BG3104 Set 1 2014-15

Division of Bioengineering, School of Chemical and Biomedical Engineering 1
Part I

Asst/Prof Poh Chueh Loo
Division of Bioengineering
Room: N1.3 B2-09; Tel: 6514 1088
CLPoh@ntu.edu.sg
BG3104
Biomedical Imaging
Topics to be covered in Part I
Medical Image processing techniques
Fundamentals of digital image representation
Basic Relationships between pixels
Edge detection and Enhancement
Spatial Filtering
Image segmentation

Magnetic Resonance Imaging (MRI)
Principles of MRI
Image formation
Spin Echoes
Contrast Mechanism
Image Quality

Books
Text Book

Medical Imaging Signals and Systems
J erry L. Prince and J onathan M.Links
Pearson Prentice Hall. 2006

References

Digital Image Processing.
Rafael C. Gonzalez and Richard E. Woods.
International Edition, Prentice Hall. 2008.

OVERVIEW
Medical Imaging allows us to see the inside of the human
body without cutting it open, for example, through surgery.

In this course, we will cover the most common imaging
methods in Radiology today. For example,
MRI
X-ray/Computed Tomography (CT)
Ultrasound Imaging
Optics

Each of these different methods is a different imaging
modality, and the signals that arise are fundamentally
different.

Chest X-ray
Example of Biomedical images
Head CT
Magnetic Resonance (MR) Images
Multi Slice Magnetic Resonance (MR) Images
Ultrasound
Learning objectives
Medical Image processing techniques
Able to appreciate the importance of image processing and have an
understanding of the various applications of image processing in
medical imaging.

Able to determine the relationship of pixels in terms of neighbours,
adjacency and paths.

Have an understanding of fundamental image processing techniques
(namely, image enhancement and image segmentation) used in
processing medical images.

Able to perform spatial image filtering (convolution) given different
filtering masks/kernels such as Averaging, Sobel, Prewitt, and
Laplacian.
Why do we need digital image processing?

To improve pictorial information for human
interpretation

To process image data for storage, transmission, and
representation for machine perception.
Digital Image Processing
Definition of an image

An image maybe defined as a 2-dimensional function,
f(x, y) where x and y are spatial (plane) coordinates.

The amplitude of f at any pair of coordinates (x, y) is
called the intensity or gray level of the image at that
point.

An image is called a digital image when x, y, and the
amplitude values of f are all finite, discrete quantities.

Pixels is the term to denote the elements of a digital
image. (a digital image is composed of a finite number
of elements, each of which has a particular location
and value)
Fundamentals of image and signal
processing
Coordinate convention used in the course
Fundamentals of image and signal
processing
For example,
1. Coordinates at the origin are
(x, y) = (0, 0)

2. The next coordinates along
the first row of the image is
(x, y) = (0, 1)
signify second sample
along the first row.
A digital image can be represented in a compact matrix form:
|
|
|
|
|
|
.
|
\
|

=
) 1 , 1 ( ... ) 1 , 1 ( ) 0 , 1 (
. . .
: : :
) 1 , 1 ( ... ) 1 , 1 ( ) 0 , 1 (
) 1 , 0 ( ... ) 1 , 0 ( ) 0 , 0 (
) , (
N M f M f M f
N f f f
N f f f
y x f
The right side of this equation is by definition a digital image.

Each element of this matrix array is called a pixel.

Representing digital images
in Matrix form
The digitization process requires decisions about values of M, N and
for the number of discrete gray levels, L, allowed for each pixel.

M and N have to be positive integers.

Due to processing, storage, and sampling hardware considerations,
L, the number of gray levels typically is an integer power of 2:
k
L 2 =
Assumption: the discrete levels are equally spaced and that they
are integers in the interval [0, L-1]
The number of bits (b) required to store a digitized image is given by
When M = N, this equation becomes
k N b =
2
k N M b =
The number of bits required to store square images with various
values of N and K. The number of gray levels corresponding to each
value of k is shown in parentheses.

When an image can have 2
k
gray levels, it is common practice to
refer to the image as a k-bit image for example, an image with 256
possible gray-level values is called an 8-bit image.
Image Storage Sizes

Digital images used up disk space and system bandwidth.

A simple 8-bit grey scale image containing 256x256 pixels requires
65536 bytes of computer storage.

For example
A 256 x 256 x 128 MR volume at 12 bits/pixel requires 12MB of
storage!

The rapid increase of storage requirements with dimensionality is one
aspect of a general issue with processing N-dimensional data known
as the curse of dimensionality.
Medical Image File Formats

In medical imaging, there is a special format known as DICOM
3.0. DICOM stands for Digital Image COmmunications in
Medicine.

DICOM 3.0 represents a standard for the medical image data.

For all DICOM 3.0 images, the header is very long, and contains
information on the patient that has been scanned, the scan
orientation, the name of the physician/radiographer/sonographer,
other patient details, scan parameters, pixel resolution, etc.

All of this information is vital in correctly interpreting a medical
image.
A widely used definition of Spatial resolution is simply the smallest
number of discernible line pairs per unit.

Gray-level resolution similarly refers to the smallest discernible
change In gray level.

The number of gray levels is usually an integer power of 2. The most
common number is 8 bits, with 16 bits being used in some
applications where enhancement of specific gray level ranges is
necessary, for example in medical images.
Spatial and Gray-level Resolution
What is a binary image?

A binary image is one where
each pixel can take any one of two
values, typically represented as 0
or 1.

Binary images are often produced
as the result of some decision
process on pixels of an input
image - used to specify regions in
the image where some property or
condition is true.

Binary Images
Pixels in the output image are set to 1 if the corresponding input image
pixel satisfies some criterion, and are set to 0 otherwise.
Binary images are very widely used in image processing.

For example, if one needs to write an algorithm to identify and count
cells in a histology image, at some stage in the algorithm it is quite
usual to produce an intermediate binary image whose pixels contain
ones where there is a cell present, and zeros where there is no cell
present. Counting up the "blobs" of ones can then yield an estimate
for the number of cells. These cell masks can be used to keep a
record of where the cells are in the original image for further
processing.

Binary Images
An example of counting
the number of rice
Original Binary
Neighbours of a Pixel

A pixel p at coordinates (x, y) has four horizontal and vertical
neighbours whose coordinates are given by
) 1 , ( ), 1 , ( ), , 1 ( ), , 1 ( + + y x y x y x y x
This set of pixels, called the 4-
neighbours of p, is denoted by
N
4
(p).

Each pixel is a unit distance from
(x, y).

(x-1,y)

(x,y-1)

(x,y)

(x,y-1)

(x+1,y)
y
x
N
4
Neighbours of a Pixel

The four diagonal neighbours of p, N
D
(p), have coordinates
) 1 , 1 ( ), 1 , 1 ( ), 1 , 1 ( ), 1 , 1 ( + + + + y x y x y x y x
These points together with the 4-neighbours, are called the 8
neighbours of p, denoted by N
8
(p).


(x-1,y-1)

(x-1,y)

(x-1,y+1)

(x,y-1)

(x,y)

(x,y+1)

(x+1,y-1)

(x+1,y)

(x+1,y+1)

(x-1,y-1)

(x-1,y+1)

(x,y)

(x+1,y-1)

(x+1,y+1)

N
D
N
8
Connectivity

Connectivity between pixels is important in the determination of
boundaries.

To establish if two pixels are connected it must be determined
if they are neighbours
if their gray levels satisfy a specified criterion of similarity
(for example, if their gray levels are equal).

For example,
Consider a binary image with values 0 and 1,

Two pixels may be 4-neighbours, but they are said to be connected
only if they have the same value.
Adjacency (Connectivity)

Let V be the set of gray-level values used to define adjacency.

In a binary image, V = {1}if we are referring to adjacency of pixels
with value 1.

In a gray scale image, set V typically contains more elements.
For example, in the adjacency of pixels with a range of possible
gray-level values 0 to 255, set V could be any subset of these
256 values.

Consider 3 types of adjacency:

1. 4-adjacency. Two pixels p and q with values from V are 4-adjacent if
q is in the set N
4
(p).

2. 8-adjacency. Two pixels p and q with values from V are 8-adjacent if
q is the set N
8
(p).

3. M-adjacency (mixed adjacency). Two pixels p and q with values from
V are m-adjacent if
1. q is in N
4
(p), or
2. q is in N
D
(p) and the set N
4
(p) N
4
(q) has no pixels whose
values are from V.

Adjacency

Mixed adjacency is a modification of 8-adjacency. It is introduced to
eliminate the ambiguities that often arise when 8-adjacency is used.

For example
Considering the pixel arrangement shown in Figure for V= {1}. The
three pixels at the top of Figure show multiple (ambiguous) 8-
adjacency. This ambiguity is removed by using m-adjacency.
Figure. (a) Arrangement of pixels; (b) pixels that are 8-adjacent
(shown dashed) to the center pixel; (c) m-adjacency
Adjacency

Two image subsets S
1
and S
2
are adjacent if some pixel in S
1
is
adjacent to some pixel in S
2
. It is understood here and in the
following definitions that adjacent mean 4-, 8- or m- adjacent.

Example Problem

Consider the two image subsets, S1 and S2, shown in the following figure.
For V = {1}, determine whether these two subsets are
(a) 4-adjacent
(b) 8-adjacent
or (c) m-adjacent
Solution
Let p and q be as shown in the Figure below.
(a) S
1
and S
2
are not 4-connected because q is not in the set N
4
(p);
(b) S
1
and S
2
are 8-connected because q is in the set N
8
(p);

(c) S
1
and S
2
are m-connected because (i) q is in N
D
(p) and (ii) the set
N
4
(p) n N
4
(q) is empty.
A digital path (or curve) from pixel p with coordinates (x, y) to pixel q
with coordinates (s, t) is a sequence of distinct pixels with coordinates
) , ( ),..., , ( ), , (
1 1 0 0 n n
y x y x y x
n. i 1 for adjacent are ) y , (x and
) y , (x pixels and ), , ( ) , ( ), , ( ) , ( where
1 - i 1 - i
i i 0 0

= = t s y x y x y x
n n
N is the length of the path.

If (x
0
,y
0
)=(x
n
,y
n
), the path is a closed path.

4-, 8-, m- paths depends on the type of adjacency specified.

Referring to figure(b) below between the northeast and southeast
points are 8-paths, and the path in figure(c) is an m-path.
Figure. (a) Arrangement of pixels; (b) pixels that are 8-adjacent
(shown dashed) to the center pixel; (c) m-adjacency
Example Problem

Consider the image segment shown.

Let V = {0, 1}and compute the lengths of the shortest 4-, 8-, and m-
path between p and q. If a particular path does not exist between
these two points, explain why.
Solution
When V = {0; 1}, 4 path does not exist between p and q because it is
impossible to get from p to q by traveling along points that are both
4adjacent and also have values from V . Figure (a) shows that it is not
possible to get to q.
The shortest 8 path is shown in Figure (b); its length is 4. The length of
the shortest m- path (shown dashed) is 5. Both of these shortest paths
are unique in this case.
For pixels p, q and z, with coordinates (x, y), (s, t) and (v,
w) respectively, D is a distance function if

(a) D(p, q) > 0 (unless p = q)
(b) D(p, q) = D(q, p), and
(c) D(p, z) < D(p, q) + D(q, z)

The Euclidean distance between p and q is defined as
| |
2
1
2 2
) ( ) ( ) , ( t y s x q p D
e
+ =
Distance Measures
The D
4
distance (also called city-block distance) between p and q is
defined as
t y s x q p D + = ) , (
4
2
2 1 2
2 1 0 1 2
2 1 2
2
Distance Measures
The D
8
distance (also called chessboard distance) between p and q
is defined as
( ) t y s x q p D = , max ) , (
8
2 2 2 2 2
2 1 1 1 2
2 1 0 1 2
2 1 1 1 2
2 2 2 2 2
Image Enhancement in the spatial
domain
Objective of enhancement

To process an image so that the result is more suitable than the original
image for a specific application.

Image enhancement techniques are very much problem oriented.

For example,

A method that is quite useful for enhancing X-ray images may not be
necessarily be the best approach for enhancing pictures of Mars
transmitted by a space probe.
domain
Image enhancement approaches fall into 2 board categories:

Spatial Domain approaches are based on direct
manipulation of pixels in an image.

Frequency Domain techniques are based on modifying
the Fourier transform of an image.

We will only cover Spatial Domain methods in this course
domain
Spatial domain refers to the pixels composing an image.

Procedures operate directly on these pixels.

Spatial domain processes will be denoted by the expression
)] , ( [ ) , ( y x f T y x g =
where
f(x, y) is the input image
g(x, y) is the processed image
T is an operator on f, defined over some neighborhood of (x, y).
domain
Principal Approach
Using a square or rectangular subimage area centered at (x, y) in
defining a neighborhood about a point (x, y)

The center of the subimage is moved from pixel to pixel starting, say, at
the top left corner.
domain
Figure. A 3 x 3 neighborhood about a point (x, y) in an image
The operator T is applied at each
location (x, y) to produce the output,
g, at that location. The process
utilizes only the pixels in the area of
the image spanned by the
neighborhood.
The simplest form of T is when the neighborhood is of size 1 x 1.
In this case, g depends only on the value of f at (x, y), and T becomes
gray-level transformation function of the form

) (r T s =
Where
r variables denotes the gray level of f(x,y) at any point (x, y)
s variable denotes the gray level of g(x,y) at any point (x, y)
domain
For example,
If T(r) has the form shown, the transformation would produce an image
of higher contrast than the original by
darkening the levels below m
brightening the levels above m in the original image.

In this technique, known as contrast stretching, the values of r below
m are compressed by the transformation function into a narrow range
of s, toward black. The opposite effect takes place for values of r above
m.
domain
Figure. Gray level
transformation functions for
contrast enhancement
In the limiting case shown in Figure, T(r) produces a two-level (binary)
image.

A mapping of this form is called a Thresholding function.
domain
Figure. Gray level
transformation
functions for contrast
enhancement
Example of Thresholding
domain
Original MR image Thresholding using m = 150
The general approach is to use a function of the values of f in a
predefined neighborhood of (x,y) to determine the value of g at (x,y).

One of the principal approaches in this formulation is based on the use
of so-called masks (also referred to as filters)

domain
What is a mask?

A mask is basically a small 2-D array in
which the values of the mask coefficients
determine the nature of the process.

This approach often are referred to as
mask processing or filtering.

Some examples of Basic Gray Level Transformations:

1. Image negatives
2. Log Transformations (more on Fourier transform)
3. Power-Law Transformations
domain
Image negatives

The negative of an image with gray levels in the range [0, L 1] is
obtained by using the negative transformation which is given by the
expression
r L s = 1
Reversing the intensity levels of an image in this manner produces the
equivalent of a photographic negative.

This type of processing is particularly suited for enhancing white or
gray detail embedded in dark regions of an image.

domain
domain
Example
Original digital mammogram.
Negative image obtained using the
negative transformation
The original image is a digital mammogram showing a small lesion. In
spite of the fact that the visual content is the same in both images, note
how much easier it is to analyze the breast tissue in the negative image in
this particular case.
Log Transformations (more on fourier transform)

The general form of the log transformation is
) 1 log( r c s + =
Where c is a constant, and it is assumed that r > 0.

domain
domain
The shape of the log curve in
Figure shows that this
transformation maps a narrow
range of low gray-level values
in the input image into a wider
range of output levels.

The log function has the
important characteristic that it
compresses the dynamic
range of images with large
variations in pixel values.
This operation is carried out on a pixel-by pixel basis between two or
more images. This excludes the logic operation NOT, which is
performed on a single image

For example,
Subtraction of two images would result in a new image whose pixel at
coordinates (x, y) is the difference between the pixels in that same
location in the two images being subtracted.

Pixel values are processed as strings of binary numbers in logic
operations on gray-scale images,

Main logic operators
AND
OR
NOT

Enhancement using Arithmetic/Logic
Operation
NOT Logic Operator

The NOT logic operator performs the same function as the negative
transformation

Operation
Original digital
mammogram.
Negative image obtained using
the negative transformation
AND and OR operations
Used for selecting subimages in an image (i.e. masking).

Masking is referred to as region of interest processing and is used
primarily to isolate an area for processing.
Operation
(a) Original Image
(b) AND image mask
(c) Result of the AND
operation on images (a)
and (b)
(d) Original image
(e) OR image mask.
(f) Result of operation OR
on images (d) and (e)

Subtraction and addition are the most useful arithmetic operation.

Image Subtraction

The difference between two images f(x, y) and h(x, y) is expressed as

) , ( ) , ( ) , ( y x h y x f y x g =
Is obtained by computing the difference between all pairs of
corresponding pixels from f and h.

The key usefulness of subtraction is the enhancement of differences
between images.

Operation
One of the most commercially successful and beneficial uses of image
subtraction is in the area of medical imaging called mask mode
radiography.
Operation
Enhancement by image subtraction. (a) Mask image. (b) An image (taken after injection of a
contrast medium into the bloodstream) with mask subtracted out.
Spatial filtering process involve moving the filter mask from point to
point in an image.

At each point (x, y), the response of the filter at that point is calculated
using a predefined relationship.

For linear spatial filtering, the response is given by a sum of products
of the filter coefficients and the corresponding image pixels in the area
spanned by the filter mask.

Spatial filtering is used in smoothing, edge detection, noise removal,
etc.
Basics of Spatial Filtering
For example, the 3 X 3 mask shown, the result (or response), R, of
linear filtering with the filter mask at a point (x, y) in the image is

) 1 , 1 ( ) 1 , 1 ( ) , 1 ( ) 0 , 1 ( ... ) , ( ) 0 , 0 (
... ) , 1 ( ) 0 , 1 ( ) , ( ) 1 , 1 (
+ + + + + + +
+ + =
y x f w y x f w y x f w
y x f w q y q x f w R
which is the sum of products of the mask
coefficients with the corresponding pixels directly
under the mask

Note that the coefficient w(0,0) coincides with the
image value f(x, y), indicating the mask is centered
at (x, y) when the computation of the sum of the
products takes place.
Mask coefficients showing coordinate
arrangement
Pixels of image section under mask
In general, linear filtering of an image f of size M x N with a filter mask of
size m x n is given by the expression:
( )

= =
+ + =
a
a s
b
b t
t y s x f t s w y x g ) , ( ) , ( ,
2 / ) 1 ( = m a
2 / ) 1 ( = n b
1 ,...., 2 , 1 , 0 = M x
1 ,...., 2 , 1 , 0 = N y
where
The process of linear filtering is similar to a frequency domain concept
called convolution.
The response, R, of an m x n mask at any point (x, y) given by the
following expression:
=
=
+ + + =
mn
i
i i
mn mn
z w
z w z w z w R
1
2 2 1 1

...
where
the ws are mask coefficients
the zs are the values of the image gray levels corresponding to those
coefficients
mn is the total number of coefficients in the mask.

For the 3 x 3 general mask, the response at any
point (x, y) in the image is given by

=
=
+ + + =
9
1
9 9 2 2 1 1

...
i
i i
z w
z w z w z w R
Smoothing Spatial Filters are used for blurring and for noise
reduction.

Blurring is used in preprocessing steps, for example, removal of small
details from an image prior to object extraction, and bridging of small
gaps in lines or curves.

The output of a smoothing, linear spatial filter is the average of the
pixels contained in the neighborhood of the filter mask.

These filters are also call averaging filters/low pass filters

Smoothing Spatial Filters
Basic idea
The value of every pixel in an image is replaced by the average of the
gray levels in the neighborhood defined by the filter mask.

As a result, the resulting image has reduced sharp transitions in
gray levels.

Because random noise typically consists of sharp transitions in gray
levels, smoothing is applied to reduce noise.

However, edges (which are desirable) are characterized by sharp
transitions in gray levels, so averaging filters have the undesirable
side effect that they blur edges.
Example of two 3 x 3 smoothing filters
Example of 3 x 3 smoothing filters (1)

This produces the standard average of the pixels
under the mask.
=
=
+ + + =
9
1
9 9 2 2 1 1

...
i
i i
z w
z w z w z w R
Substituting the coefficient of the mask into the above equations
=
=
9
1
9
1
i
i
z R
Which is the average of the gray levels of the pixels in the 3 x 3
neighborhood defined by the mask.
Example of 3 x 3 smoothing filters (2)

This produces a weighted average.

This give more importance (weight) to some
pixels at the expense of others.

The center point has the highest weight and
reducing the weight as function of increasing
distance from the origin.

This reduces blurring in the smoothing process.
What is the effect of smoothing when the size of the filter increases?
Original n = 3
n = 5 n = 9
n = 35 n = 15
The response of nonlinear spatial filters is based on ordering (ranking)
the pixels contained in the image area encompassed by the filter, and
then replacing the value of the center pixel with the value determined
by the ranking result.

An example is the median filter

The median filter replaces the value of a pixel by the median of the
gray levels in the neighborhood of that pixel (the original value of the
pixel is included in the computation of the median)

Median filter for certain types of random noise (e.g. impulse noise)
Provide excellent noise-reduction capabilities, with considerably
less blurring than linear smoothing filters of similar size
Order Statistics filters
What is median? Recall

The median of a set of values is such that half the values in the set
are less than or equal to , and half are greater than or equal to .

Steps in performing median filtering at a point in an image

1. Sort the values of the pixel in question and its neighbors
2. Determine their median
3. Assign the median value to that pixel at the point.
Median filters
For example

Suppose that a 3 x 3 neighborhood has values (10, 20, 20, 20, 15, 20,
20, 25, 100).

These values are sorted as (10, 15, 20, 20, 20, 20, 20, 25, 100).

In a 3 x 3 neighborhood the median is the 5
th
largest value. Thus, the
median will be 20.

The principal function of the median filters is to force points with distinct
gray levels to be more like their neighbours.
Median filters

Figure 1. Application of the median filter
Median filters
Other types of order statistics filters

1. Max filter, useful in finding the brightest points in an image

2. Min filer, used in finding the darkest points in an image

3. Mean filter, simply smoothes local variations in an image
Order Statistics filters
Image Segmentation
Segmentation subdivides an image into its constituent regions or
objects.

Segmentation accuracy determines the eventual success or failure of
computerized analysis procedures.

Image segmentation algorithms generally are based on one of two
basic properties of intensity values: Discontinuity and similarity

Discontinuity : the approach is to partition an image based on abrupt
changes in intensity, such as edges in an image.

Similarity: based on partitioning an image into regions that are similar
according to a set of pre defined criteria, eg. Thresholding.
Image Segmentation
Application of Segmentation

Quantifying a pathology

The strength of medical imaging is that it provides very visual
information that often gives very important localisation information.

The downside is that quantifying a pathology from an image (which
is normally interpreted by a radiologist) is not easy. Indeed
diagnoses sometimes lack repeatability, and standardizing the
diagnostic process, particularly in uncertain cases, becomes difficult.

Because many images are now digital by nature, the possibility of
automated or computer-assisted diagnosis is theoretically possible.
Image Segmentation
Application of Segmentation

Rendering 3D data
Figure (a) Single MR T1 weighted image slice. (b) Rendered using segmentation of structures
from each slice of the volume, placing these together, constructing surfaces and finally
rendering in 3D. The difficult part is the segmentation!
Image Segmentation
Thresholding can be defined as a labelling operation on an image.

For gray scale images, thresholding is a means of distinguishing
pixels that have higher intensity from those of lower intensity.

If g(m; n) represents a binary version of the gray scale image f(m;
n), then we can express the thresholding operation as g(m; n) if
f(m; n) > T then g(m; n) = 1 else g(m; n) = 0

This is a simple example of decision process upon an input (gray-
scale) image which yields a binary image.

Thresholding is a well known way of re-quantizing an image into just
two levels; a good way of reducing data bandwidth.
Thresholding
For example, applying a thresholding operation on a gray scale image
containing pixel values in the range 0 - 255.

An image (as shown) contains pixels which correspond to an object of
interest, and also pixels which correspond to the background. We
would like to be able to label which pixels belong to the object, and
which to the background. This represents the simplest example of
image segmentation.

Thresholding
Figure: (a) Leaf on Grass, grey scale image of object and background. (b) Thresholded at a gray level of 140

Techniques to detect three basic types of gray-level discontinuities in a
digital image:
Points
Lines
Edges

The common way to look for discontinuities is to run a mask through
the image in the manner described in previous lecture.

For a 3 x 3 mask, this procedure involves computing the sum of
products of the coefficients with the gray levels contained in the region
encompassed by the mask.
Detection of Discontinuities
Recall

The response of the mask at any point in the image is given by
=
=
+ + + =
9
1
9 9 2 2 1 1

...
i
i i
z w
z w z w z w R
Where z
i
is the gray level of the pixel associated with mask coefficient w
i
.
The response of the mask is defined with respect to its center location.
Point Detection

Using the mask shown, a point is detected at the location on which the
mask is centered if
T R
Where T is a pre-defined non negative threshold and R is given by the
equation below.

This formulation measures the weighted differences between the center
point and its neighbors.

=
=
+ + + =
9
1
9 9 2 2 1 1

...
i
i i
z w
z w z w z w R
Point
detection
mask
Point Detection (For example)
Single pixel discontinuities
Line Detection

Consider the mask shown
If the first mask was moved around an image, it would respond more
strongly to lines (one pixel thick) oriented horizontally.

With a constant background, the maximum response would result
when the line passed through the middle row of the mask.
Line Detection (Example)

A digitized (binary) portion of a wire-
bond mask for an electronic circuit.

Suppose we are interested in finding all
the lines that are one pixel thick and
are oriented at -45. For this purpose,
we choose the following mask.
Line Detection (Example)
In order to determine which lines best fit the mask, we simply threshold
this image.
Edge Detection

An ideal edge is a set of connected
pixels (in the vertical direction here).

The edge is located at an orthogonal
step transition in gray level (as shown)
Edge Detection

In practice, edges are more closely modeled as
having a ramplike profile.

This is caused by imperfections determined by
factors such as quality of the image acquisition
system, the sampling rate etc.

The slope of the ramp is inversely proportional
to the degree of blurring in the edge.

The edge is no longer a thin (one pixel thick)
path. The thickness of the edge is determined
by the length of the ramp, as it transits from an
initial to a final gray level.
First and second-order digital derivatives can be use for the detection
of edges in an images.
Edge Detection
Two regions separated
by a vertical edge.

Detail near the edge,
showing a gray level
profile, and the first and
second derivatives of the
profile.
First derivative

Moving from left to right along
the profile, the first derivative is
positive at the points of
transition into and out of the
ramp.

It is constant for points in the
ramp

It is zero in areas of constant
gray level.
Two regions separated by a vertical
edge.
Edge Detection - First derivative
Second Derivative

is positive at the transition
associated with the dark side of the
edge

Is negative at the transition
associated with the light side of the
edge

Is zero along the ramp and in areas
of constant gray level.

The sign of the derivatives would be
reversed for an edge that transitions
from light to dark.
Two regions separated by a vertical edge.
Edge Detection - Second derivative
Observations

The magnitude of the first derivative can be used to detect the
presence of an edge at a point in an image (i.e., to determine if a point
is on a ramp)

The sign of the second derivative can be used to determine whether
an edge pixel lies on the dark or light side of an edge.

Edge Detection
For a function f(x, y), the gradient of f at coordinates (x, y) is defined as
the two-dimensional column vector
(
(
(
(
=
(
=
dy
f
x
f
G
G
y
x
f
The magnitude of this vector is given by
| |
2 / 1
2 2
) f (
y x
G G mag f + = =
It is a common practice to refer to as the gradient.
f
Edge Detection
The direction of the gradient vector also is an important quantity.
Let (x, y) represent the direction angle of the vector at (x, y). Then,
from vector analysis,
f
Where the angle is measured with respect to the x-axis. The direction of
an edge at (x, y) is perpendicular to the direction of the gradient vector at
that point.
|
|
.
|
\
|
=

x
y
G
G
y x
1
tan ) , (
( ) ( )
3 2 1 9 8 7
z z z z z z G
x
+ + + + =
( ) ( )
7 4 1 9 6 3
z z z z z z G
y
+ + + + =
Edge Detection
Computation of the gradient of an image is based on obtaining the partial
derivatives and at every pixel location.
x f /
An approach using masks of size 3 x 3 is given by
y f /
( ) ( )
3 2 1 9 8 7
z z z z z z G
x
+ + + + =
( ) ( )
7 4 1 9 6 3
z z z z z z G
y
+ + + + =
Edge Detection
Example of two most commonly used 3 x 3 masks for computing the gradient

These masks are used to obtain the
gradient component Gx and Gy.

The prewitt masks are simpler to
implement than the sobel masks

But sobel have slightly superior
noise-suppression characteristics, an
important issue when dealing with
derivatives.

Note that the coefficients in all the
masks show sum to 0!

Edge Detection
The computational burden of implementing Equation over an entire
image is not trivial, it is common practice to approximate the magnitude
of the gradient by using absolute values instead of squares and square
roots:
y x
G G f +
This equation is simpler to compute and it still preserves relative changes
in gray levels

However, the resulting filters will not be isotropic (invariant to rotation) in
general. But Prewitt and Sobel masks can give isotropic results only for
vertical and horizontal edges.
| |
2 / 1
2 2
) f (
y x
G G mag f + = =
Edge Detection
Examples
Edge Detection
Original image |G
x
| component of the gradient in the x-
direction.
|G
y
| component of the gradient in
the y-direction.
Gradient image, |G
x
| +|G
y
|
The Laplacian of a 2-D function f(x, y) is a second order derivative
defined as
2
2
2
2
2
y
f
x
f
f
=
Digital approximations to the Laplacian for a 3 x 3 region, one of the
two forms encountered most frequently in practice is
( )
8 6 4 2 5
2
4 z z z z z f + + + =
Digital approximations including the diagonal
neighbors is given by
( )
9 8 7 6 4 3 2 1 5
2
8 z z z z z z z z z f + + + + + + + =
Edge Detection
Masks for implementing the two equations
( )
8 6 4 5 5
2
4 z z z z z f + + + =
( )
9 8 7 6 4 3 2 1 5
2
8 z z z z z z z z z f + + + + + + + =
These masks are isotropic for rotation increments of 90and 45 respectively.
Edge Detection
There are problems using Laplacian in
its original form for edge detection

1. Sensitive to noise
2. Magnitude of the Laplacian
produces double edges this
complicates segmentation
3. Unable to detect edge direction

Role of Laplacian
1. Using its zero-crossing property for
edge location
2. Complementary purpose of
establishing whether a pixel is on
the dark or light side of an edge
Edge Detection
Example
Laplacian
Zero Crossing
Gradient Operator Sobel Original
Edges in the zero crossing image are thinner than the gradient edges.
Edge Detection
What is histogram of a image?

Histogram of a digital image with gray levels in the range [0, L 1] is a
discrete function given by

Histogram
k k
n r h = ) (
where r
k
is the k
th
gray level and n
k
is the number of pixels in the image
having gray level r
k
.
Histogram
A histogram can be normalized by dividing each of its
values by the total number of pixels in the images, denoted
by n.
Thus a normalized histogram is given by p(r
k
) = n
k
/ n, for k
= 0, 1, , L - 1.
p(r
k
) gives an estimate of the probability of occurrence of
gray level r
k
.
The sum of all components of a normalized histogram is
equal to 1.
Histograms are the basis for numerous spatial domain processing
techniques.
Image enhancement
Image compression
Image segmentation

The pollen image shown in four basic gray-level characteristics: dark,
light, low contrast and high contrast. these histogram plots are simply
plots of h(r
k
) = n
k
versus r
k
or p(r
k
) = n
k
/ n versus r
k
.

The horizontal axis of each histogram plot corresponds to gray
level values, r
k

The vertical axis corresponds to values of h(r
k
) = n
k
or p(r
k
) = n
k
/n
if the values are normalized.

Histogram
The component of
the histogram of the
dark image are
concentrated on the
low(dark) side of
the gray scale.
The component of
the histogram of the
bright image are
concentrated on the
high(bright) side of
the gray scale.
Histogram
An image with low
contrast has a
histogram that will be
narrow and will be
centered toward the
middle of the gray scale.
An image with high
contrast has a histogram
that cover a broad range
of the gray scale, and
the distribution of pixels
is not too far from
uniform. (This is what we
like to see!)
Histogram

BG3104 Set 1 2014-15

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

BG3104 Set 1 2014-15

Загружено:

Авторское право:

Доступные форматы

Division of Bioengineering, School of Chemical and Biomedical Engineering 1

Вам также может понравиться