International Journal of Computer Science Trends and Technology (IJCST) – Volume 3 Issue 3, May-June 2015

RESEARCH ARTICLE OPEN ACCESS

Role of Image Segmentation in Digital Image Processing

For Information Processing
[Link]
Research Scholar
Research and Development Centre
Bharathiar University
Tamil Nadu - India

ABSTRACT
Digital Image processing is a technique using computer algorithms to perform specific operations on an image,
in order to get an enhanced image or to extract some useful information from it. Image segmentation, an
important phase in image processing, is the division of an image into regions or categories, which correspond to
different objects or parts of objects. This step is typically used to identify objects or other relevant information
in digital images. There are generic methods available for image segmentation, but each method has to be
applied in a particular context combining the domain information of that particular domain.
Keywords :- Digital image processing, image segmentation, edge detection.
II. METHODS OF IMAGE
I. INTRODUCTION PROCESSING
In imaging science, image processing refers to
processing of images using mathematical operations There are two types of methods used in Image
in order to get an enhanced image or to extract some Processing and they are Analog image processing
useful information from it. It is a type of signal and digital image processing
dispensation in which input may be image like
video frame or photograph and output may be image 1)Analog Image Processing
or characteristics associated with that image. Image Analog Image Processing refers to the alteration
processing is a rapidly growing technology today, of image through electrical means. The most
with its applications in various aspects of science, common example is the television image. The
engineering, management, business and day to day television signal is a voltage level which varies
life activities and it is a core research area too. in amplitude to represent brightness through the
Image Processing is a technique useful in image. By electrically varying the signal, the
enhancing raw images received from displayed image appearance is altered. The
cameras/sensors placed for various applications brightness and contrast controls on a TV set serve
including enhancing images obtained from to adjust the amplitude and reference of the video
unmanned spacecrafts, space probes and military signal, resulting in the brightening, darkening and
reconnaissance flights[1]. Image Processing finds alteration of the brightness range of the displayed
applications in areas such as Remote Sensing, image.
Medical Imaging, Non-destructive Evaluation,
Forensic Studies, Textiles, Material Science, 2)Digital Image Processing
Military, Film industry, Document processing, In this case, digital computers are used to process
Graphic arts and Printing Industry. the image. The image will be converted i n to
digital form using a scanner or digital camera and
then given as input for processing operations by
digital computer. Digital image processing may
involve numerous procedures including formatting
and correcting of the data, digital enhancement to
facilitate better visual interpretation, or even

ISSN: 2347-8578 [Link] Page 312

International Journal of Computer Science Trends and Technology (IJCST) – Volume 3 Issue 3, May-June 2015
automated classification of targets and features noise (insignificant lines and contours) and the
entirely by computer. The requirements for digital presence of featureless space and the important
image processing is a computer system, sometimes features include surface details and boundaries
referred to as an image analysis system, with the such as lines, edges, and vertices. Image
appropriate hardware and software to process the preprocessing typically includes a processing step
data. There are efficient software systems, both open of transforming a source image into a new image
source and proprietory, have been developed for which is fundamentally similar to the source
image processing. The term digital image image, but differs in certain aspects, e.g. improved
processing generally refers to processing of a contrast. In this step, computer suppresses noise
two-dimensional picture by a digital computer and and sometimes enhances some object features
a digital image is an array of real numbers which are essential in understanding the image.
represented by a finite number of bits[2]. [Link] segmentation: The main aim of
The main advantage of Digital Image segmentation is to reduce the information to enable
Processing methods is its versatility, repeatability easy analysis. Segmentation is also useful in Image
and the preservation of original data precision. Analysis and Image Compression[3]. In this
The following two principal application areas are process, an image is divided into multiple parts
majorily the reasons for the growing interest and and this is typically used to identify objects or
demand in digital image processing : other relevant information in digital images. The
(i) improvement of pictorial information segmentation has two objectives - i. to decompose
for human interpretation, and the image into parts for further analysis. ii. to
(ii) processing of scene data for autonomous perform a change of representation. ie., The pixels
machine perception. of the image must be organized into higher-level
Example for the first application area are the units that are either more meaningful or more
enhancement of images in medical field such as efficient for further analysis.
MRI image, X-ray image etc to improve the In short, segmentation separates an image in to its
pictorial content of it , thus making it easier for component regions or objects. Image
human interpretation. The second application area segmentation needs to segment the object from the
focuses on procedures for extracting image background to read the image properly and to
information in a form suitable for computer identify the content of the image carefully.
processing. Examples include automatic character Because of the same reason, edge detection is a
recognition, industrial machine vision for product fundamental tool for image segmentation.
assembly and inspection, military [Link] and description: Representation
recognizance, automatic processing of is transforming raw data into a form suitable for
fingerprints etc. computer processing. Description, which is also
called feature extraction, deals with extracting
III. STEPS IN IMAGE PROCESSING features that result in some quantitative
information of interest or features which are basic
for differentiating one class of objects from
There are some common steps in image
another.
processing as listed below, though all image
[Link] & Interpretation : Recognition is the
processing systems would not require all steps
process which assigns a label to an object based on
together.
the information provided by its descriptors.
[Link] acquisition : An image is captured by a
Interpretation is the process of assigning meaning
sensor such as a TV camera and it is digitized.
to an ensemble of recognized objects.
[Link] Preprocessing : The first step in preparing
Of the different steps mentioned above, Image
the picture for higher-level processing is called
segmentation, which has emerged as an important
pre-processing and the purpose of pre-processing is
phase in image based applications, is going to be
two-fold: to eliminate undesirable features that will
dealt in detail in the following section.
hinder further processing and to extract the
desirable features that represent useful information IV. IMAGE SEGMENTATION
in the image. Unwanted image attributes include

ISSN: 2347-8578 [Link] Page 313

International Journal of Computer Science Trends and Technology (IJCST) – Volume 3 Issue 3, May-June 2015
Image segmentation, one of the significant aspects the global threshold, a single threshold value is
of image processing, is a long standing problem in used in the whole image. When T depends only
the research area of computer vision. The main aim on f(x,y) (in other words, only on gray-level values)
of segmentation is to extract the ROI(Region of and the value of T solely relates to the character of
Interest)for image analysis. The division of an image pixels and this thresholding technique is called
into meaningful structures, ie., image segmentation, global
is often an essential step in image analysis, object thresholding technique .When the pixel values of
representation, visualization, and many other image the components and that of background are fairly
processing tasks. Segmentation plays an important consistent in their respective values over the entire
role in image processing since separation of a large image, global thresholding could be used[5]. There
image into several parts makes further processing are a number of global thresholding techniques
simpler. These several parts that are rejoined will such as: Otsu, optimal thresholding, histogram
cover the entire image. Segmentation may also analysis, iterative thresholding, maximum
depend on various features like colour or texture correlation thresholding, clustering, Multispectral
that are contained in the image. Before denoising and Multithresholding.
an image, it is segmented to recover the
original image. The main aim of segmentation is ii)Local thresholding: A single threshold will not
to reduce the information and hence making easy work well when we have uneven illumination due to
analysis possible. shadows or due to the direction of illumination.
There exists several image segmentation techniques, Local thresholding is applicable in such situations.
which partition the image into several parts based on The idea is to partition the image into m x m sub
certain image features like pixel intensity value, images and then choose a threshold. If threshold T
color, texture, etc and these techniques are depends on both f(x,y) and p(x,y) , this
categorized based on the segmentation method used. thresholding is called local thresholding . This
A great variety of segmentation methods has been method divides an original image into several sub
proposed till now and they fall in to the following regions, and chooses various thresholds Ts for each
major categories. sub region reasonably. After thresholding,
discontinuous gray levels among sub images
a)Threshold based segmentation must be eliminated by gray level filtering
technique. If threshold T depends on both f(x,y)
Thresholding is the simplest method of image and p(x,y), this thresholding is called local
segmentation. Thresholding techniques may be thresholding[6] . This method divides an original
applied directly to an image, but can also be image into several sub regions, and chooses various
combined with pre and post-processing techniques. thresholds Ts for each sub region reasonably.
In threshold based segmentation, pixels are allocated After thresholding, discontinuous gray levels
to categories according to the range of values in among sub images must be eliminated by gray
which a pixel lies[4]. For example, if an image is to level filtering technique. Some of the local
be segmented, pixels with values less than 128 can thresholding techniques are simple statistical
be placed in one category, and the rest can be placed thresholding, 2-D entropy-based thresholding and
in the other category. The boundaries between histogram-transformation thresholding.
adjacent pixels in different categories of this image
can be superimposed in white on the original image. iii)Adaptive thresholding: Adaptive thresholding
The new image is said to be threshold segmented the typically takes a grayscale or color image as input
image into the two predominant types. and, in its simplest implementation, outputs a binary
There are three types of thresholding algorithms - image representing the segmentation. For each pixel
Global thresholding, Local thresholding and in the image, a threshold has to be calculated. If the
Adaptive thresholding. pixel value is below the threshold it is set to the
i)Global thresholding : The global threshold is background value, otherwise it assumes the
applicable when the intensity distribution of objects foreground value. In adaptive thresholding, different
and background pixels are sufficiently distinct. In threshold values for different local areas are used.

ISSN: 2347-8578 [Link] Page 314

International Journal of Computer Science Trends and Technology (IJCST) – Volume 3 Issue 3, May-June 2015
Adaptive thresholding is used to separate desirable input image with is neighbors. This operation
foreground image objects from the background produces a new binary image in which if test is
based on the difference in pixel intensities of each successful it will have non-zero pixel value at that
region[7]. The drawback of this method is that it is location in the input image. There are various
computationally expensive and, therefore, is not structuring elements such as diamond shaped,
appropriate for real-time applications. square shaped and cross shaped. The base of the
morphological operation is dilation, erosion,
b)Edge based segmentation opening, closing expressed in logical AND, OR
With this technique, detected edges in an image are notation and described by set analysis. The
assumed to represent object boundaries, and are morphological filter which can be constructed on
used to identify these objects. Edges characterize the basis of the underlying morphological
boundaries and are therefore a problem of operations, are more suitable for shape analysis than
fundamental importance in image processing[8] . the standard linear filters since the latter sometimes
Image Edge detection significantly reduces the distort the underlying geometric form of the image.
amount of data and filters out useless information, Morphological watersheds is also a morphology
while preserving the important structural properties based technique, but unlike the typical
in an image. morphological filters, the watersheds transformation
is not composed of the primitive morphological
c)Region based segmentation. operations. The initial concept of the watersheds
Where an edge based technique may attempt to find transformation as a morphological tool was
the object boundaries and then locate the object introduced by H. Digabel and C. Lantuéjoul[9].
itself by filling them in, a region based technique Later, a joint work of C. Lantuéjoul and [Link]
takes the opposite approach, by (e.g.) starting in the led to the development of extension of this
middle of an object and then “growing” outward algorithm to the more general framework of
until it meets the object boundaries. In this grayscale images. Watershed represents the
technique, pixels that are related to an object boundaries between adjacent catchments. If one
are grouped for [Link] thresholding combines the grey level of each point at an altitude,
technique is bound with region based segmentation. it is possible to define the watershed transform as
The area that is detected for segmentation should be the ridge forming the boundary between two
closed. Region based segmentation is also known as watersheds. This is to compute the watershed of the
Similarity Based Segmentation. There would not be said relief and watersheds thus obtained correspond
any gap due to missing edge pixels in this region to regions of the image.
based segmentation. The boundaries are identified
for segmentation. In each and every step at least one d)Model based segmentation
pixel is related to the region and is taken into Markov Random Field (MRF) based segmentation is
consideration. After identifying the change in the known as Model based segmentation. An inbuilt
color and texture, the edge flow is converted region smoothness constraint is presented in MRF
into a vector[3].From this, the edges are detected which is used for color segmentation. Components
for further segmentation. of the color pixel tuples are considered as
independent random variables for further
d)Morphological methods processing. MRF is combined with edge detection
In image processing, morphology is the name of a for identifying the edges accurately[10].
specific methodology for analyzing the geometric
structure inherent within an image. Morphological e)Clustering techniques.
techniques verify the image with a small Segmentation is also done through Clustering. A
template called structuring element. This basic clustering algorithm such as K-means is
structuring element is applied to all possible used for segmentation in textured images. It
locations of the input image and generates the clusters the related pixels to segment the image.
same size output. In this technique the output Segmentation is done through feature clustering
image pixel values are based on similar pixels of and there it will be changed according to the

ISSN: 2347-8578 [Link] Page 315

International Journal of Computer Science Trends and Technology (IJCST) – Volume 3 Issue 3, May-June 2015
color components. Segmentation is also purely tool for image segmentation. Edge detection
depending on the characteristics of the image. methods transform original images into edge
Features are taken into account for segmentation. images benefits from the changes of grey tones
Difference in the intensity and color values are in the image. In image processing especially in
used for segmentation. For segmentation of color computer vision, the edge detection treats the
image, Fuzzy Clustering technique can be used, localization of important variations of a gray level
which iteratively generates color clusters using image and the detection of the physical and
Fuzzy membership function in color space geometrical properties of objects of the scene. It is a
regarding to image space. The technique is fundamental process detects and outlines of an
successful in identifying the color region. In real object and boundaries among objects and the
time clustering based segmentation, A Virtual background in the image. Edge detection is the most
attention region is captured accurately for familiar approach for detecting significant
segmentation[3]. Image is segmented coarsely by discontinuities in intensity values.
multithresholding. It is then refined by Fuzzy C- Edges are local changes in the image intensity.
Means Clustering. The advantage is applied to any Edges typically occur on the boundary between
multispectral images. Though clustering is two regions. The main features can be extracted
sometimes used as a synonym for agglomerative from the edges of an image and this provides major
segmentation techniques, this term is used here to features for image analysis. These features are
denote techniques that are primarily used in used by advanced computer vision
exploratory data analysis of high-dimensional algorithms. Edge detection is used for object
measurement patterns. In this context, clustering detection which serves various applications such
methods attempt to group together patterns that are as medical image processing, biometrics etc.
similar in some sense. Clustering is a powerful Edge detection is an active area of research as
technique that has been used in image it facilitates higher level image analysis. There
segmentation. The cluster analysis is to partition are three different types of discontinuities in the
an image data set into a number of disjoint groups or grey level like point, line and edges[13]. Spatial
clusters. K means clustering is one of the popular masks can be used to detect all the three types of
method because of its simplicity and discontinuities in an image.
computational efficiency[11]. There are many edge detection techniques in the
literature for image segmentation. The most
[Link] BASED DETECTION commonly used discontinuity based edge
TECHNIQUES detection operators are Roberts edge detection,
Sobel Edge Detection, Prewitt edge detection,
Since edge detection is in the forefront of image
Kirsh edge detection, Robinson edge detection,
processing for object detection, it is crucial to
Marr-Hildreth edge detection, LoG edge detection
have a good understanding of edge detection
and Canny Edge Detection. The performance of
algorithms. The edge representation of an image
edge detectors are to be evaluated as there is no
significantly reduces the quantity of data to be
guarantee that the edge detectors can detect all
processed, yet it retains essential information
edges[14]. Some of the problems of the edge
regarding the shapes of objects in the scene. This
detectors are – missing valid edge points, classifying
explanation of an image is easy to incorporate into a
the noise points as valid edge points and smearing
large amount of object recognition algorithms used
edges.
in computer vision along with other image
Edge detectors often do not produce continuous
processing applications. The major property of
edges. Often the detected edges are not sharp and
the edge detection technique is its ability to
continuous due to the presence of noise and intensity
extract the exact edge line with good orientation.
variations. Therefore edge linking is used to detect
It is generally difficult to judge the performance of
the presence of edges and to connect them to their
the edge detection techniques. The performance of
neighbours to avoid breaks. Continuity is ensured by
edge detection techniques are always judged
techniques such as hysteresis thresholding and edge
personally and separately, dependent to its
relaxation[15]. Thresholding produces uniform
application[12]. Edge detection is a fundamental

ISSN: 2347-8578 [Link] Page 316

International Journal of Computer Science Trends and Technology (IJCST) – Volume 3 Issue 3, May-June 2015
regions based on the threshold criterion. Edge techniques, Computer vision, graphics, and
relaxation is the process of re-evalutation of pixel image processing, 1998, 41(2), 233-260.
classification using its context. [7] E.R. Davies , Machine Vision: theory,
algorithms, practicalities, 1997.
[Link] [8] R. Gurcan, I. Erer and S. Kent, An edge
detection method using 2-D autoregressive
lattice prediction filters for remotely sensed
Segmentation of region of interest in real world
images, In Geoscience and Remote Sensing
images is the one major hurdle for effective
Symposium, 2004, IGARSS'04,
implementation of image processing applications as
Proceedings of 2004 IEEE International ,
segmentation process is often difficult. The success
IEEE, September, 2004, Vol. 6, pp. 4219-
or failure of image segmentation determines the
4222.
success of image processing applications. Therefore
[9] J. Serra and P. Salembier, “Connected
the user has to try many segmentation algorithms
operators and pyramids,” in Image Algebra
and pick an algorithm that performs the best for the
and Morphological Image Processing IV,
given requirement[16]. Validation of a segmentation
vol. 2030 of Proceedings of SPIE, San
algorithm is carried out to evaluate the accuracy of
Diego, Calif, USA, July 1993, pp. 65–76.
the segmentation method based on a truth
[10] [Link] and [Link], Edge detection
model[17]. Several works are in progress
techniques-an overview, Pattern
researching in improving the efficiency of
Recognition and Image Analysis C/C of
segmentation algorithms.
Raspoznavaniye Obrazov I Analiz
Izobrazhenii, 8, 1998, 537-559.
REFERENCES [11] T. Kanungo , [Link], N.S. Netanyahu,
[1] [Link] and M.K. Arora, Overview of
C.D. Piatko, R. Silverman, and A.Y. Wu,
image processing, In Advanced image
An efficient k-means clustering algorithm:
processing techniques for remotely sensed
Analysis and implementation. IEEE
hyperspectral data, Springer Berlin
transactions on pattern analysis and
Heidelberg, 2004, pp. 51-85.
machine intelligence, 2002, 24(7), 881-892.
[2] K. Ramesh and [Link], Digital
[12] R. Muthukrishnan and [Link], Edge
Photoelasticity: Advanced Techniques and
detection techniques for image
Applications, Applied Mechanics Reviews,
segmentation, International Journal of
55, B69, 2002.
Computer Science & Information
[3] R. Yogamangalam and B. Karthikeyan,
Technology, 2011, 3(6), 259.
Segmentation techniques comparison in
[13] A. Chaudhary and T. Gulati, Segmenting
image processing, International Journal of
digital images using edge detection,
Engineering and Technology (IJET), 2013,
International Journal of Application or
5(1), 307-313.
Innovation in Engineering & Management
[4] P. Singh, P and R.S. Chadha, A Novel
(IJAIEM), 2013, 2(5).
Approach to Image Segmentation,
[14] [Link] and F. Ge and T. Liu, Evaluating
International
edge detection through boundary detection,
Journal of Advanced Research in Computer
EURASIP Journal on Applied Signal
Science and Software Engineering
Processing, 2006.
Research Paper,2013, 3(4).
[15] P. Dhankhar and N. Sahu, A review and
[5] H. Devi, Thresholding: A pixel-level
research of edge detection techniques for
image processing methodology
image segmentation, International Journal
preprocessing technique for an OCR
of Computer Science and Mobile
System for the Brahmi Script, Ancient
Computing (IJCSMC), 2013, 2(7), 86-92.
Asia, 1, 2006.
[16] C.A. Boiangiu and R. Ioanitescu, Voting-
[6] P.K. Sahoo, Soltani, S. A. K. C, and A.K.
Based Image Segmentation, Journal of
Wong, A survey of thresholding

ISSN: 2347-8578 [Link] Page 317

International Journal of Computer Science Trends and Technology (IJCST) – Volume 3 Issue 3, May-June 2015
Information Systems & Operations
Management, 1, 2013.
[17] S.K. Warfield, K.H. Zou and W.M. Wells,
Simultaneous truth and performance level
estimation (STAPLE): an algorithm for the
validation of image segmentation, IEEE
transactions on medical imaging, 2004,
23(7), 903-921.

ISSN: 2347-8578 [Link] Page 318