imgprocessingproject

Color Image Edge detection Algorithm

Based on Circular Shifting

RAJENDRA KUMAR YADLA

1

Chapter – 1

INTRODUCTION TO IMAGE PROCESSING

1.1 Introduction:

EARLY days of computing, data was numerical. Later, textual data

became more common. Today, many other forms of data: voice, music,

speech, images, computer graphics, etc. Each of these types of data is

signals. A Signal is a function that conveys information. Before going to

know about Digital Image Processing, let us discuss the history from

where it exists.

First the issue of digital image processing appeared relatively late

in computer history, it had to wait for the arrival of the first graphical

operating systems to become a true matter. Secondly, digital image

processing requires the most careful optimizations and especially for real

time applications. As long as people have tried to send or receive the

message through electronic media: telegraphs, telephones, television,

radar, etc. there has been the realization that these signals may be

affected by the system used to acquire, transmit, or process them.

Sometimes, these systems are imperfect and introduce noise, distortion,

or other artifacts.

Understanding the effects these systems have and finding ways to

correct them is the fundamental of signal processing. That is, we

specifically introduce the information content into the signal and hope to

extract it out later. Sometimes, these man-made signals are encoding of

natural phenomena (audio signal, acquired image, etc.), but sometimes

we can create them from scratch (speech generation, computer

generated music, computer graphics). Finally, we can merge these

technologies together by acquiring a natural signal, processing it, and

2

then transmitting it in some fashion. This fashion is called Digital Image

Processing.

Vision allows humans to perceive and understand the world

surrounding us. Computer vision aims to duplicate the effect of human

vision by electronically perceiving and understanding an image. Giving

computers the ability to see is not an easy task - we live in a three

dimensional (3D) world, and when computers try to analyze objects in 3D

space, available visual sensors (e.g., TV cameras) usually give two

dimensional (2D) images, and this projection to a lower number of

dimensions incurs an enormous loss of information. In order to simplify

the task of computer vision understanding, two levels are usually

distinguished as Low-level image processing and High-level image

processing.

1.2 What is Digital Image Processing?

According to Erza Pound, “Image is an intellectual and emotional

complex in an instant of time". The term Image, refers to a two-

dimensional light intensity function f(x, y), where x and y denote spatial

coordinates and the value of f at any point (x, y) is proportional to the

brightness (or gray level) of the image at that point. The x axis is usually

the horizontal axis. The x axis is usually the vertical axis. The origin of the

coordinate system is usually the upper left corner of the image. The x axis

is positive from left to right. The y axis is positive from the top to the

bottom of the image.

Image processing is any form of information processing for which

both the input and output are images, such as photographs or frames of

video. Image processing modifies pictures to improve them

(enhancement, restoration), extract information (analysis, recognition),

and change their structure (composition, image editing). It is a subclass of

3

http://en.wikipedia.org/wiki/Information_processing

signal processing concerned specifically with pictures. Improve image

quality for human perception and/or computer interpretation.

A Digital image is an image f(x, y) that has been discretized both in

spatial coordinates and brightness, which can be considered as a matrix

with rows and columns indices identify a point in the image, the value of

the matrix element identifies the grey level at that point.

Digital Image Processing refers to processing of digital images by

means of a digital computer. Digital Image Processing (DIP) is a

multidisciplinary science that borrows principles from diverse fields such

as optics, surface physics, visual psychophysics, computer science and

mathematics. Digital image processing is a subset of the electronic

domain wherein the image is converted to an array of small integers,

called pixels, representing a physical quantity such as scene radiance,

stored in a digital memory, and processed by computer or other digital

hardware.

Digital Images are divided into three type, they are Binary image,

Monochrome Image, Color Image. Binary Image can only encode two

levels, namely bright as 0’s and dark as 1’s. In Monochrome image the

array values are represented with in some range of values. Color image

records both brightness and color of each pixel i.e. for each pixel in an

image it records the contents of Red, Green, Blue.

There are a few fundamental steps of digital image processing that

can be applied to images for different purposes and possibly with different

objectives. They are:

4

Image Acquisition: Generally, the image acquisition stage

involves preprocessing, such as scaling. Image is captured by a sensor

(such as a monochrome or color TV camera) and digitized. If the output of

the camera or sensor is not already in digital form, an analog-to-digital

converter digitizes it.

Example: Frame grabber only needs circuits to digitize the electrical signal

from the imaging sensor to store the image in the memory (RAM) of the

computer.

5

Image Enhancement: It is the simplest and most appealing

areas of digital image processing. Basically, the idea behind enhancement

techniques is to bring out detail that is obscured, or simply to highlight

certain features of interest in an image. Enhancement is a very subjective

area of image processing.

Example:

Image Restoration: It improves the appearance of an image.

However, unlike enhancement, which is subjective, image restoration is

objective, in the sense that restoration techniques tend to be based on

mathematical or probabilistic models of image degradation.

Example:

Color Image Processing: This area is gaining more importance

because of the significant increase in the use of digital images over the

Internet.

Wavelets: This is a foundation for representing images in various

degrees of resolution. Unlike the Fourier Transforms, whose basic

6

functions are sinusoids, wavelet transform is based on small waves called

“wavelets” of varying frequency and limited duration.

Compression: This reduces the storage space required to save an

image, or the bandwidth required to transmit it. Image compression is

familiar to most users of computers in the form of image file extensions,

such as the jpg file extension used in the JPEG (Joint Photographic Experts

Group) image compression standard.

Morphological processing: These are the tools for extracting

image components that are useful in the representation and description of

shape. This is the stage in which the output of the process is not the

image but the image attributes.

Image Segmentation: This partitions an image into its

constituent parts or objects. In general, the more accurate the

segmentation, the more likely recognition is to succeed.

Representation & Description: It makes a decision whether

the data should be represented as a boundary or as a complete region.

Boundary representation focuses on external shape characteristics, such

as corners and inflections. Region representation focuses on internal

properties, such as texture or skeleton shape.Description, also called

feature selection, deals with extracting attributes that result in some

quantitative information of interest or are basic for differentiating one

class of objects from another.

Object Recognition: It is the process that assigns a label (e.g.,

“vehicle”) to an object based on its descriptors.

Knowledge Base: Knowledge about the problem domain is coded

into an image processing system in the form a knowledge databases.

7

1.3 Applications of Digital Image Processing:

The many applications of digital image processing include:

astronomy, ultrasonic imaging, remote sensing, video communications

and microscopy, among innumerable others. The few applications are as

follows;

1. Image processing technology is used by planetary scientists to

enhance images of Mars, Venus, or other planets.

2. Doctors use this technology to manipulate CAT scans and MRI

images.

3. One of the first applications of digital images was digitized

newspaper pictures sent by submarine cable between London and New

York.

4. The major application area of digital image processing techniques is

in problems dealing with machine perception which is focused on

procedures for extracting image information in a form suitable for

computer processing. Examples of the type of information used in

machine perception are statistical moments, Fourier transform

coefficients, and multidimensional distance measures.

5. Image processing in the laboratory can motivate students and make

science relevant to student learning.

6. In Archeology, Image Processing methods are to store the blurred

pictures or the image which is damaged after being photographed.

7. Image Processing is used in Astronomy, Geography, Biology,

Defense, Law enforcement, and so on.

Until a few years ago, segmentation techniques were

proposed mainly for gray-level images since for a long time these have

been the only kind of visual information that acquisition devices were able

to take and computer resources to handle.

8

1.4 What is Segmentation?

Segmentation distinguishes objects from background that is it

divides an image into parts that have a strong correlation with objects or

areas of the real world contained in the image. Simply, partitioning of an

image into several constituent components is called “Segmentation”.

Another word for object detection is segmentation. The object to be

segmented differs greatly in contrast from the background image.

Segmentation is concerned with splitting an image up into

segments (also called regions or areas) that each holds some property

distinct from their neighbor. This is an essential part of scene analysis in

answering the questions like where and how large is the object, where is

the background, how many objects are there, how many surfaces are

there... Segmentation is a basic requirement for the identification and

classification of objects in scene. The segmentation of structure from 2D

and 3D images is an important step for a variety of image analysis and

visualization tasks.

Image Segmentation is a subset of an expansive field of Computer

Vision which deals with the analysis of the spatial content of an image. In

particular, it is used to separate regions from the rest of the image, in

order to recognize them as objects. Image segmentation is a

computational process and should NOT be treated as a task.

Segmentation is an important part of practically any automated

image recognition system, because it is at this moment that one extracts

the interesting objects, for further processing such as description or

recognition.

Segmentation of an image is in practice the classification of each

image pixel to one of the image parts. If the goal is to recognize black

characters, on a grey background, pixels can be classified as belonging to

9

the background or as belonging to the characters: the image is composed

of regions which are in only two distinct grey value ranges, dark text on

lighter background. The grey level histogram, viz. the probability

distribution of the grey values, has two separated peaks, i.e. is clearly

bimodal. In such a case, the segmentation, i.e. the choice of a grey level

threshold to separate the peaks, is trivial. The same technique could be

used if there were more than two clearly separated peaks.

1.5 Procedure for Segmentation:

The goal of Segmentation is to segment a given gray level image

into regions. First we have to label the various regions of the image.

Secondly, we have to create a region adjacency graph, which will be used

to determine whether two regions are adjacent. Lastly, we have to

compute the mean value of each region. If the mean values of two

adjacent regions differ by less than one threshold, the regions will be

merged as one.

1. The first step is to read a chosen test image as;

2. Secondly, the test image had to be converted to a gray level image so

that thresholding would be possible. The four threshold values of 10, 70,

130, 190, and 250 were hand chosen so that the resulting image would be

an evenly spread 5-level gray picture. This 5-level gray image can be

seen in the following picture.

10

A region adjacency graph is created. This graph will determine

which regions are connected. The mean value of each region will then be

calculated. If the mean values of two adjacent regions differ by less than

a threshold, the regions will be merged. This process will continue until

there are not anymore regions that can be merged.

3. The result of merging the test image can be as;

It can be seen in this image that each of the objects in the

original image has a unique gray-level value. This is done so that each

object can be recognized.

1.6 Types of Segmentation:

Segmentation methods can be divided into three groups according

to the dominant features they employ. They are Thresholding, Edge-

basedsegmentation, Region-based segmentation.

11

1.6.1 Supervised and Unsupervised Segmentation:

Supervised segmentation of color images is based on the palletized

representation of the image and consists of a few steps as, down-

sampling the lattice upon which the image pixels are defined, blurring the

image through low pass filtering in order to average colors lying in almost

uniform regions, exploiting the well- established and fast methods for

color quantization in order to compute color clusters in the RGB space.

Unsupervised color image segmentation is based upon the ideas

where processing of images in their palletized format; using the low

spatial frequency content in the low-low band of the 2-level wavelet

transform of the image; building and thresholding hue and chroma

histograms. The goal of unsupervised segmentation is to create structure

for the data by objectively partitioning the data into homogeneous groups

where the within group object similarity and the between group object

dissimilarity are optimized.

1.6.2. Complete and Partial segmentation:

Complete Segmentation is a one which results in set of disjoint

regions uniquely corresponding with objects in the input image. The whole

class of segmentation can be solved successfully using low-level

processing. In this case, the image commonly consists of contrasted

objects on a uniform background -- simple assembly tasks, blood cells,

printed characters, etc. Here, a simple global approach can be used and

the complete segmentation of an image into objects and background can

be obtained.

Partial Segmentation is a one in which regions do not

correspond directly with image objects. Image is divided into separate

regions that are homogeneous with respect to a chosen property such as

brightness, color, reflectivity, texture, etc. In a complex scene, a set of

possibly overlapping homogeneous regions may result. The partially

segmented image must then be subjected to further processing, and the

12

final image segmentation may be found with the help of higher level

information.

Many segmentation methods are based on two basic properties of

gray-level values and the pixels in relation to their local neighborhood are:

discontinuity and similarity. In the first category, the approach is to

partition an image based on abrupt changes in gray level. The principal

areas of interest in this category are detection of isolated points and

detection of lines and edges in an image. Methods based on pixel

discontinuity are called boundary-based method which relies on the

gradient features at a subset of the spatial positions of an image.

The principal approaches in the second category are based on

thresholding and label region algorithm. Methods based on pixel similarity

are called region-based method which relies on the homogeneity of

spatially localized features.

1.6.3 Region Based Segmentation:

Region finding techniques are based on the absence of

discontinuities within the regions themselves. The region based

approaches try to form clusters of similar pixels. Initially, clusters can be

found by grouping adjacent pixels with same value.

Various approaches are distinguished by the particular rules

used to decide when to merge regions and when to keep them distinct.

The rules are normally based on some measure of the strength of the

boundary between the regions. The strength of the boundary at any

particular position is based on examination of the values of the pixels

which lie on either side.

If the difference in value exceeds some threshold, the

boundary location is said to be strong. If the difference is less than the

threshold, the boundary location is said to be weak. If enough weak

boundary locations exist, the regions can be merged.

13

Homogeneity is an important property of regions and is used

as the main segmentation criterion in region growing, whose basic idea is

to divide an image in to zones of maximum homogeneity. The criteria for

homogeneity can be based on gray-level, color, texture, shape, model,

etc.

A region-based method usually proceeds as follows: the image

is partitioned into connected regions by grouping neighboring pixels of

similar intensity levels. Adjacent regions are then merged under some

criterion involving perhaps homogeneity or sharpness of region

boundaries.

Over-stringent criteria create fragmentation; lenient ones

overlook blurred boundaries and over merge. Hybrid techniques using a

mix of the methods above are also popular.

Region growing techniques are generally better in noisy

images, where borders are extremely difficult to detect is one of the

advantage of region growing.

1.7 Uses of Segmentation:

1. Image segmentation has been used widely in image processing

applications like, content based image retrieval, image understanding,

object-recognition, browsing and image classification. Recently, the use of

image processing applications have spread over Web and into our daily

lives much more, thus the speed requirements of the applications are

increasing day by day.

2. Segmentation is a tool that has been widely used in medical image

processing and computer vision for a variety of reasons. The goal is to

segment the images with respect to their characteristics such as bone and

tissue types; (or) Segmentation is a tool that has been widely used in

medical image processing and computer vision for a variety of reasons.

14

Chapter – 2

ANALYSIS

Types of Segmentation

2.1 Thresholding:

Thresholding has a long but checkered history in digital image

processing. Thresholding is a commonly used enhancement whose goal is

to segment an image into object and background. A threshold value is

computed above (or below) which pixels are considered “object” and

below (or above) which “background”, and eliminates unimportant

shading variation.

One obvious way to extract the objects from the background is to

select a threshold T that separates these modes. Then, any point (x, y) for

which f(x, y) > T is called an object point; otherwise, the point is called a

background point.. Thresholding can also be done using neighborhood

operations.

Thresholding essentially involves turning a color or grayscale

image into a 1-bit binary image. This is done by allocating every pixel in

15

the image either black or white, depending on their value. The pivotal

value that is used to decide whether any given pixel is to be black or

white is the threshold.

If the threshold value is only dependent on the gray values then it is

called as “Global Thresholding”. If the threshold value is dependent on

the gray values and on some local property then it is called as “Local

Thresholding”. If the threshold value is dependent on the gray values,

some local property and some spatial co-ordinates then it is called as

“Adaptive Thresholding”.

Different ways to choose the threshold value when we covert a gray

image to binary image are;

1. Randomly selecting the gray values of the image

2. Average of all the gray values of the image:

3. Median of given gray values:

4. (Minimum value + Maximum value)/2

Algorithm for Selecting the Threshold value Randomly:

Step 1 : Read image.

Step 2 : Convert image into two dimensional matrixes (eg: 64X64).

Step 3 : Select the threshold value randomly.

16

Step 4 : If grey level value is greater than the threshold value make it

bright,

else ,make it dark.

Step 5 : Display threshold image.

Algorithm for Selecting the Threshold value using Average of all

gray values:



Step 3 : Select the threshold value by calculating average of all the gray values.

Step 4 : If grey level value is greater than the threshold value make it bright, else

make it dark.


Algorithm for Selecting the Threshold value using Median of all

gray values:



Step 3 : Select the threshold value by calculating median of all the gray values.


make it dark.


Algorithm for Selecting the Threshold value using Average of minimum

and maximum gray values:



17

Step 3 : Select the threshold value by calculating the average of Minimum and

Maximum gray values.


make it dark.


Limitations:

1. As hardware cost dropped and sophisticated new algorithms were developed,

thresholding became less important.

2. Thresholding destroys useful shading information and applies essentially

infinite gain to noise at the threshold value, resulting in a significant loss of

robustness and accuracy.

2.2 Edge Based Segmentation:

The edges of an image hold much information in that image. The

edges tell where objects are, their shape and size, and something about

their texture. An edge is where the intensity of an image moves from a

low value to a high value or vice versa. Edges in images are areas with

strong intensity contrasts – a jump in intensity from one pixel to the next.

Edge detecting an image significantly reduces the amount of data and

filters out useless information, while preserving the important structural

properties in an image

Edge detectors are a collection of very important local image pre-

processing methods used to locate (sharp) changes in the intensity function.

18

Edges are pixels where the brightness function changes abruptly. The following

figure shows several typical standard edge profiles;

Edge-based segmentation represents a large group of methods based

on information about edges in the image; it is one of the earliest segmentation

approaches and still remains very important and also relies on edges found in an

image by edge detecting operators -- these edges mark image locations of

discontinuities in gray level, color, texture, etc.

The classification of edge detectors is based on the behavioral study of

the edges with respect to the following operators:

• First Order derivative edge detection

• Second Order derivative edge detection

2.2.1. First Order edge detection Technique:

The first derivative assumes a local maximum at an edge. For a gradient image

f(x, y), at location (x, y), where x and y are the row and column coordinates

respectively, we typically consider the two directional derivatives. The two

functions that can be expressed in terms of the directional derivatives are the

gradient magnitude and the gradient orientation. The gradient magnitude is

defined by,

19

The gradient orientation is also an important quantity. The gradient orientation is

given by,

The angle is measured with respect to the x- axis. The other method of

calculating the gradient is given by estimating the finite difference.

We can approximate this finite difference as;

Let us see the most popular classical gradient-based edge detectors;

2.2.1.1. Robert edge detection operator:

The calculation of the gradient magnitude and gradient magnitude of an

image is obtained by the partial derivatives,

20

at every pixel location. The simplest way to implement the first order partial

derivative is by using the Roberts cross gradient operator. The Roberts operator

masks are;

Therefore the partial derivatives for the above two 2x2 masks are as follows;

The Roberts cross operator provides a simple approximation to the gradient

magnitude:

Using convolution masks, this becomes:

The above two masks are designed to respond maximally to edges

running at 45Â° to the pixel grid, one mask for each of the two perpendicular

21

orientations. The masks can be applied separately to the input image, to produce

separate measurements of the gradient component in each orientation (call

these Gx and Gy). These can then be combined together to find the absolute

magnitude of the gradient at each point and the orientation of that gradient. The

gradient magnitude is given by:

By using the mask given below:

The approximate magnitude is as follows

Likewise the Robert cross edge detector is used.

Algorithm:

Step 1: Read image.

Step 2: Convert image into two dimensional matrix (eg: 64X64).

Step 3: Calculate Threshold value of the image using average method.

Step 4: If grey level value is greater than or equal to the threshold value make it

bright, else make it dark.

Step 5: Display threshold image.

22

Step 6: Take 2x2 Robert Operator and place it on the Original Image.

Step 7: Find the Gradient Magnitude and replace the origin pixel by the

magnitude.

Step 8: Convolve the mask on the entire image.

Step 9: Display the Convolved image.

Limitations:

1. It is less vulnerability to noise.

2. It produces very weak responses to genuine edges unless they are very sharp.

2.2.1.2 Prewitt edge detection operator:

The Prewitt edge detector is a much better operator than the Roberts

operator. This operator having a 3x3 masks deals better with the effect of noise.

The Prewitt edge detection masks are one of the oldest and best

understood methods of detecting edges in images. Basically, there are two

masks, one for detecting image derivatives in X and one for detecting image

derivatives in Y. This Prewitt operator is obtained by setting c = 1.

An approach using the masks of size 3x3 is given by considering the below

arrangement of pixels;

23


Like wise the Prewitt edge detector is used as the masks have longer support.

Algorithm :

Step 1: Read image.






Step 6: Take 3x3 Prewitt Operator and place it on the Original Image.

Step 7: Find the Gradient Magnitude and replace the center pixel by the

magnitude.



Limitation:

These differentiate in one direction and average in the other direction,

so the edge detector is less vulnerable to noise.

2.2.1.3. Sobel edge detection operator:

The Sobel edge detector is very much similar to the Prewitt edge

detector. The difference between the both is that the weight of the center

coefficient is 2 in the Sobel operator.

24

The Sobel operator performs a 2-D spatial gradient measurement on an

image. Then, the approximate absolute gradient magnitude (edge strength) at

each point can be found. The Sobel operator is more sensitive to diagonal edges

than vertical and horizontal edges. The Sobel operator uses a pair of 3x3

convolution masks, one estimating the gradient in the x-direction (columns) and

the other estimating the gradient in the y-direction (rows). A convolution mask is

usually much smaller than the actual image. The actual Sobel masks are shown

below:

This operator consists of a pair of 3*3 convolution masks, one mask is

simply the other rotated by 90Â°. The mask is slid over an area of the input

image, changes that pixel's value and then shifts one pixel to the right and

continues to the right until it reaches the end of a row. It then starts at the

beginning of the next row.

The center of the mask is placed over the pixel you are manipulating in the

image. It is important to notice that pixels in the first and last rows, as well as

the first and last columns cannot be manipulated by a 3x3 mask. This is because

when placing the center of the mask over a pixel in the first row (for example),

the mask will be outside the image boundaries.

An approach using the masks of size 3x3 is given by considering the below

arrangement of pixels;

25


Like wise the Sobel edge detector is used.

ALGORITHM FOR SOBEL OPERATOR:

Step 1: Read image.






Step 6: Take 3x3 Sobel Operator and place it on the Original Image.


magnitude.



Advantages:

26

1. An advantage of using a larger mask size is that errors due to the effects of

noise are reduced by local averaging within the neighborhood of the mask.

2. An advantage of using a masks of odd size is that the operators are centered

and can therefore provide an estimate that is biased towards a centre pixel (i , j).

Limitations:

Although the Prewitt masks are easier to implement than the Sobel

masks, the later has better noise suppression characteristics.

2.2.1.4. Robinson Edge Detector operator:

The Robinson edge detector is similar to the Sobel edge detector. The

Robinsons row, column and corner masks are shown below:

It is symmetrical by the axis. The maximum value is found in the edge

magnitude and the edge direction is defined by the maximum value found by the

edge magnitude.

ALGORITHM FOR ROBINSON OPERATOR:

Step 1: Read image.





27


Step 6: Take 3x3 Robinson Operator and place it on the Original Image.


magnitude.



2.2.1.5. Krisch Edge Detector operator:

The Krisch edge detector is similar to the above edge detector. The

Krisch row, column and corner masks are shown below:

ALGORITHM FOR KRISCH OPERATOR:

Step 1: Read image.






Step 6: Take 3x3 Krisch Operator and place it on the Original Image.

28


magnitude.



2.2.2. Second Order edge detection technique:

Second Order derivative edge detection techniques employ

some form of spatial second order differentiation to accentuate edges. an edge is

marked if a significant spatial change occurs in the second derivative.

2.2.2.1 Laplacian and Gaussian Edge detection operator:

The Laplacian operator is a very popular operator approximating the

second derivative which gives the gradient magnitude only. The principle used in

the Laplacian of Gaussian method is, the second derivative of a signal is zero

when the magnitude of the derivative is maximum. The Laplacian is

approximated in digital images by a convolution sum. The Laplacian of a 2-D

function f(x, y) is defined as;

A 3x3 masks for 4-neighborhoods and 8-neighborhood for this operator

are as follows;

29

The two partial derivative approximations for the Laplacian for a 3x3 region are

given as,

The 5x5 Laplacian is a convoluted mask to approximate the second

derivative. Laplacian uses a 5x5 mask for the 2nd derivative in both the x and y

directions.

ALGORITHM FOR LAPLACIAN AND GAUSSIAN OPERATOR:

Step 1: Read image.


30




Step 5: Display Threshold image.

Step 6: Perform Gaussian Operation on Original image to reduce the noise.

Step 7: Take 3x3 Laplacian Operator and place it on the Original Image which is

masked by the Gaussian operator.


magnitude.



Limitations:

1. It responds doubly to some edges in the image.

Limitations for Edge Based Segmentation:

Edge-based methods center around contour detection: their weakness

in connecting together broken contour lines make them, too, prone to failure in

the presence of blurring

The main disadvantage of these edge detectors is their dependence on

the size of objects and sensitivity to noise

Further, since conventional boundary finding relies on changes in the grey

level, rather than their actual values, it is less sensitive to changes in the grey

scale distributions over images as against region based segmentation.

2.3 Algorithm for Adjacent Difference Method:

Step 1: Read image.

Step 2: Convert image into two dimensional matrixes (eg: 64X64).

31




Step 5: Display Threshold image.

Step 6: Calculate the Row Wise differences by subtracting A[i] [j+1] from A[i] [j],

where i,j are row and column indices respectively.

Step 7: Calculate the Column Wise differences by subtracting A [i+1] [j] from

A[i] [j], where i,j are row and column indices respectively.

Step 8: Calculate the Right Diagonal Wise differences by subtracting

A [i+1] [j+1] from A[i] [j], where i,j are row and column indices

respectively.

Step 9: Calculate the Left Diagonal Wise differences by subtracting A [i+1] [j-1]

from A[i] [j], where i,j are row and column indices respectively.

Step 10: If Row Wise difference value, Column Wise difference value, Right

Diagonal Wise difference value, Left Diagonal Wise difference value is

greater than or equal to the 6 then make it bright, else make it dark.

Step 11: Display Completely Segmented image.

32

Chapter -3

DESCRIPTION OF THE PROJECT

3.1 Introduction

EDGE detection is a vital step in image processing and is one

of the most crucial steps towards classification and Recognition of

objects. Color plays a crucial role in image analysis and recognition. A

color image will have a vector of three values for every pixel unlike in

gray images where a single value representing the intensity of a

pixel. Human vision system chooses color rather than shapes and

texture as the major discriminate attribute . Many algorithms have

been proposed for edge detection of color images. Of all the edge

detectors, Sobel is the standard detector and Canny is the modern

standard and is used by researchers to compare their results with the

results of Canny detector. Novak and Shafer found that 90% of the

edges are about the same in gray level and in color images. It implies

that 10% of the edges are left over in gray level images. Since

color images give more information than gray-level images, this 10%

left over edges may be extracted from color images. In general, to

extract edges from the images either gradient based methods or

33

vector based methods are used.

The real complement approach is taking as a pre-processing

step for real edge detection . This complement acts like a low pass

filter and highlights the weak intensity pixels. The nature of the

complement process is that it reduces the gray intensity distribution

to 50% i.e., a gray level image having the intensities in the range 0-

255 will be reduced to the intensities of the range 0-128 thereby the

weak intensity pixels will be highlighted. In a new method real

complement approach has been introduced. This real edge gives

efficient edge information with out discontinuities.

To achieve this, we are using different shift operations .

In the proposed method, we have used the complement approach as

a pre-processing step. Initially, the algorithm is tested on RGB

images, where in each channel is separately processed by the

complement operation followed by matrix shifting operations. The

proposed algorithm is further tested on HSV and YUV color spaces also.

In addition, the algorithm is tested on both standard and real images

and the results are compared with the results of the other methods.

All the procedures used in the algorithm runs in polynomial time.

3.2. Color Spaces

A color space relates to number of actual colors, and is a three

dimensional object which contains all realizable color combinations. Color

spaces can be either dependent to or independent of a given device.

Device-dependent spaces express color relative to some other color space

whereas independent color spaces express color in absolute terms. Each

dimension in color space represents some aspect of color, such as lightness,

saturation or hue, depending on the type of space.

A. RGB Model

34

An RGB color space is any additive color space based on RGB

color model. A particular RGB color space is defined by the three

chromaticity’s of red, green and blue additive primaries. RGB is a

convenient color model for computer graphics because the human visual

system works similar to an RGB color space.

B.HSL and HSV Model

HSL and HSV are two related representations of points in RGB

color space which attempt to describe the perceptual color

relationships more accurately than RGB. HSL stands for hue,

saturation and lightness.HSV stands for hue, saturation and value.

C. YUV Model

YUV model defines a color space in one luminance (Y) and

two chrominance (UV) components. YUV models human perception of

color in a different way from the standard RGB model. Y stands for

luminance (brightness) component and UV stands for the

chrominance (color) components. The transformation from RGB into

YUV is given by

Since the luminance and chrominance components are separate,

the YUV space is vigorously used to broadcast video systems and

hence it is also used in image and video processing.

3.3. Real Complement Approach

The Real complement approach works like a low-pass filter

strongly attenuating the lower image frequencies. This method is based

on image complement and hence the name, Real complement. The

35

method is inexpensive due to its simple arithmetic operations. The

steps involved are as follows.

Step-1: Read the given image (i).

Step-2: Obtain the complement of the image. (ic).

Step-3: Do image differencing. (i1=i-ic).

Step-4: Perform global threshold ding on i1 to get binary image (i2).

Step-5: Obtain the real complement of the image by image differencing

between i and i2.(i3=i-i2).

The third step in real Complement approach is the subtracting the

complement of the input image from the given input image. Let i be the

given image and ic be its complement.

This image subtraction makes the background uniform

depending on the intensity of the background pixels of the image. For

illustration, they categorized the

images into three cases based on the background information. The

categorization is (i) background - very bright, (ii) background

slightly dark and (iii) background intermediate range. In the first

case, the background of i1 remains almost same since the

complement of the bright intensity values when subtracted from its

original intensity leads to its nearest (bright) intensity range. In the

second case i.e., when the background is slightly dark, the same

analogy works as in the case of first case, but nearer to its original

range. In the third case, if the difference is a positive

value, a similar gray of value is retain otherwise it becomes black.

The fourth step in Real-Complement approach is the binary map

formation. This is achieved by global thresholding . The binary map of

i2 obtained in the previous step is calculated by using following

equation.

36

3.4. Circular Shift Operations

Literature reveals many edge detectors based on

gradients, filters, derivatives etc. A simple method to detect real

edges has been introduced based on shift operations. The present

approach to edge detection incorporates the simple shift operations.

Here we are shifting entire image data i.e. left, right, up and

bottom. The following are terms used for representing

shift operations in proposed algorithm.

i. RSHT: Row Shift

ii.CLSHT: Circular Left Shift

iii.CRSHT: Circular Right Shift

iv. CSHFT: Column Shift

v. CTSHT: Circular Top shift

vi.CBSHT: Circular Bottom shift

vii.DSHT: Diagonal shift.

Procedure for finding different shifting images.

A. Finding RSHT image.

i.Read given image (i1)

37

ii. Apply CLSHT (Circular Left Shift) operation on entire original image(i2).

iii. Image differencing to obtain left slope edge pixels.(i3=(i1-i2).

iv.Apply CRSHT(Circular Right Shift)Operation on entire original

image(i4)

v.Image differencing to obtain right slope edge pixels.(i5=i1-i4).

vi.To obtain RSHT image add (i3) and (i5).

B. Finding CSHT image

i. Read given image (i1)

ii .Apply CTSHT( Circular Top Shift) operation on entire original

image(i2)

iii. Image differencing to obtain top slope edge pixels(i3=i1-i2)

iv. Apply CBSHT (Circular Bottom Shift) operation on entire original

image (i4).

v.Image differencing to obtain bottom slope edge pixels.(i5=i1-i4).

vi.To obtain CSHT image add (i3) and (i5).

C. Finding DSHT Image

i.Read given image.(i1).

ii.Apply Circular shift on diagonal image data.(i2).

iii.Image differencing to obtain diagonal edge pixels.(i3=i1-i2).

3.4. Proposed Methodology

The method proposed by us for extracting edges from color

images. The block diagram of the proposed method is given in Fig. 1.

The steps of the proposed method are as follows.

Step-1: Read the given color image.

38

Step-2: Obtain the Real complement for all three channels separately.

Step-3: Find RSHT image for individual real complement.

Step-4: Find CSHT image for individual real complement.

Step-5: Find DSHT image for individual real complement.

Step-6: For Strong Real edges add result images obtained by step3, 4and

5.

Step-7: Post-processing of the resultant edge image to get the final output.

39

Chapter-4

40

EXPERIMENTAL RESULTS ON IMAGES

In this section, the results of the proposed method are

presented. Both standard images and real world images are used to

show the efficiency of the proposed method. By the results we can see

that the proposed algorithm detecting exact real edges.

41

Standard color image of a women is supplied as input to the proposed method.

The edge map for the given input image using Right shift operations.

The same algorithm applied for real world image, a tree.

42

The edge map for the given input image using column shift operations.

The edge map for the given input image using real complement approach.

The real world image of a tree is supplied as input to the proposed method.

43

The edge map for the given input image using Right shift operations

The edge map for the given input image using column shift operations.

The edge map for the given input image using real complement approach.

Chapter- 5

CONCLUSION & FUTURE WORK

5.1 Conclusion

A new approach that is proposed in this paper is used to

extract real edges from color images. The detected edges are more

accurate compare with some existed color edge detectors. This algorithm

incorporates simple shift and arithmetic operations rather than using

gradient and other operations which are computationally expensive.

And also it manipulates entire image at a time which is never

observed in normal edge detection process. Experimental results

indicate that the performance of the proposed method is satisfactory

in almost all cases and runs in polynomial time.

5.2 Future Work

Our future work entails the reconstruction of images from the

edge map.

44

REFERENCES

[1]. T. N. Janakiraman, and P. V. S. S. R.ChandraMouli, “Color Image

Edge Detection using Pseudo-Complement and Matrix Operations”,

Proceedings of World Academy of sciences, Engineering and

Techonology. ,Volume 32,August 2008.ISSN 2070-3740.

[2].J.T.Allen and T.Huntsburger,”Comparing color edge detection and

segmentation methods.”IEEE South East Con’89.

[3].T. Carron and P. Lambert, “Fuzzy color extraction by inference rules

quantitative study and evaluation performances ”, proceedings of IEEE

International Conference on Image Processing, pp.181-184, 1995.

[4]. R. Nevatia , “A color edge detector andits use in scene

segmentation”, IEEE Transactions on Systems, ManCybernetics, Vol.

7, pp.820-826, 1977.

[5]. M.A. Abidi, R.A. Salinas, C. Richardson and R.C. Gonzalez,

“Data fusion color edge detection and surface reconstruction through

regularization”, IEEE Transactions on Ind Electron.,43(3), pp.355-

363,1996.

[6]. J. Canny, “A computational approach to edge detection”, IEEE

Transactions on Pattern Analysis and Machine Intelligence, 8(6),

pp.679-698, 1986.

[7]. C.L. Novak and S.A. Shafer, Color edge detection”, Proceedings

of DARPA Image Understanding workshop, Los Angeles, CA, USA, vol.1,

pp.35-37, 1987.

[8]. R.C.Gonalez and R.E.Woods: “Digital Image Processing (3rd

Editin)”, Prentice Hall, 2007.

[9]M.Morris Mano: “Computer System Architecture” (3rd edition)

45

Prentice Hall,

2006.

46

imgprocessingproject

Documents