic methods for detecting image forgery are necessary and it has a growing demand in coming years.

1.3 Problem Formulation and scope

The image forgery problem is an ill posed [122,137]; inverse computer vision problem [108,109].This is a generic problem. The worldwide computer vision research trends are focusing on solving an ill posed and inverse problem, however ill posed and inverse problems are complex in nature doesnt have a unique solution. We treat the image forgery problem as a set of problems; we selected two sub problem of specific image forgeries, one is Copy Paste forgery other is image compositing they are entirely two different problems for computer vision researchers though they are logically connected with the purpose of application of a manipulation. In both forgery cases the image manipulators purpose may be same concealing the traces of forgery, changing the interpretation of the authentic image to gain an undue advantage. The goal of passive blind image forgery detection is to detect fake images, the most popularly used forgery creation techniques are Copy paste Forgery and image compositing their creation methodology shown in figure 3

In this dissertation, we detect the fake images created by copy paste forgery as detecting the similar regions in an image using SIFTS images local features. We detect the image compositing by image degradation problem identifying the image degraded parameters. In the first method we used supervised approach, later in the second method we used statistical methods based on low computer vision operations such as filtering mask, HSV and luminance levels. The photographic image of minimum size of 480X360 color image, the remaining scope is specific to particular solutions. Such specific scope is defined at appropriate sections.

Figure 3: Two popular forgery creation techniques

Approach
We approached the image forgery problem through a passive blind forgery detection problem. The image forgery problem will divided as separate sub problems based on the type of forgery. We considered two most commonly used image forgery problems namely, Copy Paste Forgery and Image Compositing 1.3.1 Image Authenticity

The core theme of passive blind image forgery detection is Image authenticity, the authentic image is one that witness to an actual event, time, place in totality the complete surrounding environment condition of picturesque even [130]. The image authenticity differentiates between the authentic original images with forged images. We consider image authenticity as a result of three processes such a Scene authenticity that is the authentic image holds the consistent features surrounding environment such an illumination, shade, texture and other features of the scene, second one is an image regenerative process, the unique pattern resulting in digital images due to the camera parameter and optical image formation process. 1.3.2 Methods

We first considered the Copy Paste image forgery This sub- problem is treated as a region duplication detection, first we find out the similar regions using the SIFT image features ( see chapter 4), later phase based on some criteria thats the cutoff threshold and specific features such as invariant features, we made a decision about the image forgery. The second sub -problem is Image compositing, in this approach we made few realistic assumptions such as degradation in image quality parameters in the composite image, since two or more images from different sources and taken at different condition are skillfully stitched. First method (See Chapter 5) we used a supervised method in which we used SVM classifier, the classification decision is based on image quality features and moment features.

For solving image compositing, we used the second method based on statistical methods, (See Chapter 6) here we used the low level image processing task such as luminance level , HSV color space model based color components differences, lastly we used customized filtering masks.

1.4 Thesis organization and Contribution

1.4.1 Thesis Organization

Figure 4: Thesis Structure

1.4.2

Overview of the Thesis

This section summarizes the contribution of entire dissertation work. The complete dissertation work is structured as follows: Chapter 1: Introduction This chapter introduces the problem of image forgery. Image forgeries are a serious problem due to technical advancements in photo editing and computer graphics softwares , easy availability of digital cameras and widespread penetration of internet and photo sharing and distribution by social networking working as a catalyst to the above problem. The relevancy of problems its social impact is clearly explained and motivation for research and approaches are clearly stated in this chapter, the chapter ends with thesis organization with our contributions Chapter 2: Theoretical Background This chapter contains the relevant materials for the image forensics research and particularly for image forgery detection. This chapter starts with a brief history of image forgery, image forgeries in the digital era, types of image forgeries; forgery life cycle .Later part contains the formation of image, camera model, and camera parameters. A review of relevant materials and related research works Chapter 3: Literature Review This chapter briefly describes the origin of image forgery. The landmark contribution by various researchers , the research path from its originator professor Hany Farid to till date, sudden increased interest of researchers, various resources and challenges ahead for future researchers.

10

Chapter 4: Copy Paste Forgery Detection This chapter comprising of local shift invariant features, our proposed method based on SIFT transform, supported by results and conclusions. Chapter 5: Image Compositing detection using image Qualities This chapter briefly describes the image quality features and the moment features, supervised approach to detect the image forgery based on image quality features with their results and conclusion. Chapter 6: Image Compositing detection using image processing methods This chapter contains three different image processing methods such as luminance level tests, HSV Space color model and customized filter mask, applicability of those methods and results. Chapter 7: Conclusions and a discussion of the future work. Finally, the chapter 7 summarizes the research contributions, the conclusion derived from the research, future directions of research in the digital image forensics.

1.4.3

Contributions

In this thesis, we present three new techniques to detect the image forgery based entirely new approaches. These techniques exploit the regularities in images that arise from the world plane (view), device plane (view) and image plane (view) and irregularities arises due to the image forgery mechanism as shown in figure 9 and 19. For each approach, we describe the applicable conditions for which image forgery to be detected successfully. We modeled the property being analyzed; provide a method for estimating property from a single image. We demonstrate results on real images and forged images. The images for the purpose are collected and used from reliable internet resources, and few images are taken from NIKON S3000 digital camera, the forged image database is created using adobe Photoshop and collected from internet

11

resources. In this context, we present the approaches primarily based on following cues: 1. Image invariant features. 2. Image quality Features 3. Image color Properties In general, the forgeries are created by the addition of objects in photographs from other photographs or deletion of objects from a photograph. It is difficult to match the world view conditions such as lighting, illumination, texture, surface properties, projective and perspective geometry of scans of different photographs those may be taken from different devices under different conditions. However forger may skillfully conceal the forgery effects in a forged image. In general , we assumed the authentic image have unique consistent properties those may be the result of the world view condition , device properties and conditions finally reflected in a digital image. We also assume that forgery leaves some traces such as inconsistencies, irregularities; disturbances in an image. We exploited such properties, identified and quantitatively measured and analyzed.

12

2. Theoretical Background 1.5 Introduction

The pictures always have drawn an attention of humans. Layman, scholars, labors, business persons, entrepreneurs, and every person use pictures for one or other purpose; it may vary from personal use to official purposes. The pictures always play a vital role in society, once picture are considered a credible evidences. The revolution in information storing, processing, communication using information technology , the rapid developments in internet, sophisticated high processing software and hardware devices, easy availability of digital cameras and other digital image acquisition devices, graphic and image processing softwares, the internet, social networking, all of suddenly changed the perception of humans towards a picture. New pictures are omnipresent, newspapers, journals, advertisements, books and training materials, in one word picture are everywhere; billions of images circulating in a day through internet. The reliability of images is seriously questioned for their origin and contents. The Digital image forensics a brand new research field is originated. This aims to validate the authenticity of image origin and its contents. It provides a relevant technical solution

1.6

Forgeries

The photographs are always considered as trustworthy evidence although Seeing is no longer believing always correct. Thou common persons doesnt have any reliable resources of verifying the authenticity of digital images. 1.6.1 History of Image Forgery

The history reveals that the modern digital image manipulation inspired by the dark room photo editing skills, combinations print was the earliest image forgery that requires

13

darkroom skill where the multiple fragment of an image are printed on a single photograph, Figure 5 shows the first reported fake image by Oscar G. Reijlanders

Figure 5: First officially reported image forgery

The Two ways of Life (1857) this uses the earliest popular combination print, which had used up to 30 images In the early twentieth century, cut and paste image fragments and photomontages are used for political satire, surreal art and other various purposes. Photomontage and combination print require technical skill and time consuming, often visually detectable.

14

1.6.2

Image Forgery in the Digital Era

Digital images are pliability to manipulation due to its inherent characteristics of discrete representation. The Wall street Journal reports during 1989 that 10% of the published color photographs in the United States are digitally altered and retouched, Digital images begin to gain popularity during the decade of 1990. The following few photographs are very popular manipulated images circulated in popular journals and the internet over the last two decades. Figure 6,7 and 8 shows recent popular forged images in the last decade.

Figure 6: Most popularly circulated fake images in the last decade

15

Figure 7: Most popularly circulated fake images in the last decade

16

Figure 8: Most popularly circulated fake images in the last decade

17

1.6.3

Image Forgery Life Cycle

The behavior of image forgery creation methods and image forgery detection algorithms reveals the strong interrelationship between creation and detection mechanism. And both the process is dependent and complementary to each other. Development in one filed is contributing knowledge and technology of the other. The given below figure 9 shows the life cycle of an image forgery process.

18

Figure 9: Image forgery cycle

The image forgery detection is a live problem. Image forgeries cant be eliminated totally, as long as skills and technology and other resources available, forgery will be 19

there. But there is an essential need for constraining development of forgery creation by strong legal process, high sophisticated detection process, and creation of awareness. 1.6.4 Image Forgery Types

The tampering of images is neither new nor recent. History provides evidence of tampering of images. Many rulers, politicians used the forged images either for a degrading image of their enemies or boosting the self image. The given figure 10 shows such an example.

20

Figure 10: Image forgeries in history

The images are tampered by adjusting color and contrast, airbrushing, retouching, dodging and burning, Recent years, affordable high resolution digital camera, sophisticated computer graphic and image processing softwares such as GIMP, adobe Photoshop, CorelDraw, allows the common user to produce a complex and credible forgeries. There are many 21

unaccountable ways of manipulating and tampering digital images. In order to develop a detection technique, it is essential to know the forgery creation techniques. We are presenting most commonly reported forgery types [104,128,138]

1.6.4.1 Compositing:

The compositing is a popular image forgery technique, in this technique; two or more than two image contents are combined to gather to form a new single image. Generally one image will be taken as a base image. The other images contents skillfully overlaid on a base image, the background and adjacent pixels are carefully retouched. The figure 11 and 12 show the image compositing examples. If the whole process is carried out carefully, matches in a composting are visually not perceptible. The composting is also known as image splicing.

22

Figure 11: Image Composite

23

Figure 12: Another example for image compositing

24

Copy Paste:
The copy paste forgery is called by different names such a copy move forgery, image cloning. This forgery is created similar to compositing only difference is here only single image will be used. The undesired image contents such a person (s) or an object (s) is concealed by replacing the similar image person (s) / object (s) or background. The figure 13 shows copy paste forgery

Figure 13: Example for Copy paste forgery

25

1.6.4.2 Computer generated

In the computer generated image, the entire image is created by computer and graphics software like Maya, Alias, 3DMax, Discrete, etc using an algorithm; generally such images are used in the advertising industry. The good quality computer generated images are really hard to distinguish from real photographic images by visual inspection. Figure 14 shows the difficulties in identifying computer generated images from camera images using bare visualization

Figure 14: Can you distinguish real images from computer generated images?

Real images : 1 & 4 Rest: CG

26

Figure 2. 44 Can you identify real images?

1.6.4.3 Retouching

This method of manipulation involves the many subtasks such an airbrushing, smudging, blurring, painting and copy pasting of the regions within same images, This is a broad class of localizes manipulation. Generally retouching used in personal photos to hide some finer details like age, marks. Etc purposes are looking younger and smarter. Figure 15 shows such an example

Figure 15: Example for retouching

27

1.6.4.4 Enhancing This is a global image manipulation technique. In this method manipulation is achieved by adjusting a color, contrast, blurring, sharpening. However basically the contents of images are not altered by interpretation of images are changed by above tasks. The figure16 shows an example for image Enhancing

Figure 16: Example for image Enhancing

28

1.6.4.5 Morphed Morphing is a special effect generally used in motion pictures and animations that changes the one image to other images through a seamless transition. Most often it is used in fantasy or surreal sequences in which one person turning into another by means of technological means. This is achieved by cross fading techniques in films; today morphing images are created by computer softwares using complex algorithms. The given figure shows such a transition. The figure 17 shows the example of the morphed picture

Figure 17: Example for morphing

29

1.6.4.6 Painted This type of manipulation is created by photo editing or drawing software like adobe Photoshop, CorelDraw. This is similar to artistic painting on a traditional canvas. Here the real challenge of creating a good painting is a person with good painting ability and knowledge of such software is essential.

1.7 Forgery Detection Approaches

The Image forgery detection researchers used two entirely different approaches to ensure the authenticity and integrity of the image. Those approaches are first one is active methods second one is passive or blind methods. Figure 18 shows the overview of various approaches

30

Figure 18: Image forgery detection methods

1.7.1

Active Methods

In Active methods a certain information or data is embedded in the image either during the image acquisition process or imaging device itself. The same embedded information or data is verified to ensure the authenticity of the image. The digital watermark, digital signatures, digital fingerprints is examples of active methods. The problem with active method is the requirement of special hardware or software to embed the data or information into an image before distribution. However, most of the image circulating over the internet for example images on photo sharing websites, newspaper and journals are not watermarked. This may be due to the fact that the requirements of special hardware or software, such facility are not available or unpopular among contributor, distributor or circulator, the image contributors over the 31

internet or newspaper or journal may not bother to watermark their images. This shows the inability of applying active methods for detecting digital image forgery. 1.7.2 Passive Methods

The passive methods dont depend on pre-embedded or pre-extracted data or information such a watermark or holograms. These methods entirely depend on the image function. Passive methods basically assume that images posses a certain smoothness, consistency, continuity, regularity, and/or periodicity and exhibits certain patterns, all the above features are specific to every image those may come from a variety of resources like world, devices, or image itself. The figure 19 shows the influence of various parameters on an image and those parameters leaves specific inherent properties on images.

Figure 19: Parameters influencing on image formation

The passive methods assume that the images inherent properties such a smooth, continuity, consistency, regularity, and /or periodicity are disturbed in forged images.

32

These methods measures, quantifies and locating inherent properties such as smoothness, continuity, consistency, regularity, and /or periodicity of an image and detect the differences in the measurements. The decision of forgery is made either automatic or semiautomatic with human interpretation depending on specific methods The Passive methods are broadly classified as a supervised method and unsupervised methods. Recently, the passive (blind) forgery detection methods have begun to gain popularity among researchers worldwide. In order effectively to design a passive digital image forgery detection system, it is necessary quantitatively to characterize the inherent properties of a physical imaging system, scene, and image display. Basically the procedure is to model the image forgery detection model (IFD) that provides the knowledge about forgeries.

1.8 Image Formation

Before, we can intelligently analyze an image, we need to understand the geometry of a scene, image formation processes and image acquisition devices such a camera. The digital images are made up of a discrete color or intensity values. The real world three dimensional (3D) geometric features are projected as a two dimensional (2D) features in an image. However, images are not made up of two dimensional (2D) features. Instead, there is a sequence of bits representing a discrete color or intensity values. Where these values do comes from? How do they relate to the lighting in the environment, surface properties, geometry, camera optics and sensor properties? The figure 20 and 21 shows the components influencing the photometric image and their interrelationship.

33

Figure 20: Components of image formation

34

Figure 21: simple image formation model

1.8.1

Lighting

Light is essential for the formation of a photometric image, the scene must be illuminated by light sources; light sources are classified as a point sources, area sources. The point sources are generally originated at infinity (sun for outdoor scenes) or single place (single bulb). Point light source has location, intensity and color spectrum. The intensity of a light from a point source obeys the general inverse square law. The modeling an area light sources are more complicated. For example the simplest florescent ceiling light is modeled as a finite rectangular area emitting a light equally in all directions. The more complex light distribution such an illumination incident on outside objects can be modeled using environment maps [61] originally referred as reflection map [14] 1.8.2 Reflectance and Shading

As per snells law, when light hits the objects surface, the part of light may be reflected, transmitted and absorbed by the object surface. As shown in figure 22 by snells law.

35

Figure 22: Snell's Law

Many models are developed illustrating the interaction. Diffuse, specular corresponding to scenes. 1.8.3 Optics

and phong

shading models are popular ones. They also used to compute the global illuminations

The light from a scene passes through

a camera lens before reaching a sensor. For

many applications, the lens modeled as a simple ideal pinhole that project all the rays from the common center of projection as shown in figure 23 by simple pinhole camera model.

36

Figure 23: Simple pinhole camera model

For image forensics, it is necessary to deal with issues such as focus, exposure and vignetting and aberrations. We need to use a more sophisticated model using a principle of optics the figure 24 shows such a basic thin lens model.

Figure 24: Thin lens camera model

37

1.9

Camera Parameters

The camera plays a pivotal role in image forensics, the scenes & objects of the external world reflected as an image through the digital camera .The light sources from external world would reach imaging sensors, reflecting from one or more surfaces and passing through the optic lenses of camera and finally converted into digital (RGB) values that we considered as a digital image. The exposure, nonlinear mappings, Sampling & aliasing and, noise are the parameters produce a specific regularities and periodic patterns in an image. In manipulated images these regularities and patterns are disturbed. It is essential to understand the camera model. The figure 25 is a simple modern digital camera model developed by [63, 98, 166,]. Figure 25 shows the various processing phases of modern digital camera including the typical post processing steps

Figure 25: camera sensing pipeline

38

The shutter speed, sampling pitch, fill factor, chip size, analog gain, sensor noise, and the resolution (and quality) of the analog to digital converter are the major parameters that affect and influence the performance and quality of pictures in a digital image. 1.9.1 Camera Settings

Shutter Speed: the motion blur, shallow depth in images are determined by the amount of light reaching a sensor chip, the amount of light required for formation of the image is controlled by shutter speed. It also determines the over exposed and under exposed images. Sampling pitch: The physical distance between neighboring sensor chip cells of the imaging chip, smaller the pitch more prone to noise and less sensitive to incident light. Fill factor: Fill factor determines the active sensing area; higher fill factor are preferable as it reduce aliasing and capture more light. Chip size: Analog gain : The ISO setting on modern digital cameras provides the automatic gain control, higher gain means better performance in low lighting condition but amplifies the noise. ISO setting also provides control for shutter speed and aperture. Sensor noise: The fixed pattern noise, dark current noise, shot noise, amplifier noise and quantization noise [63,166].The noises are added from various resources at various processing phases. The total amount of noise present in images is a combination of all the above noises [62]. ADC resolution: The resolution of analog to digital conversion (ADC) process and its noise level. These parameters produce specific patterns in images. The patterns are estimated by calibrating the camera by taking the repeated shots of the same scene and plotting the estimated noise as a function of brightness.

1.9.2

Sampling and aliasing:

The aliasing is produced during the image acquisition process if Shannon sampling theorem is not satisfied during sampling. Shannons Sampling Theorem shows that the minimum sampling [135] rate required to reconstruct a signal from its instantaneous samples must be at least twice the highest frequency, fs 2fmax The maximum frequency in a signal is known as the Nyquist frequency and the inverse of the minimum sampling frequency rs = 1/fs is known as the Nyquist rate. 39

1.10 Discussion
Apart from the above basics, the problem of image forgery detection is a part of passive image forensics; this is a new research area leading to a new branch called as a computational photography or photogrammetry. The computational photography is interdisciplinary in nature; it needs the knowledge of computer vision, computer graphics, and optics.

40

3. Literature Review

1.11 Introduction to digital image forensics

There is extensive literature on Digital image forensics. Image forgery detection is a part of digital image forensics literature. Forgery detection problem in images is solved using the two approaches one is an Active methods other is passive blind approaches, former assumes the data or information is embedded into the image either during an image formation process or after imaging (during post processing), hence the special hardware or a software is required . The passive methods are natural that doesnt insert such additional data or information to the image; hence its a natural forgery detection method for common mans images. Active Methods Passive Blind Methods In this chapter, we examine the previous works of image forgery detection, active methods, passive blind methods, introduction of digital forensics, forgery detection of composite images and copy paste images. however the image forgery is considered in more broader sense by H.farid [38, 40], he objects the use of computer generated images for a pornography, later Ng [129,130] considered even computer generated images as fake images, hence we includes the introduction of digital image forensics, the goal of digital image forensics is to ensure the authenticity, consistency and integrity of images and their origination.

41

1.12 Active Forgery detection Methods

The watermarking belongs to the data hiding category, watermarking and digital signatures belongs to active forgery detection methods, these methods are also known as a conventional image authentication method [172,174,181]. The active methods can be divided into data hiding approaches [173,176,145] and digital signature approaches [90,100,150,167]. The active methods require an additional hardware or software to embed a data or information such a watermark or content signature .I imperceptible digital code to an original image at the time of recording or at the time of postprocessing of the image to be protected. The main theme of digital watermarking is to add a watermark such that the watermark is not visually traceable and shouldnt alter the image qualities for monitoring image manipulations. Fragile watermarks [121,134,145] are sensitive to any minor modifications, the robust watermarking techniques are semi fragile[93] and content based digital signature [90,150] these are resistant to modifications such a resizing and compressions. The security of the watermarking secrete key is an issue, due to above reason the Friedmans trustworthy digital camera an idea failed [45]

Figure 26: steps in active image forgery detection

42

1.13 Passive Blind Forgery detection Methods

The passive blind image forgery detection methods are alternative methods for detecting the image forgeries [40]. These methods dont depend on prior information or a data. These methods entirely depend on the image function. These methods based on the assumption that the imaging scene environment, scene geometry, image formation process, image device, projective and perspective views, all together produce a unique, consistent, authentic image. All the above parameters leave specific and consistent patterns in images, while the forgeries in the images alter the specific consistent patterns of an authentic image. There are various works based on the scene authenticity are reported [74, 75]. The figure 27 shows the parameters of the scene authenticity.

Figure 27: Scene authenticity parameters

43

The source illumination, surface properties of objects, geometry of objects, shadows, inter-reflections etc. influences the formation of image and form a consistent pattern in an image. In [65, 66, 97], the consistency of image fragments is verified using the estimation camera response function. The interpolation patterns are created by demosaicing, the forgeries disturbs the interpolation patterns, in [140], Demosaicing patterns are used detect the forgeries. 1.13.1 Copy paste Forgery detection In copy paste forgery, the part of image objects or background is added to the other part (s) of the same image or the part of image object or background deleted from the same image, the entire forgery is created using a single image. This type of image forgery is also called as copy move forgery, image cloning, and region duplication. The figure 28 shows the model of copy paste image forgery creation.

44

Figure 28: Copy paste forgery creation model

In copy paste forgery, the cloned regions may be of any shape and size and located anywhere on the image. In this forgery, the duplicated region is a portion of the same image; hence it is hard to visualize the forgery through visual inspection, the color, texture, noise will be same as of image since the duplicated region is part of an image. The best method of detecting the copy paste forgery is exhaustive search; however the exhaustive search is computationally very expensive and practically impossible. Two computationally efficient duplicate region detection algorithms are proposed in [51, 87], 45

the other methods are [103,111,138]. The second approach is autocorrelation properties proposed by Fridrich et al. [51], however, this method is effective when the duplicate regions are of large size. In block matching methods [51,103,111], the first image is divided into small blocks; these blocks are sorted in some order. The features of blocks are extracted and compared for similarities. The differences in the methods lie in the selection of features, method of feature extraction, methods of searching a similar block The figure 29 shows the copy paste forgery detection frameworks

Figure 29: Copy paste forgery detection frameworks

46

The copy paste forgery detection majorly comprising of following phases Image decomposition Feature extraction Duplicate matching of blocks Forgery decision 1.13.1.1 Image decomposition and block representation The test image is divided into a small image blocks, the effectiveness the forgery detection depends on size of image blocks, smaller size image blocks give more effective forgery detection methods. Fixed size image blocks are used by [51,94], the overlapping image blocks are used by Jessica Fridrich et al [51], the [51] uses the discrete cosine transforms (DCT) for image decompositions while wavelet transformed used in [87,123,183] 1.13.1.2 Feature extraction The feature extraction is a significant step in the duplicate region detection since the duplicated region is part of the same image hence it has very similar image characteristics such a color, texture, noise. etc. The biggest challenge is selection of features that differentiates the blocks or choosing image invariant features in a block and matching the invariant features in duplicate image blocks. The effectiveness of copy paste forgery depends on selection of features. 1.13.1.3 Duplicate matching of blocks Once the image is decomposed into a small image blocks later the features of each blocks are extracted. Based on feature extracted image blocks are matched for similarities, generally the clusters of image blocks are formed based on nearest matching blocks. Dybala et. al. [32] uses nearest neighbor search, while radix sort is used by [94]. Lexicographical sorting is used in [38].

47

1.13.1.4 Forgery decision

The matching of similar blocks is not sufficient to determine the copy paste forgery. There is the essentiality of some more condition to take decisive decisions regarding image forgery. Sevinc Bayram et al [16],Brandon Dybala et al [32] proposed a clone detector former used the rotation and scale invariant FourierMellin transform of image blocks, later method uses the filtering operation and nearest neighbor search operation for detecting the duplicate regions. Jessica Fridrich et al [51] used the discrete cosine transform (DCT) of overlapping image blocking and lexicographical sorting to detect the copy move forgery, while radix sort is used by HweiJ.Lin et al [94] to detect the copy paste forgery. The wavelet transforms are used to detect the copy move forgery in [87,123,183], however Guohui Li.et.al [87] uses single value decomposition (SVD) for image block representation while A.N.Myna [123] uses log polar coordinates in association with wavelet to detect copy move forgery. Jing Zhang et al [183] use the phase correlation with wavelet transform to find duplicate regions. Weiqi Luo. et. al [103] uses the intensity based character features to expose the copy move image forgery while Aaron Langille and Minglun Gong [81] use the intensity patterns for searching similar blocks using kdtree. Babak Mahdian and StanislavSaic [111] use the moment invariants, kd tree and principal component analysis (PCA) analysis for detecting the nearest duplicate regions. Although most of these methods require high computational time, high false positive and false negative ratios few methods require as human interpretations of results.

48

1.13.2 Image composite forgery detection Image compositing is most popular image forgery. The figure 30 shows the creation of image compositing. The photo compositing is the result of cutting and joining a two or more photographs with seamless transition without leaving any visual clues about the joining from other photographs.

Figure 30: Composite image creation model

The image compositing is also known as a photomontage and image splicing. The image compositing detection assumes that the image scene authenticity properties and conditions such as illuminations, object surface properties, shadow, noise, interreflections. Perspective and projective views etc. are rare and difficult match in a composite image. The image composite detection techniques are able to detect the above inconsistent properties in different parts of the same images, the image edges, boundaries and colors, and image qualities may be affected by image compositing. The image splicing is analyzed in a general way [31, 56,168,186], using supervised approaches [31,168], and using statistical methods [56,186] as seen in the literature.

49

The pixel continuity is disturbed due to cut and joining of other images. This feature is exposed to detect the image splicing using support vector machine by Dong et al [31] . He further demonstrates that the correlation and coherency of pixels are not continued at the stitching points. In [65,66], proposed methods uses the camera response function to determine an image splicing , in [59,60], Yu-Feng Hsu and Shih-FuChang estimates the camera response function using geometric invariants, while [65,66] uses edge based profile. Zhouchen Lin et al [96] uses the inverse camera response function (CRF) using the analyzing the edges in different patches and verifies the consistency. Hany Farid and Mary J.Bravo [42] propose the several computational methods for detecting the inconsistencies in shadows and reflections. While Wei Zhang et al [186] detect the image compositing using the geometric and photometric constraints on shadows . Sandeep Gholap and P.K.Bora [56] estimate the illuminant color to expose the image splicing. MicahK.Johnson and HanyFarid [75] estimated the direction of light from specular heighlights that appears on the image eye to detect the image compositing by demonstrating the scene authenticity are differs in composite images.

50

4. Copy Paste Forgery Detection 1.14 Introduction

The copy paste image forgery is most common image forgery after image splicing. In copy paste image forgery, the forgerys main purpose is to conceal either a person (s) or object (s) in the image. The aim of forgery may be a gain subdues advantage by boosting a self image as shown in figure 31& 32 either by adding desired or eliminating the undesired person (s) or an object (s).

Figure 31: Original image

Figure 32: Copy paste forged image

The copy paste forgery, the undesired person(s) or an object(s) in a scene are replaced by the virtually identical region from the same image. In this chapter, we present a technique that effectively and efficiently detects duplicate regions in an image. We show the efficacy of our technique of a credible forged images collected from various internet resources and our own image forgery datasets created by adobe Photoshop. We also show the robustness, sensitivity to add noise and lossy compression schemes. The experimental result shows the correct results of copy paste forgery as long as cloned regions are not rotated and scaled.

1.15 Related Literature

The main task in the detection of copy paste forgery is to find duplicate regions in an image without prior knowledge of shape and location. The simplest approach is to compare exhaustively every possible pair of image regions. However, such an approach is computationally expensive and practically improbable. 51

Recently, several researchers purposed methods for copy move forgery detection. Farid and Popescu [138] proposed duplicate image region detection by decomposing the image using principal component analysis(PCA), detection of duplicated region is achieved by lexicographical sorting. Mahdian and Saic [111] presented a copy move forgery detection based on blur moment invariants. Discrete Cosine transforms (DCT) and quantization techniques are used to detect copy move forgery by Fridrich [51]. Popescu,and Farid [138] adopted PCA for representing a small fixed size block and successfully uncover a copy move forgery

1.16 SIFT Features

Image consist of many objects, each objects many features, very few features of a objects are invariants to changes in image scale and rotation, these features are used for robust matching .The SIFT image features ,features of an objects that are not affected by many of the objects transforms such as rotation and scaling. These features are not influenced by the affine transform distortion, addition of noise, changes in illumination and 3D View points.

1.17 SIFT Features based forgery detection system

1.17.1 The Proposed method Our approach is based on David Lowes Scale invariant feature transform (SIFT) algorithm to extract the robust local image features. Those features are used to discover the copied regions in the images and any geometrical transformation if applied. However the copied regions of the image will the quite similar to that of the original region. We will match the SIFT features of the copied regions and original regions to reveal the tampering in the regions. The figure 33 shows the schematic diagram of our proposed system

52

Figure 33: Schematic Diagram of proposed model

Our forgery detection system works out in four steps Step 1: SIFT feature extraction Step2: Matching SIFT features Key points Step 3: Cluster a matched key point Step 4 : Forgery decision Step 5: Estimate the geometric deviations if forgery occurred 1.17.2 SIFT Feature Extraction Let P={p1,p2,p3, pn} --- set of key points Q={q1,q2,qn}------- Sift descriptors, of a given test image. The feature vector qi is matched with each key points to find a similar local patch in a test image. The nearest neighbor from the (n-1) key points of the image gives a best candidate match. That is a minimum Euclidean distance of a key point feature in SIFT space. The Euclidean distance is not sufficient to decide the duplicated region in the image due to high dimensional feature space. We use new parameter Distance ratio defined as given below.

53

For the sake of simplicity, The given a key point and T be a global threshold Similarity vector D= {d1, d2, d3 dn-1} ------------sorted Euclidean distances wrt (with respect to) other descriptors. We decide the regions are duplicating only if to 1 The above matching procedure fails to handle multiple key points matching; this reflects failure to find a multiple forged region in a single image. This is a limitation of our algorithm. 1.17.3 Clustering The tree structure is created by performing a hierarchical clustering on the xy coordinates of the matched key point. The algorithm works as below 1 Assign each key point to a cluster 2 Calculate reciprocal distances amongest the clusters 3 Find the closest pair of clusters 4 Merge the clusters iteratively so that finally single cluster remains. The final merging decision is based on threshold & Linkage methods. 1.17.3.1 Linkage Methods In the computer science literature several linkage methods exists, we use Single and Centroid linkage methods and empirically find out best cutoff threshold Let P and Q are given two clusters contain np and nq objects, where Xpi and Xpj 1.17.3.2 Single linkage We use the smallest Euclidean distance between objects in the two clusters where T in the range of 0

54

1.17.3.3 Centroid Linkage We use The Euclidean distance between the centroid of the two clusters

Inconsistency Coefficient (IC) that characterize the clustering operation. That is the average distance amongst clusters and doesnt allow for clusters to join together.

1.17.4 Forgery Decision The threshold Ta affects the performance of copy paste forgery, we empirically find out the best Threshold for copy paste forgery. The method decides the forgery if two or more clusters with at least three key points matched.

1.18 Experiment and Results

We carried out the experiments on our dataset, we build our own data set by collecting an authentic and copy forged image regions from authentic internet resources and we created few forged images using a Photoshop. The data set contains 57 images and same number of copy paste forged image. We have various size of images the minimum size is of 50K byte the minimum forged region is approximately 1% of the whole image. 55 authentic

Table 9: Results of SIFT method

Ta

single True positive % 39.53 68.71 85.92 94.18 97.32 95.45 82.56 63.63 True negative % 2.13 5.32 8.12 8.13 8.14 7.17 6.37 4.45

Centroid True Positive % 22.61 57.34 88.22 93.43 96.57 95.69 90.73 83.93 True negative % 1.7 3.53 6.78 7.82 7.95 7.67 7.32 4.93

0.7 0.9 1.1 1.3 1.5 1.7 1.9 2.1

56

Figure 34: Results of SIFT

57

5. Image composite Detection using supervised approach 1.19 Introduction

The image compositing is most common image forgery. The forger uses two or more images objects or background to form a single composite image. The composite image changes the interpretation and meaning of the authentic images.

1.20 Proposed Method

Our image forgery detection proposed model based on image qualities and Markov process based features the framework of the model is shown in fig 35

Figure 35: Proposed model based on image features

58

5.2.1Image qualities: In computer vision research there is a rich set of literature available on
image qualities. We selected an image quality feature based on a study of Avcibas. In[2,3] ,Aviabas present a large set of image quality features, which are sensitive to discriminative to based few features of forgeries such as compression, watermarking, blurring and distortions. We selected such an eighteen feature which is sensitive to image forgery operations .Those features are Mean Errors (D1-D4), Correlation (C1-C5), Spectral Errors (S1-S5), HSV Norms (H1-H2) a.) Mean error features : Mean absolute error D1, mean square error D2, modified Infinity norm D3, L*a*b perceptual error D4 b).Normalized cross-correlation C1, image fidelity C2,Czenakowski correlation C3, mean angle similarity C4, mean angle-magnitude similarity C5. c).Pratt edge measure E1, edge stability measure E2. Spectral phase error S1, spectral phase-magnitude error S2, block spectral magnitude error S3, block spectral phase error S4, block spectral phase-magnitude error S5. HVS absolute norm H1, HVS L2 norm H2. e). d)

5.2.2Moment based Features: The forging operation assumed to be disturbs the continuity,
smoothness, regular pattern, smoothness, consistency and periodicity of pixel correlations Our moment based feature extraction procedure is shown in fig 36

59

Figure 36: Moment Extraction Procedure

Multi- block discrete cosine transforms (MBDCT): the block discrete cosine transforms coefficient are able to reflect the disturbances (changes) in the local frequency distributions. We use multiblock discrete cosine transform to pick up local frequency disturbances effectively. The 2D block DCT coefficients are represented by
` a 2 n @1 X X F s,t = ffffA n x=0
n @1 y=0

` a ` a 2y = 1 ` a 2 x fffffffffffff +1 ffffff V x V y cos cos fffffffffffffffffff x,y

2n

2n

----1

Where f(x,y),x,y=0,1 denotes a nXn image

60

5.2.3 Prediction Error 2D Array: This is used for dimension reduction purpose. It also
serves the additional purpose of enhancing the statistical artifacts introduced by forgeries. The prediction context is shown in Fig 37

Figure 37 Prediction Context

We predict the pixel value x using the neighbouring pixel a,b and c, the prediction 2D array is represented as

x = sign x AL aM+L bM+L cM

f f f f

` aR L M L M L Mc

------------------------2

The prediction error 2D array can be expressed by

x = x @x = x @sign x AL aM+L bM+L cM

Discrete wavelet transforms

f f f f

S ` aR L M L M L M

-------------------------3

The wavelet transforms are suitable to pick up transient and localized changes in spatial and frequency domain. Moments and Marginal moments The 1D Characteristic function (CF) is the DFT of the first order histogram of each wavelet sub band. The absolute moments of 1D CF are defined by
k ff ff 2

f f f f f f f f f f f f f f f f ff f f f f f f f f f f f f f f f f f f f il 1 Ml = f ---------------------------4 k ff ff L ` a M M X L LH x i M
2

x. i

H x f f f f f f f f f f f f f f f f f f f i

` a

i=l

61

Where H(xi) is the CF component t frequency xi , Here K= total number of different values assumed by all of the coefficients in the sub-band under consideration, and L= order of the moment, which is an integer value The 2D characteristic function is the 2D DFT of the second order histogram of the image and a MBDCT coefficient 2D array. The second order histogram is defined as
b c

hd j1 , j 2 , =

N j , jf2 . , 1 ff f f f f f f f f f f f f f f f f f ff f f f f f f f f f f f f f f f ff f f f f f f f N T ,
b c

--------------------------------5

Where

@ the distance between two pixel,

@ Angle of line linking these two pixels with respect to the horizontal axis
Error! Bookmark not defined. while second is J2 Number of pixel pairs for which the first pixel value is J1

N T , -Total number of pixel pairs in the image with separation ( , ,).

Two Marginal moments of the 2D CF are given by
k f ff f 2 k ff f f 2

M X X u. iL L Hui ,v jM

M u ,j =

j = if i = 1 f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f fQ

c L X X L L Hui ,v j
j=i i=1

k f ff f 2

k f ff f 2

----------------------6

L M M X X v. j L L H ui ,v j M

K f ff fff 2

k ff ff 2

cM

M v ,j =

j i = 1 f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f ff f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f

X X | H ui ,v j |
j i=1

K ff ff ff 2

k ff ff 2

------------------------7

62

Where H(ui,vj)- 2D CF component at DFT frequency (ui,vj), l- order of moment, integer

1.21 Experiment and Results

We used the image dataset [24] of Columbia image splicing detection and evaluation data. Other images are collected from internet: 933 authentic images and 912 forged (spliced images ) and 67 forged as well as the same number of authentic images were collected from various resources from the internet. SVM classifier and matlab code [20] is used for randomly selected 65%,75% and 85% images from the above databases for training purposes and renaming are used for testing purposes. The results are shown in table 2

63

Table 10: Results

Data Set

Training Ratio %

True Positive %

True Negative %

Accuracy %

Columbia data Set

65

82.37

79.41

80.89

75 85

83.56 85.41

80.73 82.11

82.14 83.76

Our data Set

(Collected from authentic internet resources)

65

80.81

77.15

78.98

75 85

81.32 83.12

78.38 80.19

79.85 81.65

64

Figure 38: Correctly judged as authentic images

65

Figure 39: Correctly judged as composite images

Figure 40: misjudged as spliced images but these are authentic images

66

Figure 41: misjudged as real images but these are composite images

1.22 Algorithm
1.) Extract Image Quality Metrics (IQMs). a. Divide the test image into 4 regions. b. Extract features from every region. 2) Extract moment based features. a. Apply wavelet transform to this image and obtain all the sub-bands including the test image itself. b. Obtain histogram for each sub-band. c. Apply DFT to the histogram of each sub-band to obtain its characteristic function. d. Apply Eqn -- (4) to calculate moments.

67

e. Apply Eqn-- (3) to obtain a prediction - error 2-D array. f. Repeat a. to d. To obtain prediction-error 2-D array. g. Obtain 2-D histograms for the test image. h. Apply 2-D DFT to each 2-D histogram to obtain the 2-D characteristic function. i. Apply Eqn--(6) and Eqn-- (7) to calculate marginal moments. 3) Apply 22, 44, 88, nn BDCT to the given image. Round those BDCT coefficients to nearest integers, and repeat step 2). 4) Repeat step 1) to 3) to obtain features of all images. 5) Obtain the best parameter of C and g which will be used in training. 6) Train a part of images using SVM and obtain the SVM model. 7) Predict the remaining images using an SVM model.

Conclusions and Discussions

Sensitive quality features [2,3] and Markov process model were utilized to detect forgery in image data using histogram moments. Experiments were conducted on famous databases providing and authentic and forged images to find true positive and true negative results duly these trainings of data for various ratios like %65, 75% and 85% respectively .The results obtained are superior to so far used techniques.

68

6. Image composite detection using Image Processing Methods

1.23 Introduction
The image forgery can be identified by the specific patterns relating to image attributes which disturbed by the forgery operations. Particularly image composting is created by the two or more images sources, naturally all the different images are taken from the different devices and at different world view conditions. The host image conditions are expected to reflect in image portions of the altered images. The abrupt and unnatural luminance levels, colors and edges are able to detect the image forgeries. In this chapter we used such methods to detect the image forgeries.

1.24 Related Works

The researchers used a variety of methods to detect the image forgeries, however very few researchers such as Lukas used image processing methods such a filtering using standard Roberts, sobels and prewitts mask identify the image forgeries, we extend the Lukas concept by designing a customized mask. In addition above methods we used luminance level testing HSV for detecting the image forgery.

1.25 Forgery Detection Based on Image processing Methods

1.25.1 Analysis of Luminance levels The luminance of an image is the measurement of perceived level of brightness [147]. In image compositing the two or more images from a different camera and different lighting conditions. Intuitively, there is some sort of discrepancy occurs in two image portions and adjoin places. The lighting conditions on the composite difference defined in a copied section of borrowed images, it may not be visually detectable. The figure 1 shows the differences in luminance levels.

69

Figure 42: Example of changes in luminance levels

Figure1 (a) is original figure1 (b) with image with a low luminance value, figure1 (c) with highest luminance value. 1.25.2 Analysis using HSV color space model The color components such as hue, saturation and brightness (value) are disturbed at point of compositing in an image and HSV values of the copied portion of the image will be different than the authentic image. The forgery can be detected by a thorough analysis of the HSV color model. In color theory, hue refers to pure color one without a tint or shade, one element in the color wheel [33].Technically (in the CIECAM02 model), as "the degree to which a stimulus can be described as similar to or different from stimuli that are described as red, green, blue, and yellow,"[33] (the unique hues) The fully saturated color is the purest version of that color. The primary colors (RGB) are true, they are fully saturated [33] as saturation increases, and colors appear more "pure." As saturation decreases, colors appear more "washed-out."[33] The value also called as a brightness or lightness, it indicates the amount of white in the color. How dark or light of a color is referred as a value or brightness.

70

Figure 43: a) original in RGB format b) HSV The figure shows the difference in the HSV levels in the original and manipulated one. 1.25.3 Detection of tampering based on customized filtering masks Several edge detectors based on sobel and prewitts masks are used to detect the image forgery by Lukas[101,102]. Image filtering is used as an effective tool in image analysis and understanding problems. Image filtering is useful in image sharpening, smoothing, noise removal, edge detection and many pre and post processing operations. Filtering can be performed in the spatial domain and frequency domain. Filtering provides the alternate view of an image and therefore uncovers small anomalies the image tampering creates anomalies. The customization of masks information to uncover a forgery or providing further validations. In the spatial domain, a filter is defined by a mask or the kernel, which is a small array such a 3X3 ,5X5 etc, which is applied to each pixel and their neighbors within an image. 71 may prove vital

The center of the kernel is aligned with a current pixel and is square with an odd number (3,5,7 etc.) of elements in each dimension , this process is known as a convolution, This way of pixel grouping provides a way to show a trend in an image, such as brightness level across a particular area. The abstract representation pixel and their eight connected neighbors are shown below.

Xi-1,j-1 Xi,j-1 Xi+1,j-1

Xi-1,j Xi,j Xi+1,j

Xi-1,j+1 Xi,j+1 Xi+1,j+1

Let Xi, j be pixel at location i. j in image X, the remaining pixels are eight way connected neighbor. The pixel value, which is integer is extracted and manipulated with mask (Kernel) Let the mask be

72

The filters output pixel will be result of Convoluted Image pixels and the mask. The Filters result will be mathematically represented as

The result of the above operation emphasizes the trends in the image, particularly, abrupt pixel variability in the edges, more importantly at tampered regions. This is due to the averaged eight connected neighbor pixels that are used to determine new pixel value. To see more effective results an image is divided into blocks and convolution is performed on block bases. The smaller the block size will give better results.

1.26 Methods based on Luminance level

The properties of the authentic image depend on world view conditions at which photograph is taken such as the amount and intensity of light, camera model and camera representation of color and post processing operations performed by software or hardware particular camera model. In forged image above properties may slightly differ in and around tampered regions. This is because generally image composite forgery is created by two or more images, which may be taken from different camera models, and at different world view conditions. The tampered images may be visually may not be traceable , however the image analysis of luminance levels reveals the discrepancy and unnatural abrupt changes in their luminance levels at tampered regions and around the tampered places We convert the color or gray scale image to a binary image based on some configurable threshold. The binary image is one in which pixel has only two values either 0 (filled with black) or 1(filled with white). The pixel value gets either 0 or 1 based on threshold

73

The luminance level of threshold of image is represented by any fractional value between 0.0 to 1.0, the complete black is represented by 0.0 while full white is represented by 1.0, For example, if threshold value T l=0.75 then its pixel value is assigned to 0 if its luminance value is equal or less than 0.75,otherwise 1

= 1 otherwise The value of threshold Tl is to be decided on the image to be analyzed. However initially it is better to set the threshold to 0.5, that seems to be better approximate value. Later based on the result threshold value may be changed in either direction, The ultimate goal of setting the threshold is revealing the unnatural and abrupt luminance levels in an image. We conclude those abnormality will be the result of tampering. Figure 43 shows the original and its binary counterpart at threshold T l=0. 3. After continuing the thresholding, we got clear abnormality at threshold T l=0. 6.The original figure and its results with the suspected forgery region are clearly shown in figures 44 by encircling. More results with their originals are enclosed in Annexure

Figure 44: Test image

74

Figure 45: luminance level at threshold 0.3

Figure 46: Inverted result of figure 4 with threshold at 0.6

1.27 Methods based on HSV color space model

The standard color images are represented in RGB color space, in which each pixel value is represented by three values by the amount of red, green and blue, In HSV model hue, saturation and value color space model, the amount of hue(tint), saturation(shade) and value(luminance- value of brightness) represents each color

75

The analysis of color images of HSV models able to reveal the abnormalities and abrupt changes in images. The figure shows the forged image in RGB format and its HSV color space showing the abnormalities that portion is encircled showing the possible forgery in the image. The more results and their originals are attached at Annexure

1.28 Methods based on customized masks

Figure 47: Test image

Figure 48: Forgery detected in HSV

76

1.29 Methods based on customized filter masks

Lukas analyzed the forgery detection based on filters [101]. He used the standard mask such as Sobel. Robert, Prewitt and Marr masks. These methods has a limited application and ability of forgery detection, However these masks provide the foundation for image filtering. We are extending the concept of filtering based on customized mask to tailor the filtering that is resulting in a better revelation of forgery traces. The concept of the kernel or filter mask and convolution are explained in section 6.3.3 We designed the customized masks by empirically, for example the first customized mask is

Customized Mask

The Customized Mask is designed in such a way that the sum of all connected neighbors weight is equal to the centers pixels weight. This mask effectively filters out statistically similar areas in the image and only shows dissimilar regions or areas. These different dissimilar results are aroused from prominent edges or maybe from the forged regions of an image. The results provide the alternate way of viewing results on screen or in printouts. The results are dark, sometimes inverted results reveals the abnormalities. First we convert the image to gray scale image so that the statistical data of the original authentic image doesnt affect by convolution with a customized mask. The given figure shows the test image and its result from customized mask and the inverted result of customized result. The inverted result clearly shows the traces of forgery. The more results are enclosed in Annexure 77

Figure 49: Test image

Figure 50: Result of customized filter mask

78

Figure 51: inverted result of customized mask result

1.30 Experiments and Results

We are presenting three image processing methods; those are luminance level tests, HSV of image and filtering using a customized mask. In all the above methods we assume that the forgeries in images leave the traces in an image such as discrepancies in the consistent patterns in their color components, luminance levels, unsharp edges, sudden abrupt changes in image properties. However those abnormalities and inconsistencies may not to be visually detectable. The annexure I contains the test images, their attributes, these images are collected from results of all the methods are also shown in Annexure For the purpose of testing we use both authentically forged as well as original (unforged) images. Before the conduct of tests we know the tests images category. We confirmed 79 authentic resources from the worldwide web. Few images are taken from our NIKON s 3000 digital camera.The few

the status of images by the following list of experiments; the list of tests and their measurement metric of forgery are tabulated below.

Table 11: Image processing based methods for forgery detection

Test Luminance test

Detection Method Level Luminance level

Metric Measure perceived variability in luminance levels in test image. Identifies areas of luminance discrepancy, if any exist. Measure perceived variability in HSV levels in test image. Identities areas of color and/or light discrepancies, if any exist. Measure perceived object edges in test image. Identifies areas containing double or ghost edges, if any exists.

HSV level Test

HSV color space

Custom Method test

filtering

Custom Filtering

We performed the test on most commonly used image formats, The image formats used for tests and their respective tests are tabulated below

Table 12: summary of forgery detection methods and image formats used

Forgery detection method Luminance level Threshold HSV Color Space Customized Filter Mask

Grey image X

Scale BMP X X

RGB X X X

PNG X X X

JPEG X X X

80

1.30.1 Result analysis of luminance level test We tabulate the results of the luminance level test on ten forged images at different levels of forgeries and the results and interpretation of results. The test image and their attributes are given in annexure
Table 13: Summary of Luminance based test results

Forgery

Inconclusive Signs of Tampering

Possible Signs Definitive of Tampering Signs of is (further tests or Tampering analysis required) X X X

Test image 1 Test image 2 Test image 3 Test image 4 Test image 5 Test image 6 Test image 7 Test image 9 Test image 10

X X X X X X

The luminance level test fails to exhibit the sign of the forgeries in test images 4, 6 and 10 However the above images are forged, the luminance level fails to show the sign of forgery, the reason is the test image 4 ,6 and 10 are forged region are small in those images and contains the similar luminance levels as a background. The remaining test images shows the un- even or un- natural luminance level in nearby forged area. In those images the light sources in test images and host images are inconsistent resulting in differences in luminance levels. 1.30.2 Result Analysis of HSV test We tabulate the results of the HSV test on ten forged images at different levels of forgeries and the results and their interpretation of results. The test image and their attributes 81

Table 14: Summary of HSV based test results

Forgery

Inconclusive Signs of Tampering

Possible Signs Definitive of Tampering Signs of is (further tests or Tampering analysis required) X X X

Test image 1 Test image 2 Test image 3 Test image 4 Test image 5 Test image 6 Test image 7 Test image 9 Test image 10

X X X X X X

1.30.3 Result Analysis of Customized Filter We tabulate the results of customized filter tests on ten forged images at different levels of forgeries and the results and interpretation of results. The test image and their attributes

Table 15: Summary of Customized filter results

Forgery

Inconclusive Signs of Tampering

Possible Signs Definitive of Tampering Signs of is (further tests or Tampering analysis required) X

Test image 1 Test image 2 Test image 3

X X 82

Test image Test image Test image Test image

5 6 7 9

X X X X X

We are summarizing the results of all the three methods on both unforged and forged images. We used the 25 unforged images and 20 forged images. The results are tabulated as below.

Table 16: Summary of All three methods

Image dataset

Luminance level False test False negative 12 15

HSV Color space False positive 12 10 False negative 12 20

Customized mask False positive 4 5 False negative 8 15

Original

positive 8

images Forged 10 images

The result shows the significant differences due to the small set of test images. However the above methods are able to identify the forgeries if the forged region is significant in size and forgery operation is not retouched.

1.31 Results and Discussions

It is well known fact that there is no single technique to determine all the type of forgeries. This chapter introduced three simple and effective methods for detecting the image forgery, specifically for image compositing type of forgery. Each method focused on the image attributes, which are changed by the forgeries. For testing, we had small image dataset, however overall results were found satisfactory. It was realized that user of our technique will have to decide on his/her own discretion which method will suit which type of case. Need was also felt to extend the work for using technique of

83

fusion to obtain most optimum results utilizing all the three methods to complement each other.

84

7. Conclusion and Future Work 1.32 Summary

In this thesis, we are presenting techniques for detecting the forged images; however, in this thesis, we analyzed the area of image forensics relating to the image tamper detection techniques. We reviewed the literature of image forensics. The current state of the art of digital forensics majorly identifies the image forgery detection, image source identification, natural image or computer generated images and linage (flow) mapping of digital images. We have clearly explained the theoretical framework for blind image forgery detection so that any interested researcher can build up different techniques based on the above framework. The thesis broader goal is to decide the image forgery based on passive blind methods with a single image. We took two major types image forgeries - first one is copy paste forgery, second one is image compositing. We use a supervised method as well as statistical methods for determining the image compositing.

1.33 Conclusions of Research

In this thesis, we focused on two most infamous types of images forgeries .First one is a copy paste image forgery. Second one is image compositing. In the copy paste image forgery method, we developed an algorithm based on David lows SIFT (Shift invariant feature transformed) image features. We applied the lows concept for the detecting the image forgery. The key concept behind our algorithm is that local SIFT features dont change their properties even after completion of copy paste operation. In copy paste image forgery, the few objects or persons are removed from images, the removed image regions are filled by the other image regions of the same image. This is a typical problem of region similarity. We are able to detect the similar image regions based on local SIFT features. Later we made a forgery decision based on lexicographical sorting and key point matching. Our results are satisfactory and encouraging for future researchers.

85

In image compositing detection, we proposed two entirely different approaches. In first method, we used supervised learning concept. We trained the images using SVM (support vector machine) using an RSF kernel. The classification decisions are based on the image quality features and images moment features. In second Approach, we used the statistical methods. We are able to detect the image forgery based on abrupt, unnatural, sudden changes in illumination levels, color, and edge of images based on luminance test, HSV color space test and image filtering using customized masks

1.34 Recommendations for Future Research

Images are used everywhere. Images are also used in multimedia contents. Today it is essential to assure the authenticity and security of entire digital contents. Hence it is very important to ensure the authenticity of the images. The image forgery detection is one such effort. The digital image forensics is a new research area, since it poses more challenges to researchers, providing equally good opportunities. The worldwide researchers herein follow two different approaches. The first one is an active image authentication technique such as watermarking, digital signatures etc. The second approach is passive blind image authentication methods, which we used in the current thesis. The passive blind image forensics is another way for image authentication. Below we suggest few new methods for researchers in the area of passive blind image forensics. 1.34.1 Multimodal forensics Currently, the research in digital forensics using passive blind methods are reported in images [38, 51], Apart from images, only a few studies [36, 42,170] are reported in the area of video forensics. There is wide scope for extending the passive blind digital forensics approaches to other multimedia contents, particularly, video and speech. 1.34.2 Color Theory The research in passive blind image forensics carried out and reported so far in respect of color images has been scarce. The color images are invariably converted either to gray level image or a binary image [47, 51,111,138]. Color provides a rich set of features, which gives more cue for manipulations. Future research should consider the using the 86

properties and features of color images directly by exploiting a color theory concepts to expose the image forgeries. 1.34.3 3D scene Consistency The matching a 3D scenes feature is very difficult in a forged image. Again the estimating and quantifying the 3D scene character features are not so easy. The research may try to find the distortion and inconsistencies in forged image due to 3D scene matching. So far reported research using the consistency properties are limited [70, 73,111,132] 1.34.4 Fusion The passive blind techniques used the supervised and statistical approaches to detect the image forgeries. They used the image statistics [40]. Device properties [102, 22] and tampering artifacts [29, 86]. Future researcher may use fusion approaches and knowledge from other fields. 1.34.5 Evolution The passive blind image forensics research is entirely based on image function. The advances in image acquisition devices, sophisticated image processing and computer graphics software are rapidly evolving. These evolutions in these technologies and devices need to be explored for new opportunities for researchers Morphing Several cases of morphing based image forgeries are also being noted recently. Detecting such forgeries is other area for future research.

87

BIBILOGRAPHY

1. M.Arnold, M.Schmucker, S.D.Wolthusen,Techniques and Applications of Digital Watermarking and Content Protection, Artech house,Inc, Norwood, MA, USA, 2003. 2. I.Avcibas, B.Sankur, K.Sayood,Statistical Evaluation of Image Quality Measure, Journal of Electronic Imaging,2002 Vol 11,pp206-223. 3. I.Avcibas, N.Menon,B.Sankur,Steganalysis Using Image Quality Metrics, IEEE Transactions on Image processing, 2003.Vol.12,pp21-229. 4. I.Avcibas, S.Bayram, N.D.Memon, M.Ramkumar, B.Sankur, A classifier design for detecting image manipulations,in:ICIP,2004, pp. 26452648. 5. S.Battiato, G.Messina,Digital forgery estimation into dct domain: a critical analysis,in:MiFor09: Proceedings of the First ACM Workshop on Multimedia in Forensics, ACM, NewYork, NY,USA,2009, pp.3742. 6. G.Baxes, A Digital Image Processing: Principles and Applications, New York: John Wiley & Sons, Inc, 1994. 7. S.Bayram, I.Avcibas, B.Sankur, N.Memon, Image manipulation detection with binary similarity measures, in: Proceedings of 13 th European Signal Processing Conference, vol.1, Antalya, Turkey, 2005, pp.752755. 8. S.Bayram.,I.Avcibas, B.Sankur, N.Memon, Image manipulation detection, Journal of Electronic Imaging15(4)(December2006) 041102-1041102-17. 9. S.Bayram.S, H.T.Sencar, N.D.Memon, Classification of digital camera-models based on demosaicing artifacts, Digital Investigation 5(12) (2008)4959. 10. S.Bayram.H.T.Sencar,N.D.Memon,I.Avcibas, Source camera identification based on cfa interpolation,in: ICIP(3),2005,pp.6972. 11. S.Bayram.S,H.TahaSencar,N.Memon, An efficient and robust method for detecting copymoveforgery,in:ICASSP09:Proceed-ings of the 2009 IEEE International Conference on Acoustics, Speech and Signal Processing, IEEE Computer Society, Washington, DC,USA, 2009,pp.10531056.

88

12. S.Bhattacharjee, M. Kutter, Compression tolerant image authentication, In IEEE International Conference on Image Processing 1998, ICIP98, proceedings,Chcago,IL,USA.1998 13. J.F.Blinn,M.E.Newell, Texture and reflection in computer generated images. Communications of the ACM 1976, 19(10), 542547. 14. S.Bravo-Solorio, A.K.Nandi, Passive forensic method for detecting duplicated regions affected by reflection, rotation and scaling ,in: European Signal Processing Conference,2009,pp.824828. 15. G.Cao,Y.Zhao , R.Ni, Detection of image sharpening based on Histogram aberration and ringing artifacts ,in: IEEE International Conference on Multimedia and Expo,2009,pp.10261029. 16. G.Cao,Y.Zhao, R.Ni, Edge-based blur metric for tamper detection ,Journal of Information Hiding and Multimedia Signal Processing1(1) (2010)2027. 17. H.Cao.A.C.Kot, A generalized model for detection of demosaicing Characteristics , in:ICME,2008,pp.15131516. 18. H.Cao. A.C.Kot, Accurate detection of demosaicing regularity for digital image forensics,IEEE Transactions on Information Forensics and Security4(4)(2009) 899910. 19. O.Celiktutan, I.Avcibas,B.Sankur, Blind identification of source cell-phone model, IEEE Transactions on Information Forensics and Security 3(3) (September2008)553566. 20. C.Chang,C.J.Lin,LIBSVM: A Library for Support Machines, http://www.csie.ntu.edu.tw/~cjlin/libsvm/index.html. 21. M.Chen.M.Goljan,J.Lukas, Determining image origin and integrity using sensor noise ,IEEE Transactions on Information Forensics and Security 3(1) (March2008)7490. 22. W.Chen.Y.Q.Shi,W.Su, Image splicing detection using 2-dphase congruency and statistical moments of characteristic function, in: SPIE Electronic Imaging :Security, Steganography and Watermarking of Multimedia Contents, San Jose ,CA, USA, January2007.

89

23. Y.L.Chen., C.-T.Hsu, Detecting doubly compressed images based on quantization noise model and image restoration, in: IEEE Workshop on Multimedia Signal Processing, 2009. 24. H. R. Chennamma, Lalitha Rangarajan, Image Splicing Detection Using Inherent Lens Radial Distortion, IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 6, November 2010, pp. 149-158,2010 25. H. R. Chennamma,Lalitha Rangarajan,Veerabhadrappa, Face Identification from Manipulated Facial Images using SIFT, CoRR, abs/1105.4712:2011 26. H. R. Chennamma, Lalitha Rangarajan, M. S. Rao: Robust Near Duplicate Image Matching for Digital Image Forensics. IJDCF 1(3): 62-79 (2009) 27. Columbia DVMM Research Lab, Columbia Image Splicing Detection Evaluation dataset, http://www.ee.columbia.edu/ln/dvmm/downloads/AuthSplicedDataSet/AuthSplic ed DataSet.htm 28. I.Cox,J.Bloom, M. Miller,Digital Watermarking, Principles and Practice. Morgan Kaufmann 2001 29. S.Dehnie,H.T.Sencar,N.D.Memon,Digital computer generated and digital Atlanta,USA,2006,pp.23132316. 30. A.E.Dirik,S.Bayram,H.T.Sencar,N.Memon, New features to identify computer generated images ,in: IEEE International Conference on Image Processing ,ICIP07,vol.4,2007,pp.433436. 31. A.E.Dirik,N.Memon, Image tamper detection based on demosaicing Artifacts, in:ICIP(09),Cairo,Egypt,November2009,pp.429432. 32. A.E.Dirik,H.T.Sencar,N.Memon,Digital single lens reflex camera identification from traces of sensor dust, IEEE Transactions on Information ForensicsandSecurity3(3)(September2008) 539552. 33. J.Dong,W.Wang,T.Tan,Y.Shi, Run-length and edge statistics based approach for image splicing detection, in: Digital Water- marking, 7th International Work shop,IWDW2008,Busan,Korea,November 10-12,2008,pp.7687. image forensics images for ,in: identifying ICIP,IEEE, camera

90

34. B.Dybala.B.Jennings,D.Letscher,

Detecting

filtered

cloning

in

digital

images,MM&Sec07:Proceedings of the 9th Workshop on Multimedia & Security,ACM,NewYork,NY,USA,2007,pp.4350. 35. M.Fairchild, "Color Appearance Models: CIECAM02 and Beyond." Tutorial slides for IS&T/SID 12th Color Imaging Conference. 36. N.Fan,C.Jin,Y.Huang, A pixel-based digital photo authentication framework via demosaicking inter-pixel correlation ,in: MM& Sec 09:Proceedings of the 11th ACM Workshop on Multimedia and Security, ACM,NewYork,NY,USA,2009,pp.125130. 37. Z.Fan,R.L.deQueiroz, Identification of bitmap compression history: jpeg detection and quantizerestimation ,IEEE Transactions on ImageProcessing12(2) (2003)230235. 38. H.Farid, Detecting digital forgeries using bispectral analysis, Technical Report AIM-1657,AILab,Massachusetts Institute of Technology, 1999. 39. H.Farid, Digital image ballistics from JPEG quantization ,Technical Report TR2006-583,Department of Computer Science,Dartmouth College, 2006. 40. H.Farid, Exposing digital forgeries in scientific images ,in: ACM Multimedia and SecurityWorkshop,Geneva,Switzerland,2006. 41. H.Farid.Exposing digital forgeries from jpeg ghosts ,IEEE Transactions on Information Forensics and Security1(4)(2009)154160. 42. H.Farid.A survey of image forgery detection ,IEEE Signal Processing Magazine 2(26)(2009)1625. 43. H.Farid, Lyu, S., Higher-order wavelet statistics and their application to digital forensics, In IEEE Workshop on Statistical Analysis in Computer Vision, Madison, Wisconsin 2003 44. H.Farid.M.Bravo, Image forensic analyses that elude the human visual system ,in: SPIE Symposium on Electronic Imaging ,SanJose, CA, 2010. 45. X.Feng.G.Doerr,Jpeg recompression detection,in: SPIE Conference on Media Forensics and Security,2010.

91

46. C.S.Fillion,S,G.Sharma,Detecting content adaptive scaling of images for forensic applications,in: Proceedings of the SPIE, Electronic Imaging, MediaForensicsandSecurityXII,2010. 47. G.L.Friedman, The trustworthy digital camera: restoring credibility to The photographic image, IEEE Transactions on Consumer Electronics, 39(4):905910.1993 48. J.Fridrich, Goljan, M., and Baldoza, A. C, New fragile authentication watermark for images, In IEEE International Conference on Image Processing, volume 1,2000 49. J.Fridrich, Digital image forensics ,IEEE Signal Processing Magazine 2 (26) (2009)2637. 50. J.Fridrich,, Goljan, M., Baldoza, A. C. New fragile authentication watermark for images, In IEEE International Conference on Image Processing, volume 1. 2000 51. J.Fridrich, Lukas, Estimation of primary quantization matrix in double compressed jpeg images,in:ProceedingsofDFRWS,vol.2, Cleveland, OH,USA,August2003. 52. J.Fridrich.T.Pevny, Detection of double-compression for applications in steganography ,IEEE Transactions on Information Security and Forensics3(2) (June2008)247258. 53. J.Fridrich.D.Soukal,J.Lukas, Detection of copymove forgery in digital images ,in: Proceedings of Digital Forensic Research Work-shop, IEEE Computer Society,Cleveland,OH,USA,August2003,pp. 5561. 54. D.Fu.,Y.Q.Shi,W.Su, Detection of image splicing based on HilbertHuang transform and moments of characteristic functions with wavelet decomposition ,in: International Workshop on Digital Watermarking, Jeju,Korea,November2006,pp.177187. 55. D.Fu.,Y.Q.Shi,W.Su, A generalized Benfords law for jpeg coefficients and its applications in image forensics ,in: SPIE Electronic Imaging: Security, Steganography, and Watermarking of Multimedia Contents, SanJose, CA,USA, January2007. 92

56. A.T.Gallagher.T.Chen, Image authentication by detecting traces of demosaicing , in: Proceedings of the CVPRWVU Workshop, Anchorage, AK, USA ,June2008 ,pp.18. 57. A.C.Gallagher, Detection of linear and cubic interpolation in jpeg compressed images,in:CRV05:Proceedings of the 2nd Canadian Conference on Computer and Robot Vision(CRV05),IEEE Computer Society, Washington,DC,USA,2005,pp.6572. 58. S.Gholap,P.K.Bora,Illuminant colour based image forensics ,in: TENCON 2008,TENCON2008.IEEERegion10Conference,IEEEComputerSociety,Hyderabad ,India,November2008,pp.15. 59. T.Gloe.A.Winkler,K.Borowka, Efficient estimation and large-scale Evaluation of lateral chromatic aberration for digital image forensics, in: SPIEConferenceonMediaForensicsandSecurity,2010. 60. V.Gopi, Digital image forgery detection using artificial neural network and independent component analysis , Applied Mathematics and Computation 194 (2) (2007) pp.540543. 61. E.S.Gopi,N.Lakshmanan,T.Gokul,S.KumaraGanesh,P.R.Shah, forgery detection using artificial neural network and coefficients,in:CCECE,2006,pp.194197. 62. H.A.Gou.Swaminathan,M.Wu, Noise features for image tampering detection and steganalysis ,in:ICIP(6),IEEE,San Antonio, USA,2007,pp.97100. 63. N.Greene, Environment mapping and other applications of world projections, IEEE Computer Graphics and Applications 1986, 6(11), 2129. 64. M.D.Grossberg, Nayar, S. K, Modeling the space of camera response functions, IEEE Transactions on Pattern Analysis and Machine Intelligence 2004, 26(10), 12721282. 65. G.E.Healey,R.Kondepudy, Radiometric CCD camera calibration and noise estimation, IEEE Transactions on Pattern Analysis and Machine Intelligence 1994, 16(3), 267276. 66. D.Y.Hsiao,S.-C.Pei, Detecting digital tampering by blur estimation, in:SADFE05:Proceedings of the First International Workshop on Systematic 93 Digital auto image regressive

Approaches to Digital Forensic Engineering (SADFE05) on Systematic Approaches to Digital Forensic Engineering, IEEEComputerSociety,Washington,DC,USA,2005,p.264 67. Y.F.Hsu, S.-F.Chang, Detecting image splicing using geometry invariants and camera characteristics consistency , in: ICME,2006, pp. 549552. 68. Y.F.Hsu, S.-F.Chang, Image splicing detections using camera response function consistency and automatics estimation ,in: ICME, 2007,pp.2831. 69. H.Huang,W.Guo,Y.Zhang, Detection of copymove forgery in digital images using sift algorithm,in:PACIIA08:Proceedingsof the 2008IEEEPacificAsiaWorkshopon Computational Intelligence and Industrial Application, IEEE Computer Society, Washington, DC, USA, 2008,pp.272276. 70. Y.Huang,Y.Long, Demosaicking recognition with applications in digital photo authentication based on a quadratic pixel correlation model, in: IEEE Conference on Computer Vision and Pattern Recognition, Anchorage,AK,USA,2008,pp.18. 71. W.Jing,Z.Hongbin, Exposing digital forgeries by detecting traces of image splicing ,in: 8th International Conference on Signal Processing, Guilin,China,November2006,pp.1620. 72. M.Johnson, Farid, H. Exposing digital forgeries by detecting inconsistencies in lighting. In : ACM Multimedia and Security Workshop, New York, NY 2005 73. M.Johnson,H.Farid, Exposing digital forgeries through chromatic aberration , in: ACM Multimedia and Security Workshop, Geneva, Switzerland, 2006. 74. M.Johnson.H.Farid, Metric measurements on a plane from a single Image, Technical ReportTR2006-579,Department of Computer Science, DartmouthCollege,2006. 75. M.Johnson.H.Farid, Detecting photographic composites of people , in: 6 th International Workshop on DigitalWatermarking,Guangzhou, China,2007. 76. M.Johnson.H.Farid, (2007)450461. Exposing digital forgeries in complex lighting environments, IEEE Transactions on Information Forensics and Security 3(2)

94

77. M.Johnson.H.Farid ,Exposing digital forgeries through specular highlights on the eye, in: 9th International Workshop on Information Hiding,SaintMalo,France,2007. 78. E.Kee,H.Farid, Digital image authentication from thumb nails ,in: Proceedings of the SPIE ,Electronic Imaging ,Media Forensics and Security XII,2010. 79. N.Khanna.G.T.-C.Chiu,J.P.Allebach,E.J.Delp, Forensic techniques for classifying scanner ,computer generated and digital camera images , in: IEEE International Conference on Acoustics Blind ,Speech source and camera Signal Processing,LasVegas,USA,April2008,pp.16531656. 80. M.Kharrazi,H.T.Sencar,N.D.Memon, in:ICIP,2004,pp.709712. 81. M.Kirchner,Fast and reliable resampling detection by spectral analysis of fixed linear predictor residue ,in: Proceedings of the 10th ACM workshop on Multimedia and security , ACM , NewYork, NY, USA,2008,pp.1120. 82. M.Kirchner. J.Fridrich, On detection of median filtering in digital images ,in: Proceedings of the SPIE, Electronic Imaging, Media Forensics and Security XII, SanJose, CA, USA, January 2010. 83. A.Langille,M.Gong,An and Robot efficient match-based IEEE duplication Computer detection Society, algorithm,in:CRV06:Proceedings of the 3rd Canadian Conference on Computer Vision(CRV06), Washington,DC,USA,2006,pp.64. 84. T.V.Lanh, K.S.Chong,S.Emmanuel,M.S.Kankanhalli,A Survey on digital camera image forensic methods ,in:ICME,2007,pp.1619. 85. S.Lee,D.A. Shamma,B.Gooch, Detecting false captioning using common-sense reasoning ,DigitalInvestigation3(Suppl.1)(2006) 6570. 86. A.Leykin A, F.Cutzu, Differences of edge properties in photographs and paintings ,in: ICIP(3),2003,pp.541544. 87. A.Leykin.F.Cutzu,H.Riad, Distinguishing paintings from Photographs, Computer Vision and Image Understanding100 (3) (2005) 249273. 88. C.T.Li, Detection of block artifacts for digital forensic analysis ,in:e-Forensics, 2009,pp.173178. 95 identification,

89. G.Li,Q.Wu,D.Tu,S.Sun, A sorted neighborhood approach for detecting duplicated regions in image forgeries based on dwt and svd, in:ICME,2007,pp.17501753. 90. W.Li, Y.Yuan,N.Yu, Detecting copy-paste forgery of jpeg image via block artifact grid extraction, in: International Workshop on Local and NonLocalApproximationinImageProcessing,2008. 91. Z.Li, J.B.Zheng, Blind detection of digital forgery image based on the local entropy of the gradient,in:IWDW, 2008,pp.161169. 92. C.Y.Lin, S.F.Chang, Generating robust digital signature for image/video authentication, in: ACM Multimedia Workshop, 1998, pp.115118. 93. C.Y.Lin, S.F.Chang, A robust image authentication method surviving JPEG lossy compression, In SPIE Storage and Retrieval of Image/Video Database, volume 3312,1998 94. C.Y.Lin,S.F.Chang, A robust image authentication method distinguishing JPEG compression from malicious manipulation, IEEE Transactions on Circuits and Systems for Video Technology2001, 11(2):153-168. 95. E.T.Lin, C.I. Podilchuk, E.J Delp, Detection of image alterations using semifragile watermarks. In SPIE International Conference on Security and Watermarking of Multimedia Contents II 2000, volume 3971, pages 152-163. 96. H.J.Lin, C.W.Wang,Y.T.Kao, Fast copymove forgery detection,WSEAS TransactionsonSignalProcessing5(5)(2009)188197. 97. W.Y.Lin,S.Tjoa,H.V.Zhao,K.J.R.Liu,Image source coding forensics via intrinsic fingerprints ,in:ICME,2007,pp.11271130. 98. Z.Lin,,J.He,X.Tang,C.K.Tang, Fast, automatic and fine-grained tampered jpeg image detection via dct coefficient analysis ,Pattern Recognition 42(11) (2009)24922501. 99. Z.Lint,,R.Wang,X.Tang,H.Y.Shum,Detecting doctored images using camera response normality and consistency ,in:CVPR05: Proceedings of the 2005 IEEEComputer Society Conference on Computer Vision and Pattern 1,IEEE Recognition(CVPR05)Volume ComputerSociety,Washington,DC,USA,2005,pp.10871092. 96

100. C.Liu,R. Szeliski, S.B.Kang, C.L. Zitnick, W.T.Freeman, Automatic estimation and removal of noise from a single image, IEEE Transactions on Pattern Analysis and Machine Intelligence,2008 ,30(2), 299314. 101. Q.Liu, A.H.Sung, A new approach for jpeg resize and image splicing detection ,in:MiFor09:ProceedingsoftheFirstACM workshop on Multimedia in Forensics, ACM, NewYork,NY,USA,2009, pp.4348. 102. C.S.Lu,H.M.Liao,Structural digital signature for image authentication: an incidental distortion resistant scheme ,in:MULTIMEDIA 00: Proceedings of the 2000ACM workshops on Multimedia,ACM Press, NewYork,NY,USA,2000,pp.115118. 103. J.Lukas,Digital image authentication using image filtering techniques, in: Proceedings of ALGORITMY 2000, Conference on Scientific Computing,Podbanske,Slovakia,September2000, pp. 236244. 104. J.Lukas, J.Fridrich,M.Goljan,Detecting digital image forgeries using sensor pattern noise, In SPIE Electronic Imaging, Photonics West 2006. 105. W.J.Luo,G.Huang,G.Qiu, Robust detection of region-duplication forgery in digital image ,in:ICPR06:Proceedings of the18th International Conference on Pattern Recognition, IEEE Computer Society, Washington,DC,USA,2006,pp.746 749. 106. W.Luo,W.J.Huang,G.Qiu,A novel method for block size forensics based on morphological operations,in:IWDW,2008,pp. 229239. 107. W.Luo,Z.Qu,J.Huang,G.Qiu,A novel method for detecting cropped and recompressed image block ,in: IEEE International Conference on Acoustics,SpeechandSignalProcessing,vol.2, HI,USA,April2007,pp.217220. 108. W.Luo,Z.Qu,F.Pan,J.Huang, A survey of passive technology for digital image forensics ,Frontiers of Computer ScienceinChina1(2) (2007) 166179. 109. S.Lyu,H.Farid, How realistic is photorealistic? IEEE Transactions on Signal Processing 53(2)(2005)845850. Honolulu,

97

110. A.D.Mohammad , Introduction to Inverse Problems in Imaging and Vision In: Inverse Problems in Vision and 3D Tomography Edited by:Ali MohammadDjafari. 15-58 ISTE-WILEY 2010 111. A.D.Mohammad, Inverse Problems in Vision and 3D Tomography Edited by:Ali Mohammad-Djafari. ISTE-WILEY 2010 isbn:9781848211728 112. D.Mahajan, R. Ramamoorthi, B. Curless, A theory of spherical harmonic identities for BRDF/lighting transfer and image consistency,. In European Conference on Computer Vision, Graz, Austria 2006 113. B.Mahdian ,S.Saic, Detection of copymove forgery using a method based on blur moment invariants ,Forensic Science International 171 (23)(2007)180189. 114. B.Mahdian,S.Saic, Blind authentication using periodic properties of interpolation, IEEE Transactions on Information Forensics and Security 3(3)(September2008) 529538. 115. B.Mahdian,S.Saic, Blind methods for detecting image fakery ,in: IEEE International Conference on Security Technology, IEEE Computer Society, Prague, Czech Republic, October2008,pp. 280286. 116. B.Mahdian,S.Saic, Computational Detection Sciences of Resampling its supplemented ,IEEE with noise inconsistencies analysis for image forensics ,in: International Conference on and Applications Computer Society,Perugia,Italy,July2008,pp.546556. 117. B.Mahdian,S.Saic,A cyclostationarity analysis applied to image forensics , in: IEEE Workshop on Applications of Computer Vision (IEEE WACV),December2009. 118. B.Mahdian,S.Saic, Detecting double compressed jpeg images ,in: The 3rd International Conference on Imaging for Crime Detection and Prevention(ICDP09),London,UK,December2009. 119. B.Mahdian,S.Saic, Detection and description of geometrically transformed digital images ,in: Media Forensics and Security, Proceedings ofSPIEIS&TElectronicImaging,vol.7254,SanJose, CA, USA,January2009. 120. B.Mahdian,S.Saic, Using noise in consistencies for blind image forensics, ImageVision Computing27(10)(2009)14971503. 98

121. N.Memon,P.Vora,Authentication techniques for multimedia Content, In SPIE Multimedia Systems and Applications 1999, volume 3528-pages 412-422. 122. T.G.Matthias Kirchner,On Resampling detection in re-compressed Images, in :IEEE Workshop on Information Forensics and Security, December 2009,pp.21 25. 123. P.Moulin, The role of information the origin watermarking and its application to image watermarking ,Signal Processing 81(6)(2001) 11211139. 124. J.Marroquin, S. Mitter, T. Poggio, Probabilistic Solution of Ill-Posed Problems in Computational Vision, Journal of the American Statistical Association, Vol. 82, No. 397 (Mar., 1987), pp. 76-89 125. A.N.Myna,M.G.Venkateshmurthy,C.G.Patil, Detection of region duplication forgery in digital images using wavelets and log-polar on mapping, in:ICCIMA07:ProceedingsoftheInternationalConference Society, Washington, DC,USA,2007, pp.371377. 126. L.Nataraj,A.Sarkar,B.S.Manjunath, Adding Gaussian noise to denoise jpeg for detecting image resizing ,in: International Conference on Image Processing,2009,November2009. 127. R.Neelamani,R.L.deQueiroz,Z.Fan,S.Dash,R.G.Baraniuk,Jpeg compression history estimation for color images ,IEEE Transactions on Image Processing 15(6) (2006)13651378. 128. T.T.Ng, Camera response function signature for digital forensicspart II: signature extraction ,in: IEEE Workshop on Information Forensics and Security,December2009,pp.161165. 129. T.T.Ng,S.F.Chang,A model for image splicing ,in: IEEE International Conference on Image Processing(ICIP),Singapore,October2004. 130. T.T.Ng,S.F.Chang, A non-line system for classifying computer graphics images from natural photographs, in: SPIE Electronic Imaging, SanJose,CA,January2006. 131. T.T.Ng,S.F.Chang,J.Hsu,L.Xie,M.P.Tsui,Physics-motivated distinguishing photographic images and computer features graphics, for in: Computational

Intelligece and Multimedia Applica- tions (ICCIMA2007), IEEE Computer

99

MULTIMEDIA05:Proceedings of the 13th Annual ACM International Conference on Multimedia, ACM, NewYork, NY, USA,2005,pp.239248. 132. T.T.Ng,S.F.Chang,C.Y.Lin,Q.Sun,Passive-blind Rights,Elsevier,Hawthorne,NY,USA,2006. 133. T.T.Ng,S.F.Chang,Q.Sun, Blind detection of photomontage using higher order statistics ,in: IEEE International Symposium on Circuits and Systems (ISCAS),Vancouver,Canada,2004. 134. T.T.Ng,S.F.Chang,M.P.Tsui,Lessons learned from online classification of photorealistic computer graphics and photographs , in: IEEE Workshop on Signal Processing Applications for Public Security and Forensics (SAFE),April2007. 135. T.T.Ng,M.P.Tsui,Camera response function signature for digital forensics part I:theory and data selection ,in IEEE Workshop on Information Forensics and Security,December2009,pp.156160. 136. N.Nikolaidis, I.Pitas, Robust image watermarking in the spatial domain in Proceedings of IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP '98), pp. 29692972, 137. A.V.Oppenheim,R.W.Schafer,J.R.Buck,Discrete-Time Signal Processing, Prentice Hall 1999, Englewood Cliffs, New Jersey, 2nd edition. 138. M.C.Poilpre,P.Perrot,H.Talbot, International Conference on Image Forensic tampering detection and using Bayer in interpolation and jpeg compression ,in: e-Forensics 08: Proceedings of the1st Applications Techniques Telecommunications, Information, and Multimedia and Workshop, ICST, Brussels, Belgium ,Belgium,2008, pp.15.ICST(Institute for Computer Sciences, SocialInformatics and Telecommunications Engineering). 139. T.Poggio,C.Koch,Ill-Posed Problems in Early Vision: From Computational Theory to Analogue Networks, Proceedings of the Royal Society of London. Series B, Biological Sciences, Vol. 226, No. 1244. (Dec. 23, 1985), pp. 303-323. 140. A.Popescu,H.Farid, Exposing digital forgeries by detecting duplicated image regions,TechnicalReportTR2004-515,Department of Compute Science, Dartmouth College, 2004. 100 image forensics, in: W.Zeng,H.Yu,C.Y.Lin(Eds.), Multimedia Security Technologies For Digital

141. A.Popescu,H.Farid, Exposing digital forgeries by detecting traces of resampling, IEEE Transactions on Signal Processing 53(2) (2005) 758767. 142. A.Popescu,H.Farid, Exposing digital forgeries in color filter array interpolated images, IEEE Transactions on Signal Processing 53(10) (2005)39483959. 143. A.Popescu, Statistical tools for digital image forensics , Ph.D. Thesis, Department of Computer Science, Dartmouth College, Hanover, NH,2005. 144. S.Prasad,K.R.Ramakrishnan, On resampling detection and its application to image tampering ,in: Proceedings of the IEEE International Conference on Multimedia and Exposition, Toronto, Canada, 2006,pp.13251328. 145. Z.Qu Z,W.Luo,J.Huang, A convolutive mixing model for shifted double jpeg compression with application to passive image authentication, in: IEEE International Conference on Acoustics, Speech and Signal Processing,LasVegas,USA,April2008,pp.42441483. 146. Z.Qu,G.Qiu,J.Huang, Detect digital image splicing with visual cue in:InformationHiding,11thInternationalWorkshop,IH2009,Darmstadt,Germany, June8-10,2009,pp.247261. 147. C.J.Rey,J.L.Dugelay, A survey of watermarking algorithms for image authentication,EURASIP interactive services 2002). 148. A.S.Rocha,S. Goldenstein, Is it fake or real ? in: XIX Brazilian Symposium on Computer Graphics and Image Processing, Manaus, Brazil,April2006. 149. J.Sachs, Digital Image Basics. Digital Light & Color. 1999. 150. Saiqa Khan, Arun Kulkarni, Reduced Time Complexity for Detection of CopyMove Forgery Using Discrete Wavelet Transform, International Journal of Computer Applications (0975 8887)Volume 6 No.7, September 2010 151. Saiqa Khan, Arun Kulkarni, An Efficient Method for Detection of Copy-Move Forgery Using Discrete Wavelet Transform, International Journal on Computer Science and Engineering,Vol. 02, No. 05, 2010, 1801-1806 Journal on applied Signal Processing 2002(June2002)613621(special issue on image analysis for multimedia

101

152. B.L.Shivakumar, S.Santhosh Baboo, Detecting Copy-Move Forgery in Digital Images: A Survey and Analysis of Current Methods , Global Journal of Computer Science and Technology Vol. 10 Issue 7 Ver. 1.0 September 2010 153. S.Santhosh Baboo, P Subashini,Digital Water-marking for Mobile and PDA Devices based Images GVIP 05 Conference, 19-21 December 2005, CICC, Cairo, Egypt 154. G.Sankar,V.Zhao,Y.H.Yang, Feature based classification of computer graphics and realimages,in:ICASSP09:Proceedingsof the 2009 IEEE International Conference on Acoustics, Speech and Signal Processing, IEEE Computer Society, Washinton, DC,USA, 2009, pp.15131516. 155. A.Sarkar,L.Nataraj,B.S.Manjunath, Detection of seam carving and localization of seam insertions in digital images ,in: MM&Sec 09:Proceedings of the11th ACMWorkshop on Multimedia and Security, ACM, NewYork,NY, USA,2009 ,pp.107116. 156. M.Schneider,S.F.Chang,A robust content based digital signature for image authentication ,in: IEEE International Conference on Image Processing (ICIP96), 1996. 157. H.T.Sencar,,N.Memon, Overview of state-of-the-art in digital image forensics, Indian Statistical Institute Platinum Jubilee Monograph series titled Statistical Science and Interdisciplinary Research, December2008,pp.120. 158. H.T.Sencar, Applications: Ramkumar, Content A.N.Akansu, Security natural in Data Digital model Hiding Fundamentals and Multimedia, approach and to Academic splicing (ACM in:

Press,Inc.,Orlando,FL,USA,2004. 159. Y.Q.Shi,C.Chen,W.Chen,A detection ,in: ACM image on Workshop Multimedia versus Security

MMSEC07),ACM,NewYork,NY,USA,September2007,pp. 5162. 160. Y.Q.Shi,C.Chen,G.Xuan,W.Su,Steganalysis International Workshop on Digital Guangzhou,China,December2007. 161. Y.Q.Shi,W.Chen,G.Xuan,Identifying computer graphics using hsv color model and statistical moments of characteristic functions, in:ICME,2007,pp.11231126. 102 splicing detection, Watermarking (IWDW07),

162. M.Sorell ,Conditions for effective detection and identification of primary quantzation of re-quantized jpeg images ,in : e-Forensics 08: Proceedings of the 1st International Conference on Forensic Applications and Techniques in Telecommunications, Information, and Multimedia and Workshop, ICST, Brussels, Belgium, Belgium, 2008, pp.16. ICST(Institute for Computer Sciences, SocialInformatics and Telecommunications Engineering). 163. M.Stamm, K.Liu, Blind forensics of contrast enhancement in digital images ,in: IEEE International Conference on Image Processing, SanDiego,CA,USA,2008,pp.31123115. 164. Y.Sutcu, B.Coskun,H.T.Sencar,N.Memon, Tamper detection based on regularity of wavelet transform coefficients ,in: ICIP: IEEE International Conference on imageProcessing, IEEE,San Antonio, USA,2007,pp.397400. 165. P.Sutthiwan,J.Ye,Y.Q.Shi, An enhanced statistical approach to identifying photorealistic images,in:IWDW09:Proceedingsof the 8th International Workshop on Digital Watermarking,Springer-Verlag, Berlin,Heidelberg,2009,pp.323335. 166. A.Swaminathan,M.Wu,K.J.R.Liu, Image tampering identification using blind deconvolution,in:ICIP,2006,pp.23092312. 167. A.Swaminathan ,M.Wu,K.J.R.Liu, Non intrusive component forensics of visual sensors using output images ,IEEE Transactions on Information Forensics and Security2(1)(2007) 91106. 168. A.Swaminathan,M.Wu,K.J.R.Liu, Hiding traces of Resampling in digital images ,IEEE Transactions on Information Forensics and Security 3(1) (March2008)101117. 169. A.Swaminathan,M.Wu,K.J.R.Liu, Component forensics ,IEEE Signal Processing Magazine26(2)(March2009)3848. 170. S.Tjoa,W.Y.S.Lin,K.J.R.Liu, Transform coder classification for digital image forensics ,in: ICIP(6),2007,pp.105108. 171. S.Tjoa,W.Y.S.Lin,H.V.Zhao,K.J.R.Liu, Block size forensic analysis in digital images ,in: IEEE International Conference on Acoustics, Speech and Signal Processing ,Honolulu HI,USA, April 2007.

103

172. Y.Tsin,V.Ramesh,T. Kanade, Statistical calibration of CCD imaging process, In Eighth International Conference on Computer Vision (ICCV 2001), pages 480 487, Vancouver, Canada. 173. C.H.Tzeng,W.H.Tsai,A new technique for authentication of image/video for multimedia applications ,in: Proceedings of the 2001 Workshop on Multimedia and Security, ACM Press, NewYork, NY,USA,2001,pp.2326. 174. W.Wang,J.Dong,T.Tan, Effective image splicing detection based on image chroma ,in: IEEE International Conference on Image Processing,2009. 175. X.Wang,B.Xuan,S.LongPeng, Digital image forgery detection based on the consistency of defocus blur ,International Conference on Intelligent Information Hiding and Multimedia Signal Processing IEEE Computer Soci ty,LosAlamitos,CA,USA,2008,pp.192195. 176. W.Wang,H.Farid, Exposing digital forgeries in video by detecting double MPEG compression. In ACM Multimedia and Security Workshop 2006,Geneva, Switzerland 177. W.Weimin,W.Shuozhong,T.Zhenjun, Estimation of rescaling factor and detection of image splicing ,in:11th IEEE International Conference on Communication Technology, IEEE Computer Society, Hangzhou,China,2008,pp.676679. 178. M.Wu, B. Liu, Watermarking for image authentication. In IEEE International Conference on Image Processing.1998 179. M.Wu, Multimedia data hiding ,Ph.D.Thesis, A dissertation presented to the faculty of Princeton university in candidacy for the degree of doctor of philosophy,June2001. 180. P.W.Wong, A watermark for image integrity and ownership verification, In Proc. IS&T Conference on Image Processing, Image Quality and Image Capture Systems,1998 pages 128{140. Portland, Oregon 181. M.Wu,B.Liu,Watermarking for image authentication,. In IEEE International Conference on Image Processing ICIP 98,Chicao,IL1998. 182. W.Wu,B.Liu, Multimedia Data Hiding ,Springer-Verlag NewYork, Inc., Secaucus,NJ,USA,2002.

104

183. Q.Wu ,G.H.Li,D.Tu,S.-J.Sun, State of the art and prospects on blind digital image forensics technology based on authenticity detection, Acta Automatica Sinica(2008)ISSN=0254-4156. 184. Q.Wu ,S.J.Sun,W.Zhu,G.-H.Li,D.Tu, Detection of digital doctoring in exemplarbased inpainted images, in: Seventh International Conference on Machine Learning and Cybernetics, Kunming, China,July2008,pp.12221226. 185. Z.Xiu-ming,X.Guo-rong,Y.Qiu-ming,T.Xue-feng,S.Yung-qing,Re-sampling detection in information forensics ,International Journal of Automation and Computing26(11)(2007)25962597. 186. S.Ye,Q.Sun,E.C.Chang,Detecting digital image forgeries by measuring in consistencies of blocking artifact ,in: ICME,2007, pp.1215. 187. M.Yeung, F.Mintzer,An invisible watermarking technique for image verification. In IEEE International Conference on Image Processing 1997, 188. C.Zhang,H.Zhang,Detecting digital image forgeries through weighted local entropy ,in: IEEE International Symposium on Signal Processing and InformationTechnology, Giza,Egypt, December 2007,pp.6267. 189. J.Zhang,,Z.Feng,Y.Su,A new approach for detecting copymove forgery in digital images ,in: IEEE Singapore of International in Conference jpeg2000 Application, onCommunicationSystems,2008,pp.362366. 190. J.Zhang,H.Wang,Y.Su,Detection Intelligent Information double-compression Technology images,in:IITA08: Proceedings of the 2008 Second International Symposium on IEEEComputerSociety,Washington,DC,USA,2008,pp. 418421. 191. W.Zhang,,X.Cao,Z.Feng,J.Zhang,P.Wang, Detecting photographic composites using two-view geometrical constraints, in: IEEE International Conference on Multimedia and Expo,2009, pp. 10781081. 192. W.Zhang,,X.Cao,J.Zhang,J.Zhu,P.Wang, Detecting photographic composites using shadows ,in: IEEE International Conference on Multimedia and Expo,2009,pp.10421045. 193. Z.Zhang,J.Kang,Y.Ren, An effective algorithm of image splicing detection,in:CSSE08:Proceedings of the 2008 International Conference on 105

Computer

Science

and

Software

Engineering,

IEEE

Computer

Society,Washington,DC,USA,2008,pp.10351039. 194. J.Zheng.,M.Liu, A digital forgery image detection algorithm based on wavelet Homomorphicfiltering,in:IWDW,2008,pp.152160. 195. L.Zhou ,D.Wang,Y.Guo,J.Zhang, Blur detection of digital forgery using mathematical morphology ,in:KES-AMSTA07:Proceedings of Verlag,Berlin,Heidelberg,2007,pp.990998. the1st KES International Symposium on Agent and Multi-Agent Systems, Springer-

106

 LIST OF PUBLICATIONS: RESEARCH ARTICLES (PUBLISHED / ACCEPTED)

Peer reviewed Journals

S. Math,R.C.Tripathi,Digital Forgeries:Problems and Challenges, International journal of computer Applications(ISBN 0975-8887),Vol.5,No12,August 2010.

S.Math.R.C.Trpathi,Image quality feature based Detection algorithm for forgery in images,International journal of computer graphics and Animation(ISBN),Vol.1No1,April 2011.

S.Math,R.C.Tripathi,Copy paste forgery detection using image invariant features, The International Journal of Multimedia & Its Applications (ISSN : 0975-5578(Online); 0975-5934 accepted

S.Math,R.C.Tripathi, Image Composite Detection using Customized Masks, International Journal of Information Sciences and Techniques (IJIST)ISBN 0975-5934- accepted

107

Conferences
S.Math, Ranjeet kumar, R.c.tripathi, Image Compositing Detection based on HSV Color Space Model,AICTE Sponsored National conference on Emerging Trends in Computing and information technology,NCETCIT11, RKGIT for Women,Ghaziabad. S.Math,R.C.Tripathi,Copy paste forgery detection using image invariant features,FECIT2011 ISM Dhanbad

S.Math,R.C.Tripathi, Copy paste forgery detection using supervised approach,

2011 IEEE Conference on Computer Applications & Industrial Electronics, Shah Alam, Malaysia

108

 Annexure

Figure 52: Original test images

Figure 53 original images

II

Figure 54: forged images

III

Figure 55: Result from HSV model

IV

Figure 56 Results from customized mask

Figure 57 Results from luminance test

VI

VII