ISSN: 2277 9043 International Journal of Advanced Research in Computer Science and Electronics Engineering Volume 1, Issue 2, April

l 2012

FULL- REFERENCE METRIC FOR IMAGE QUALITY ASSESSMENT Ms. Shraddha N. Utane Department Of Information Technology Sipnas COET(Amravati University) Amravati, India Prof. V. K. Shandilya Department Of Information Technology Sipnas COET(Amravati University) Amravati , India

Abstract-The quality of image is most important factor in image processing, to evaluate the quality of image various methods have been used. Proposed system defines one of the best methods in image quality assessment. Proposed system calculates the image quality assessment using normalized histogram. Sender send the image to the receiver, after receiving the image, receiver compare the image with senders image using normalized histogram. In proposed work, MGA transforms perform excellently for reference image reconstruction, have perfect perception of orientation, are computationally tractable, and are sparse and effective for image representation. MGA is utilized to decompose images and then extract features to mimic the multichannel structure of HVS. Additionally, MGA offers a series of transforms including wavelet, curvelet, bandelet, contourlet transform etc. These different types of transforms are used to capture different types of image geometric information. Contrast Sensitivity Function(CSF) and Just Noticiable Difference(JND) isused to produce a noticeable variation in sensory experience. After calculating the normalized histogram of both the reference and distorted image, we are checking the quality of both the images. Index Terms- Multiscale Geometric Analysis(MGA), Fullreference(FR) metric, Human Visual System(HVS).

I. INTRODUCTION The objective of image quality assessment (IQA) [1] is to provide computational models to measure the perceptual quality of an image. In recent years, a large number of methods have been designed to evaluate the quality of an image, which may be distorted during acquisition, transmission, compression, restoration, and processing which lead to image degradation. In poor transmission channels, transmission errors or data dropping would happened, which lead to the imperfect quality and distortion of the received video data. Therefore, how to evaluate the image quality [2]has become a burning problem. In recent years, digital camera is equipped in most of the mobile products like cellular phone, PDA and notebook computer. Image quality is the most important criteria to choose mobile products. In some cases, the benchmarks or reviews of products are based on subjective image quality test and thus are dependent on tester and environment. The subjective image quality assessment often misleads the decision for the image quality control parameters of Image Signal Processing (ISP) algorithm.The recent years has demonstrated and witnessed the tremendous and imminent demands for image quality assessment (IQA) methods at least in the following three ways: 1) They can be exploited to monitor image quality for controlling quality of processing system. 2) They can be employed to benchmark image processing systems and algorithms. 3) They can also be embedded into image processing systems to optimize algorithms and parameter settings. Existing IQA metrics

[10] can be categorized into subjective and objective methods.The former is based on the quality which is assessed by human observers, and the latter provides an objective index or real value which is obtained from any an assessment model to measure the image quality. Because human observers are the ultimate receivers of the visual information contained in an image, subjective method [5] whose results are directly given by human observers is probably a reliable way to assess the quality of an image. The subjective method is that the observers are asked to evaluate the picture quality of sequences using a continuous grading scale and to give one score for each sequence. A number of different subjective methods are represented by ITUR Recommendation BT.500[4]. The subjective quality measurement has been used for many years as the most reliable form of quality measurement. However, subjective experiment requires human viewers working in a long period, and repeated experiments are needed for many image objects, it is expensive and time consuming, and cannot be easily and routinely performed for many scenarios, e.g., real time systems. Moreover, there has not been any precise mathematical model for subjective assessment currently. As a consequence, there arouses the requirement of an objective quality metric that accurately matches the subjective quality and can be easily implemented in various image systems, leading to the emergence of the objective IQA. The objective IQA [3] is proposed to provide a computational model to measure the perceptual quality of an image. It makes use of the variation of several original or distorted image characteristics which is caused by degradation to represent the variation of the image perceptual quality. Many objective quality metrics for predicting image distortions have been investigated. Metrics are usually obtained from either reference or distorted images to reflect a number of image characteristics. Evaluation results obtained from a good objective IQA method should be statistically consistent with subjective methods. According to the availability of a reference image, there is a general agreement [1] that objective quality metrics can be divided into three categories: full-reference (FR), no-reference (NR), and reduced-reference (RR) method. II. FULL- REFERENCE METRIC (FR)

In this metric we are sending the reference or original image to the receiver side. Then with the help of quality evaluation system we are comparing both the original and degraded image. Evaluation system consist of two metrics within this class are the PSNR (Peak Signal-to-Noise Ratio) and the MSE (mean square error), due to simplicity of their computation. Conventional FR IQA methods calculate pixel-wise distances, e.g., peak signal-to-noise ratio(PSNR)

149
All Rights Reserved 2012 IJARCSEE

ISSN: 2277 9043 International Journal of Advanced Research in Computer Science and Electronics Engineering Volume 1, Issue 2, April 2012

and mean square error (MSE), between a distorted image and the corresponding reference, but they have not been in agreement with perceived quality measurement widely [1].

the system is depicted in figure 2. As an example, Webster et al. propose a reduced reference system in which the side information consists of two distinct types of measurements: spatial measurements extracted from the frames edges, and temporal measurements extracted from frames differences.

Fig.1.1 Full reference metrics III. NO-REFERENCE METRIC (NR) : Fig 1.3 Reduced-reference metric

The evaluation system has no reference to any side information regarding the original media. This kind of metrics is the most promising in the context of video broadcast scenario, since the original images or video are in practice not accessible to end users.

V.

PROPOSED WORK

Fig 1.2 No-reference metric

In the no-reference objective metrics scenario (figure 2), quality rating is attained through analysis of the received media only. Noreference objective metrics are relatively rare in literature, but some proposals have been made. Generally, the proposed algorithms evaluate some specific quality features that result from image or video transmission, like block effect in block-based DCT compression methods, edge discontinuity, etc. This kind of analysis is possible by taking into account both human visual system models and natural image models. By contrast, designing objective No-Reference (NR) quality measurement algorithms is a very difficult task. This is mainly due to the limited understanding of the HVS, and it is believed that effective NR quality assessment is feasible only when the prior knowledge about the image distortion types is available [8].

IV.

REDUCED-REFERENCE METRIC (RR):

The evaluation system has access to a small amount of side information regarding the original media, i.e. features or descriptors extracted from the original. In a reduced reference metrics scenario, the content provider transmits additional information together with the video. This class of metrics requires additional bandwidth (or an additional channel) to transmit the side information. Generically speaking, side information usually consists of relevant features extracted from the original media which are transmitted and compared with the analogous features extracted from the degraded media. The amount of additional information that is transmitted through the side channel is highly dependent of the design of the system. The general architecture of

In proposed framework implementation of image quality assessment will be carried out by considering various types of images from LIVE database. In this we are considering the reference images and distorted images distorted by some parameter. The working of proposed framework is listed below. Module 1: Image Decomposition for Feature Extraction. In which we will decompose the image using MGA framework for feature extraction. It [5] is popular to analyze signals in both time and frequency domains simultaneously and adaptively. By using multiscale operation, it extracts effective features to represent signals, especially for nonstationary signals . In our proposed work, we are using bior wavelet transform to decompose the image into highpass subbands and a lowpass residual subband. Both the reference and distorted images are decomposed by using wavelet transform. Module 2: CSF Masking In second module after decomposition of the images we are applying the CSF masking .CSF [6] measures how sensitive we are to the various frequencies of visual stimuli, i.e., we are unable to recognize a stimuli pattern if its frequency of visual stimuli is too high. It Re-weights MGA decomposed coefficients to mimic the nonlinearities inherent in HVS. It measures sensitivity to various frequency of visual stimuli. Module 3: JND Threshold After applying the CSF masking we are calculating the threshold value.It measures the minimum amount, by which stimulus intensity must be changed to produces a noticeable variation in sensory experience. Lower the JND threshold, visual quality of reconstructed image is better. Module 4: Normalized Histogram for Image Representation Histogram is constructed for both the reference image and the distorted image and both images are compared for quality. Module 5: Evaluation of Quality of Image

150
All Rights Reserved 2012 IJARCSEE

ISSN: 2277 9043 International Journal of Advanced Research in Computer Science and Electronics Engineering Volume 1, Issue 2, April 2012

We will evaluate the quality of image using proposed framework. In this we will evaluate the quality of image by using different metric such as SSIM ,FSIM etc.

Input Image
.

Evaluation Of Quality Of Image Reference image Distorted image FSIM Image Decompositi on Using Wavelet transform Image Decompositi on using Wavelet transform SSIM WSNR

Fig.3 Evaluation Metric For Image Quality

Applying the CSF Masking and calculating the Threshold value

Applying the CSF Masking and calculating the Threshold value

VI.

STRUCTURAL SIMILARITY INDEX METRIC

Construct Normalized Histogram

Construct Normalized Histogram

Calculate Quality of Image Evaluation of Quality Of of image

Calculate Quality of image

The structural similarity (SSIM) [9]index is a method for measuring the similarity between two images. The SSIM index is a full reference metric, in other words, the measuring of image quality based on an initial uncompressed or distortion-free image as reference. SSIM is designed to improve on traditional methods like peak signal-to-noise ratio (PSNR) and mean squared error (MSE), which have proved to be inconsistent with human eye perception. The difference with respect to other techniques mentioned previously such as MSE or PSNR, is that these approaches estimate perceived errors on the other hand SSIM considers image degradation as perceived change in structural information. Structural information is the idea that the pixels have strong inter-dependencies especially when they are spatially close. These dependencies carry important information about the structure of the objects in the visual scene.The SSIM metric is calculated on various windows of an image. The measure between two windows and of common size NN is: VII. FEATURE SIMILARITY INDEX METRIC

Fig.2 Framework of the proposed model

Image quality assessment (IQA) aims to use computational models to measure the image quality consistently with subjective evaluations. The well-known structural-similarity (SSIM) index brings IQA from pixel based stage to structure based stage. In this work, a novel feature-similarity (FSIM)

151
All Rights Reserved 2012 IJARCSEE

ISSN: 2277 9043 International Journal of Advanced Research in Computer Science and Electronics Engineering Volume 1, Issue 2, April 2012

[8]index for full reference IQA is proposed based on the fact that human visual system (HVS) understands an image mainly according to its low-level features. Specifically, the phase congruency (PC), which is a dimensionless measure of the significance of a local structure, is used as the primary feature in FSIM. Considering that PC is contrast invariant while the contrast information does affect HVS perception of image quality, the image gradient magnitude (GM) is employed as the secondary feature in FSIM. PC and GM play complementary roles in characterizing the image local quality. After obtaining the local similarity map, we use PC again as a weighting function to derive a single quality score. Extensive experiments performed on six benchmark IQA databases demonstrate that FSIM can achieve much higher consistency with the subjective evaluations than all the state-of-the-art IQA metrics used in comparison. Although FSIM is designed for grayscale images (or the luminance components of color images), the chrominance information can be easily incorporated by means of a simple extension of FSIM, and we call this extension FSIMC. VIII. WEIGHTED SIGNAL TO NOISE RATIO (WSNR)

framework has better effectiveness not only than the reduced-reference method, WNSIM, but also than the full reference method as MSSIM, in terms of CC, ROCC, OR, MAE, and RMSE. In addition, different transforms have different advantages in different distortions. For example, wavelet transform is optimal to Gblur distortion, curvelet transform and PSNR perform best onWNdistortion, and WBCT outperforms the other methods for FF distortion. 3) Low data rate: The proposed framework has a relatively low data rate for representing features of the reference image, i.e., a relatively low bits are used to represent the image features (e.g., for wavelet-based decomposition,there are only 10 features/image utilized for representation;contourlet-based decomposition corresponds to 17 features/image; WBCT-based decomposition corresponds to 25 features/image; and HWD-based decomposition corresponds to 17 features/image). XI. CONCLUSION In this proposed framework , a Full-reference image quality assessment framework is proposed by incorporating merits of multiscale geometry analysis (MGA), contrast sensitivity function (CSF), and the Webers law of just noticeable difference (JND). In comparing with existing image quality assessment approaches, the proposed one has strong links with the human visual system (HVS): sparse image representation is utilized to mimic the multichannel structure of HVS, CSF is utilized to balance magnitude of coefficients obtained by MGA to mimic nonlinearities of HVS, and JND is utilized to produce a noticeable variation in sensory experience. In this framework, images are represented by normalized histograms, which correspond to visually sensitive coefficients. The quality of a distorted image is measured by comparing the normalized histogram of the distorted image and that of the reference image.

In [10], a different approach to PSNR was presented: As the human visual system (HVS) is not equally sensitive to all spatial frequencies, a contrast sensitivity function (CSF) is taken into account. The CSF is simulated by a lowpass or bandpass frequency filter.First of all, the difference of the reference and the distorted image is computed. Then the difference is transformed into frequency domain using 2dimensional fast Fourier transform. The obtained error spectrum is weighted by the CSF resulting in weighted error spectrum. The last thing to do is to compute the power of the weighted error spectrum and the power of the signal (also transformed into frequency domain). IX. Metric FSIM PraposedFramework TABLE Reference Image 0.8611 77.9088 Distorted Image

XII. 0.8567 33.5261

ACKNOWLEDGEMENT

I express my sincere thanks to Prof.V.K.Shandilya for her valuable guidance.

Table 1. Quality Evaluation Parameter

X.

ADVANTAGES

XIII.

REFERENCES

1) General purpose: A number of different transforms, e.g., wavelet, curvelet, bandelet, contourlet, WBCT and HWD,for image decomposition can be applied in the proposed framework for IQA. All these transforms can work well for different image distortions, and WBCT and HWD perform much better than the others, especially for JPEG and JPEG2000 images. 2) Sound effectiveness: The objective assessment of the proposed framework performs consistently with the subjective perception, and can evaluate visual quality of JPEG and JPEG2000 image effectively. Particularly, by applying WBCT and HWD for image decomposition, the

[1]. Z. Wang and A. C. Bovik, Modern Image Quality Assessment. New York: Morgan & Claypool, 2006. .[2] Wang, Z., Bovik, C. A., and Lu, L. G., "Why is image quality assessment so difficult?" in Proc. IEEE Int. Conf. Acoustics, Speech, and Signal Processing, Florida, USA, 4, 3313-3316 (2002). [3] Wang, Z. and Bovik, C. A., Modern Image Quality Assessment, New York: Morgan and Claypool Publishing Company (2006). [4] Methodology for the Subjective Assessment of the Quality of Television Pictures, Recommendation ITU-R Rec. BT. 500-11.

152
All Rights Reserved 2012 IJARCSEE

ISSN: 2277 9043 International Journal of Advanced Research in Computer Science and Electronics Engineering Volume 1, Issue 2, April 2012

[5] S. Gabarda and G. Cristobal, Blind image quality assessment through anisotropy, J. Opt. Soc. Amer. A, vol. 24, pp. B42B51, 2007 [6] M. Miloslavski and Y.-S. Ho, Zerotree wavelet image coding based on the human visual system model, in Proc. IEEE Asia-Pacific Conf.Circuits and Systems, 1998, pp. 57 60 [7] S. Mallat, A theory for multiresolution decomposition: The wavelet representation, IEEE Trans. Pattern Anal. Mach. Intell., vol. 11, no. 7, pp. 674693, Jul. 1989

[8] FSIM: A Feature Similarity Index for Image Quality Assessment Lin Zhanga, Student Member, IEEE, Lei Zhanga,1, Member, IEEE Xuanqin Moub, Member, IEEE, and David Zhanga, Fellow, IEEE [9] Video Quality Metrics Mylne C. Q. Farias Department of Computer Science University of Braslia (UnB) Brazil. [10] Damera-Venkata, N. et al. Image Quality Assessment BasedonDegradationModelhttp://www.ece.utexas.edu/~bev ans/papers/2000/imageQuality/

153
All Rights Reserved 2012 IJARCSEE