The Color Dissimilarity based Method among Other Segmentation

Methods: A Comparison

I Gede Made Karma

, I Ketut Gede Darma Putra

, Made Sudarma

and Linawati

Doctoral Program of Engineering Science, Faculty of Engineering, Udayana University, Badung, Bali, Indonesia

Information Technology, Faculty of Engineering Udayana University, Badung, Bali, Indonesia

Electrical Engineering, Faculty of Engineering Udayana University, Badung, Bali, Indonesia

Keywords: Segmentation, Color Dissimilarity, Ground-truths, Object Detection.

Abstract: The segmentation process plays a very important role in the process of detecting and recognizing objects in

an image. Although many segmentation methods have been developed, there is no method that can give good

results and is generally accepted. This study aims to find a better segmentation method in finding patterns or

ground-truths of objects in an image, so that these objects can be easily recognized. Based on the results

shown by all the methods being compared, several methods were not successful in showing the pattern of

objects contained in the sample image. The segmentation method based on color dissimilarity, which is a

method that emulates the way humans recognize an object based on visible color differences, shows the best

results compared to other methods. This method is a very suitable method to be used in the process of detecting

and recognizing objects in an image.

1 INTRODUCTION

Our interest in an image actually lies in certain parts.

This part usually has unique and special properties. In

image processing, this part is commonly referred to

as the target or foreground. The other part is called

the background. It takes an extraction process to

separate these parts so that they can be analyzed and

identified. This is where image segmentation comes

into play. Based on the features it has, whether in the

form of gray pixels, colors or other textures, the

segmentation process is carried out. The goal is to

find these targets (Jun, 2010).

In image processing, object identification can be

done after going through a number of process stages.

This series of processes consists of noise removal,

segmentation, feature extraction and classification.

Of all the stages of this process, the segmentation

process is the most important and most determining

the success of the object identification process (P.

Agrawal et al., 2010). The process of segmenting an

image is a process of partitioning an image into

https://orcid.org/0000-0003-3837-5345

https://orcid.org/0000-0002-1960-6462

https://orcid.org/0000-0002-8331-0519

https://orcid.org/0000-0001-8303-2822

several parts. The process of image partitioning is

carried out based on the similarity of existing

characteristics in the image, such as similarity in

color, intensity, texture, shape and others, so that it

can be used to find and identify objects and

boundaries in an image (Kumar et al., 2014). This

segmentation process will change the image

representation in such a way that it becomes easy to

analyze (Dass et al., 2012).

Image segmentation is the first step in image

analysis. The image segmentation process divides the

image into a collection of pixels which are connected

into a single unit. The resulting set of pixels is a

region, which is determined on the basis of visual

properties extracted from the local features of the

image (Ivanovici et al., 2013). The segmentation

process will divide the image into several segments in

such a way that it can find objects and image

boundaries. Image segmentation is used to identify

objects and backgrounds in the image (Saini & Arora,

2014). Image segmentation is the process of assigning

a label to each pixel in an image such that pixels with

Made Karma, I., Darma Putra, I., Sudarma, M. and Linawati, .

The Color Dissimilarity based Method among Other Segmentation Methods: A Comparison.

DOI: 10.5220/0010746800003113

In Proceedings of the 1st International Conference on Emerging Issues in Technology, Engineering and Science (ICE-TES 2021), pages 131-141

ISBN: 978-989-758-601-9

131

the same label share certain visual properties (Jayagar

& Jeyakumari, 2015). The segmentation technique

mainly converts complex images into simple images.

There are many factors that influence the results

of the image segmentation process. Apart from the

problem domain, the pixel color factor, color

intensity, texture, similarity and image content will

determine the outcome of this segmentation process.

The number of these factors, of course, is not all

capable of being considered or handled by every

existing segmentation method. So it is very natural

that there is no segmentation method that applies to

all types of images (M. W. Khan, 2014). This is due

to the variety of existing image features (Vidhya et

al., 2016). Each image segmentation technique has its

own advantages and disadvantages. Some of the

existing techniques were developed with a specific

purpose, so that they will be more suitable and

perform better when applied to these applications.

Therefore, the performance measure is more

determined by the application (Faiza Babakano,

2015).

Various approaches have been proposed in image

segmentation, both related to certain aspects and

improvements from previously known methods.

These efforts are practically unceasing to optimize

and refine existing techniques in order to obtain more

perfect results (Phonsa & Manu, 2019). Although

various segmentation techniques have been

developed with more promising results, they are still

a challenge and continue to be developed in an effort

to find better techniques (Fahad & Morris, 2006).

Because there is no universal segmentation technique,

in order to streamline the segmentation process, these

various techniques often have to be combined in their

application (Zaitoun & Aqel, 2015). The selection of

this segmentation technique and the level of

segmentation carried out are determined based on the

type of image and the characteristics of the problems

being faced related to this image (B. Kaur & Kaur,

2015). A good segmentation technique is a

segmentation technique that is able to maintain image

features in an efficient and short computation time (D.

Rasi, 2016). The existing images are of course very

diverse, both in terms of color, intensity, texture,

features and so on, which are used as a basis for

segmentation. It becomes very natural that a

segmentation technique cannot be applied in general.

By considering the various advantages of each

existing technique, it is necessary to consider an

alternative combination of these existing techniques

(Song & Yan, 2017).

Some image segmentation techniques in principle

work based on the gray level of an image. But in fact,

this technique works even better when applied to

color images. This is possible because the colors in an

image have a lot of information and are clear (Shakti,

2013). Knowledge of pixel color is the basis for

developing the segmentation method (K. & S., 2014).

The existence of pixels in an image greatly affects the

image segmentation process. Although the

segmentation process can be carried out on the basis

of the features of the image, the results will not be

good if it ignores the existence of pixels and the

relationship between these pixels. This is due to the

existence of pixels that often overlap between regions

in an image (Karmakar & Dooley, 2002). The

segmentation process which is based on a certain

combination of color intensity and texture gives better

segmentation results (Taneja et al., 2015).

There are 3 (three) factors that can be considered

as assessing whether a segmentation method is good

or not. The three factors are precision (reliability),

accuracy (validity), and efficiency (viability). The

difficulty is that each of these factors affects one

another. Improvement in one factor will impact on

other factors (Udupa et al., 2006). There are some

general guidelines that can be applied in choosing a

segmentation technique. If the benchmark is pixel

clusters, then area-based segmentation techniques are

most appropriate. If based on pixel classification,

without considering the cluster connection, then the

clustering segmentation technique is better

(Hosseinzadeh & Khoshvaght, 2015). To get good

results and apply to various applications and images

is certainly not easy. Continuous research is needed

to develop better segmentation techniques. One

alternative that needs to be considered is to involve

user interaction (Chandel et al., 2012).

In this research, we will compare various well-

known and widely used segmentation techniques. The

aim is to get to know and at the same time know the

advantages and disadvantages of each of these

segmentation techniques. By understanding the

characteristics of each of these segmentation

techniques, it will be easier to choose the appropriate

technique for the segmentation problem at hand.

2 IMAGE SEGMENTATION

TECHNIQUES

As previously mentioned, there are various

techniques and algorithms that have been applied in

image segmentation. These various techniques and

algorithms are then grouped with various approaches

as well. The groupings include:

ICE-TES 2021 - International Conference on Emerging Issues in Technology, Engineering, and Science

132

a. contextual technique and non-contextual

technique.

The segmentation technique is grouped on the

basis of its treatment of the features possessed by

the image. The contextual technique fully

considers the relationship between image features,

whereas the non-contextual technique completely

ignores (Sekar & Ilanchezhian, 2015);

b. based on discontinuity dan based on similarity.

The segmentation technique is grouped based

on how to partition the image based on the

intensity of the gray level of the image. In the

based on discontinuity technique, image

partitioning is carried out when there is a sudden

change, whereas in the based on similarity

technique, image correction is carried out on the

basis of the similarity of the gray level of the

image (Matta, 2014; Narkhede, 2013; Sonawane

& Dhawale, 2015);

c. structural, stochastic and hybrid techniques.

In structural techniques, image sorting is

carried out based on the structural information of

the image part, while in stochastic techniques it is

carried out based on the discrete pixel value of the

image (D. Kaur & Kaur, 2014). Hybrid technique

is a segmentation technique that combines both

techniques, namely using discrete pixels and

structural information together (Inderpal &

Dinesh, 2014).

Regardless of how they are grouped, each of these

techniques has a different way from one another,

which will be described in the following section.

2.1 Threshold Method

This method performs image segmentation based on

a certain threshold value, which is usually

predetermined. By specifying an appropriate

threshold value, a gray image can be converted to a

binary image. In general, the conversion of a gray

image to a binary image is carried out by referring to

a threshold value (T). Furthermore, for all gray level

values of the image that are equal to or less than T,

they are classified as black (0), while otherwise

classified as white (1). Because this method works on

the basis of a gray image, the color image will be

converted to a gray image first. This threshold

resulting image can be defined as (Yan et al., 2005):



x,y



 

1 if f



x,y



T

0 if f



x,y



T

(1)

Referring to the properties of an image, this

method is then applied with several variants, namely

global, local and adaptive. The global threshold

variant is applied to a segmentation process based on

the gray level values. The segmentation process,

which refers to the gray level values and local

properties of the image, applies the local threshold

variant. Whereas for the segmentation process based

on the gray level value, the environment and pixel

coordinate properties of the image will apply an

adaptive threshold variant (Al-Amri & Kalyankar,

2010).

This method partitions the image using a

threshold value. Image segmentation is done by

comparing the threshold value with all image pixels.

Each image pixel is then given an object label or

image background (Vidhya et al., 2016) (Kang et al.,

2009). This method has the advantage of simple

calculations and fast operation. Excellent

performance is produced when the image has

contrasting objects and backgrounds. Poor results

will be obtained when the images do not have

sufficient gray scale differences (Yuheng & Hao,

2017). The key to the success and weakness of this

method is the accuracy in selecting / determining the

threshold value (Sivakumar & Meenakshi, 2016).

2.2 Edge based Method

This method performs the segmentation process by

finding the pixels that are the boundaries of the

objects contained in an image. This boundary is a

closed area and the number of closed areas is equal to

the number of objects in the image (Rani & Sharma,

2012). Previously, the image is converted to gray

image first, and edge detection is done with the help

of the operator (Ansari et al., 2017). This method

partitions an image based on a significant change in

image intensity. An edge is a set of pixels that are

related to one another, and at the same time it is the

boundary of two regions that have different gray

values. Therefore, these pixels are commonly referred

to as edge points. This edge point can be determined

based on the approximate intensity gradient of the

pixels (Senthilkumaran & Rajesh, 2009).

There are various edge detectors applied to this

edge-based segmentation method (Pardhi & Wanjale,

2016; Yogamangalam & Karthikeyan, 2013). This is

intended to simplify image analysis. Detector

functions to reduce data processing, but on the other

hand, structural information related to object

boundaries can still be maintained (Canny, 1986).

Different techniques are used to perform edge

detection and among them the gradient-based method

and the gray histogram are the most commonly used

techniques for edge detection (W. Khan, 2013). In the

gray histogram method, segmentation is based on

separating the foreground from the background by

The Color Dissimilarity based Method among Other Segmentation Methods: A Comparison

133

selecting a threshold value. Whereas in a gradient

based method, segmentation is based on sudden

changes between the intensity of two regions

(Zhengdong, 2015).

There are several variations of this method based

on the type of operator used, namely the Robert,

Sobel, Prewitt, Canny and Laplace operators (Saluja

et al., 2013). Sobel operator is used for edge detection

in two directions (horizontal. Vertical) (Maurya et al.,

2020). This operator is used to calculate the gradient

estimate of the image enrichment function. Similar to

the Sobel operator, the Prewitt operator detects edges

based on the difference in pixel intensity in the image.

2.3 Region based Method

The concept of this method is actually very close to

human perception, that is, features are extracted based

on certain aspects such as ratio, circularity, invariance

and so on from color, shape and texture (S et al.,

2014). This area-based segmentation technique works

with the concept of homogeneity. The image will be

divided into areas where there are connected pixels,

which have similar characteristics, which are

different from other regions (Kaganami & Beiji,

2009). By relating the different regions in this image

to the object and its background, this method is very

suitable for object detection and recognition (Lalaoui

& Mohamadi, 2013). This method performs sorting

of areas on certain criteria which include color,

intensity or object. In order to obtain a wider area,

taking into account the characteristics of existing

pixels, an area-based segmentation method was

developed, namely region growing and region

splitting and merging (Sivakumar & Meenakshi,

2016).

In the region growing method, the formation of a

homogeneous region begins with a pixel at a certain

position. Then this area is expanded by adding

adjacent pixels that have the same property, namely

the predetermined homogeneity criteria. The

similarity rate of pixels in a certain region will be

greater than the similarity rate in other regions

(Qiong, 2015). Properties that can be used as criteria

include grayscale level, texture, color or shape. Then

this criterion will determine whether a pixel next to it

can join in the region or not (Szénási, 2014). The

advantage of this method is its ability to separate

areas with certain characteristics and is able to show

boundaries well. Unfortunately, the high computation

costs, uneven grayscale and noise are not handled

properly, so the results are not perfect (Yuheng &

Hao, 2017).

In the technique of separation and merging of

regions, an image is divided into four quadrants with

certain criteria to distinguish its homogeneity. Sorting

is done repeatedly with the same criteria, until sorting

is no longer possible. The sorting results are then

recombined to get the desired results (Kaganami &

Beiji, 2009; Kang et al., 2009).

2.4 Clustering based Method

This method refers to the process of grouping the

same or similar data into a particular group or cluster.

In the image segmentation process, this method

classifies the image based on the pixels in it. This

method differentiates and classifies pixels according

to certain requirements and rules. Pixels are classified

according to similarity using a mathematical

algorithm (Wang & Yang, 2010). Pixels with the

same characteristics are grouped into a cluster. As a

result, the image will be divided into a number of

clusters with different pixel coherence for each

cluster (Chandhok et al., 2012).

The grouping of pixels into clusters is based on

the principle of maximizing the similarity of pixels in

a cluster and maximizing the differences between

clusters. The process is carried out repeatedly by

paying attention to the characteristics of each cluster

formed (Jain et al., 1999).

Clustering techniques can be classified into two

categories, namely hard clustering and soft clustering.

In hard clustering, it emphasizes that there are

differences or clear boundaries between one cluster

and another. A pixel belongs to one cluster only. In

soft clustering, pixel grouping is based on several

similarity criteria. Similar pixels will merge into the

same cluster. The similarity criteria used can be

distance, connectivity or intensity (Sravani & Deepa,

2013).

One of the most popular and widely used

clustering methods is K-Means clustering (Acharya et

al., 2013). A very simple and fast clustering

algorithm, which aims to group data into K clusters

(Luo et al., 2003). Grouping is carried out on the basis

of the proximity of the grouped elements to the center

of the cluster. The distance calculation is done on the

various properties which form the basis of grouping

(Bhatia, 2004). In color image segmentation, besides

using RGB space, the K-Means algorithm can also be

applied to the HIS space, so that the hue and intensity

components can be fully considered (Zhang & Wang,

2000).

In the K-Means method, the number of clusters

has been determined, with the center of the cluster

being initially determined randomly. Pixels are then

ICE-TES 2021 - International Conference on Emerging Issues in Technology, Engineering, and Science

134

grouped into a cluster based on their proximity. The

average for each cluster is then calculated, then

proceed with regrouping based on the proximity of

the pixel value to the average value on the cluster.

This process is repeated until there are no significant

changes or in a certain number of repetitions

(Thilagamani & Shanthi, 2011).

2.5 Watershed based Method

This method includes methods that take advantage of

the similarity category in an area, using mathematical

morphological approaches and real-life analysis, such

as areas affected by floods (Saini & Arora, 2014).

Image is considered as a surface gradient topography

(Tang, 2010) with the image pixels having the highest

gradient will be the boundary of an area, such as a

watershed, with continuous boundaries, without gaps

(Seerha & Kaur, 2013). Therefore, to understand the

working concept of this method, imagine a watershed.

An image is analogous to a watershed whose

direction follows the color intensity gradient in the

image. The segmentation process begins by

determining the part of the image that has a minimum

intensity, and serves as the base point for watersheds

(Bleau & Leon, 2000). When it rains, the watershed

will be flooded, forming a puddle area. To prevent

standing water from mixing with puddles in other

areas, a dam or water barrier is built (Vartak &

Mankar, 2013).

The Watershed method in principle succeeds in

utilizing the concept of the concept and property

discontinuity of an image. By applying grayscale

mathematical morphology, this method is able to

distinguish well an object in the foreground from its

background in an image (Rahman & Islam, 2013). Its

weakness is the complexity of the calculation,

sensitivity to blur and often results in excessive image

segmentation (Jayapriya & Hemalatha, 2019).

2.6 PDE based Method

The basic idea of this PDE (Partial Differential

Equations) based method is to convert a certain curve,

surface or image into a partial differential equation

(PDE) with certain initial conditions and limitations.

The segmentation results are obtained from the

solution of this equation (Wei & Chan, 2016). In this

method, image segmentation is carried out by

utilizing active contours on an image. The active

contour in the form of a curve in an image is defined

explicitly by using a multidimensional function.

Several PDE methods used for image segmentation

are the Snakes, Level-Set, and Mumford Shah method

(Xin-Jiang et al., 2009).

To overcome the weaknesses of the Snakes

model, an adaptive PDE model has been developed,

namely the fuzzy PDE contour model, which applies

the Fuzzy C-Means classification to the PDE

geometric contours when segmenting images. The

aim is to find the boundary of the segmentation area

of an image (Bueno et al., 2004).

The problem of changing the object topology in

the image is resolved naturally by developing a level

set function (Sliž & Mikulka, 2016). In the Level-Set

method, the curve or surface of an image is

represented as a set of zero levels of a higher

dimensional surface. This method turns out to provide

a more accurate numerical implementation. Thus, this

method is also able to better handle topological

problems (Xin-Jiang et al., 2009). The Level-Set

method is quite simple and adaptable for calculating

and analyzing the interface changes of an image,

either two or three dimensions (Zhou et al., 2010).

The PDE method is proven to be able to overcome

the geometric complexity of an image. The image

generated from the segmentation process using the

PDE method can increase the texture contrast of the

image (Sofou & Maragos, 2008). With this capability,

the PDE method is believed to be able to produce a

better segmentation process than other methods

(Shahzad et al., 2008).

Apart from segmentation purposes, the PDE

method is also suitable for denoising and enhancing

image coherence (Yoruk & Akgul, 2004). The

application of a nonlinear isotropic diffusion filter to

the PDE method will allow the edge enhancement of

an image. The algorithm used is additive operator

splitting (AOS), which allows the application of

filters recursively and separately. To produce good

denoising, the PDE method applies the non-convex

variational image recovery method. In this case an

AOS scheme is used in the decomposition of the

Gaussian pyramid (Weickert, 2001).

2.7 ANN based Method

This method is actually inspired by the human

nervous system model, by simulating the learning

process carried out by the human brain. This neural

network or neuron is a parallel system that works with

guidance and has the ability to learn, just like humans

(Amanpreet Kaur & Kaur, 2015). In implementation,

this artificial neural network system is trained with a

number of training samples, as if the system is

involved in the learning process (S. Agrawal & Xaxa,

2014). The system receives input, processes and

The Color Dissimilarity based Method among Other Segmentation Methods: A Comparison

135

provides output based on a familiar concept or pattern

(Amritpal Kaur & Kaur, 2014).

In its development, currently various neural

network models have been developed with a variety

of learning algorithms, exploration methods and

various analysis methods as well (Tianhao &

Tianzhen, 2005). In the image segmentation process,

various approaches are used with different levels of

success. In general, there are two categories of

network methods namely supervised and

unsupervised methods. In the supervised method, it

takes the presence of an expert to supervise and

simultaneously carry out the learning process in the

system. Whereas in the unsupervised method, the

practical system works automatically. Even if expert

involvement exists, it is minimal and intended to

improve system performance (Amza, 2012).

This ANN paradigm has been widely used in

image processing techniques, including in image

segmentation. One of the ANN methods that is

considered feasible for image segmentation is the

PCNN (Pulse-Coupled Neural Network) model

(Carata & Neagoe, 2016). A model inspired by the cat

visual system by simply modeling the cortical

neurons in the visual area of the cat brain (Cheng et

al., 2008), is believed to be able to produce good

segmentation results, even though the input image is

in poor condition. Each pixel of the image is

represented by a neuron. The initial conditions of

these pixels and their environment will determine the

state of the neurons. These neurons will form a

network of temporal pulses, which are presented in a

two-dimensional array of laterally connected layers

of neurons and pulses. This neural network is then

used as the basis for image processing applications

(Kuntimad & Ranganath, 1999) (Harris et al., 2015).

Image segmentation is performed based on temporal

correlation which depends on lateral modulation and

bridging waves to synchronize pixels in the same area

(Zhan et al., 2017). Related to the problem of setting

PCNN parameters in the image segmentation process,

it can be overcome by applying artificial intelligence

techniques, for example by applying algorithms based

on particle swarm optimization (PSO), genetic

algorithms (GA) and differential evolution (DE)

(Hernández & Gómez, 2016).

2.8 Color Dissimilarity based Method

This segmentation method works following the

concept of human work identifying objects. An object

can be recognized by humans, because the eye's

ability to distinguish the color of the object. Each

color has its own different R, G and B values, with a

value range of 0 - 255. That is, conceptually, there are

millions (256x256x256) of color types available.

However, not all of these colors can be recognized /

distinguished by humans. Apart from the lighting

factor, the color similarity factor is also difficult to

identify. A new color can be distinguished when the

value of R, G or B, singly, in pairs or all three, differs

from a certain value (Karma, 2020).

By comparing two adjacent pixels in an image, the

segmentation process is carried out with the concept:

( k

ꓴ k

)  f(x,y,i) = 255 (2)

where:

= (∆R

ꓴ ∆R

ꓴ ∆G

ꓴ ∆B

)

≥ 9

= ((∆R

ꓵ ∆G

) ꓴ (∆R

ꓵ ∆B

) ꓴ (∆G

ꓵ

∆B

) ꓴ (∆R

ꓵ ∆B

) ꓴ (∆R

ꓵ ∆B

) ꓴ

(∆G

ꓵ ∆B

)) ≥ 7

= ((∆R

ꓵ ∆G

ꓵ ∆B

) ꓴ (∆R

ꓵ ∆G

ꓵ

∆B

)) ≥ 6

The results obtained are then converted into black

and white images, and to clarify and reinforce the

results, it ends with a morphological process (Karma

et al., 2020).

3 RESULTS AND DISCUSSION

Comparisons of the various segmentation methods

discussed earlier are carried out on the basis of the

results they provide. The main criteria for comparison

is the extent to which the results given by each

method are able to lead us to the detection of objects

in the image. The extent to which each method is able

to produce a pattern that is the ground-truth of the

object contained in an image. Image segmentation

process is a process that plays an important role in the

detection and recognition of objects. Therefore, a

segmentation process is declared good, if the process

is able to provide results that facilitate the process of

detecting and recognizing the object.

Testing is carried out using the MATLAB

program code from each method on 5 (five) existing

samples. In this study, comparisons were made of the

results of the segmentation of the proposed Color

Dissimilarity method with other methods, such as

Thresholding (Bhosale, 2015), Edge (Rajan, 2016),

Region Growing (Kroon, 2008), K-Means Clustering

(Analyst, 2018), Watershed (MathWorks, n.d.),

Fuzzy C-Mean (Li et al., 2011) and Pulse Coupled

Neural Network (Hernández & Gómez, 2016).

Sample images and results from the segmentation

process of each method are presented in Table 1.

ICE-TES 2021 - International Conference on Emerging Issues in Technology, Engineering, and Science

136

Table 1: Sample image and results of each method.

Sample image

Thresholding

Edge

Region growing

KMeansClustering

Watershed

Fuzzy C-Mean

Pulse Coupled Neural

Network

Color dissimilarity

(Proposed Method)

The Color Dissimilarity based Method among Other Segmentation Methods: A Comparison

137

This study only uses 5 (five) samples, with sample

characteristics that are expected to represent the

characteristics of the image in general. The

characteristics of the image used are colored or gray

images, images that have objects with different

background colors or close to the object's color, and

images that have several objects attached or

separated. This study wants to find out which

segmentation method is able to provide better

segmentation results, in terms of our ease in detecting

objects in the image.

As previously stated, in general, there is no perfect

segmentation method that is capable of providing

consistent and always good results. Based on the

results of tests carried out on the five existing

samples, a number of methods seemed unsuitable for

use in the process of detection and object recognition,

as desired. This is due to the inability of this method

to produce a pattern or ground-truth of the object in

the image. However, it must be admitted that there are

relatively good results for certain samples.

When examined from the results given, the

Watershed method shows the most undesirable

results, because it is practically unable to produce

patterns from existing objects. The K-Means

Clustering method is relatively the same as the

Watershed method, except in sample number 2, this

method is able to show patterns of existing objects.

Fuzzy C-Mean and Pulse Coupled Neural Network

methods give practically the same results, with

unclear and inconsistent patterns, as in the results

from sample number 5. The same thing happens to the

Thresholding method. There is an object pattern in

sample number 4, but it is wrong in sample number 5.

The Region Growing method does not appear to

produce a pattern from existing objects, except to

change the coloring of the original image. The Edge

method is able to produce clear borders based on

color differences in the image. These borders provide

a detailed pattern and tend to be complex and difficult

to spot. The relatively good pattern is actually

produced by the color dissimilarity method. Of the

five existing samples, three samples were able to

show their object patterns clearly. Meanwhile, the

other two, number 1 and 3, formed a pattern, but it did

not clearly show the pattern of the object.

4 CONCLUSIONS

This research also shows that there is no perfect and

generally applicable segmentation method. However,

the Color Dissimilarity method turned out to be able

to produce a pattern or ground-truth of the objects

contained in the segmented image. This means that

this method is very fixed when applied in the process

of detecting and recognizing objects in an image.

REFERENCES

Acharya, J., Gadhiya, S., & Raviya, K. (2013).

Segmentation techniques for image analysis: A review.

International Journal of Computer Science and

Management Research, 2(1), 1218–1221.

Agrawal, P., Shriwastava, S. K., & Limaye, S. S. (2010).

MATLAB implementation of image segmentation

algorithms. Computer Science and Information

Technology (ICCSIT), 2010 3rd IEEE International

Conference On, 3, 427–431.

Agrawal, S., & Xaxa, D. K. (2014). Survey on Image

Segmentation Techniques and Color Models. Savita

Agrawal et Al,/(IJCSIT) International Journal of

Computer Science and Information Technologies, 5(3),

3025–3030.

Al-Amri, S. S., & Kalyankar, N. V. (2010). Image

segmentation by using threshold techniques. Journal of

Computing, 2(5), 83–86.

Amza, C. (2012). A review on neural network–based image

segmentation techniques. De Montfort University,

Mechanical and Manufacturing Engg., The

GatewayLeicester, LE1 9BH, United Kingdom, 1–23.

Analyst, I. (2018). Image segmentation using k means

clustering.

https://www.mathworks.com/matlabcentral/answers/3

92715-image-segmentation-using-k-means-clustering

Ansari, M. A., Kurchaniya, D., & Dixit, M. (2017). A

Comprehensive Analysis of Image Edge Detection

Techniques. International Journal of Multimedia and

Ubiquitous Engineering, 12(11), 1–12.

doi:10.14257/ijmue.2017.12.11.01

Bhatia, S. K. (2004). Adaptive K-Means Clustering.

FLAIRS Conference, 695–699.

Bhosale, A. (2015). Segmentation using Thresholding.

MATLAB Central File Exchange.

https://www.mathworks.com/matlabcentral/fileexchan

ge/50288-segmentation-using-thresholding

Bleau, A., & Leon, L. J. (2000). Watershed-based

segmentation and region merging. Computer Vision

and Image Understanding.

doi:10.1006/cviu.1999.0822

Bueno, G., Martínez-Albalá, A., & Cosías, P. (2004).

Fuzziness and PDE based models for the segmentation

of medical images. IEEE Nuclear Science Symposium

Conference Record. doi:10.1109/nssmic.2004.1466702

Canny, J. (1986). A computational approach to edge

detection. IEEE Transactions on Pattern Analysis and

Machine Intelligence, PAMI-8(6), 679–698.

Carata, S., & Neagoe, V. (2016). A Pulse-Coupled Neural

Network approach for image segmentation and its

pattern recognition application. 2016 International

Conference on Communications (COMM), 61–64.

doi:10.1109/ICComm.2016.7528317

ICE-TES 2021 - International Conference on Emerging Issues in Technology, Engineering, and Science

138

Chandel, G. S., Kumar, R., Khare, D., & Verma, S. (2012).

Analysis of Image Segmentation Algorithms using

MATLAB. International Journal of Engineering

Innovation & Research, 1(1), 51–55.

Chandhok, M. C., Chaturvedi, S., & Khurshid, A. A.

(2012). An approach to image segmentation using K-

means clustering algorithm. International Journal of

Information Technology (IJIT), 1(1), 11–17.

Cheng, D., Zhao, W., Tang, X., & Liu, J. (2008). Image

segmentation based on pulse coupled neural network.

11th Joint International Conference on Information

Sciences.

D. Rasi, J. S. (2016). A Survey on Image segmentation

algorithms. International Journal of Computer Trends

and Technology (IJCTT), 35(4), 170–174.

doi:10.14445/22312803/IJCTT-V35P132

Dass, R., Priyanka, & Devi, S. (2012). Image Segmentation

Techniques. International Journal of Electronics &

Communication Technology, 3(I), 66–70.

Fahad, A., & Morris, T. (2006). A faster graph-based

segmentation algorithm with statistical region merge.

Advances in Visual Computing, 286–293.

Faiza Babakano, J. (2015). Performance Metrics for Image

Segmentation Techniques: A Review.

Harris, M. A., Van, A. N., Malik, B. H., Jabbour, J. M., &

Maitland, K. C. (2015). A pulse coupled neural network

segmentation algorithm for reflectance confocal images

of epithelial tissue. PloS One, 10(3), e0122368.

Hernández, J., & Gómez, W. (2016). Automatic tuning of

the pulse-coupled neural network using differential

evolution for image segmentation. Mexican Conference

on Pattern Recognition, 157–166.

Hosseinzadeh, M., & Khoshvaght, P. (2015). A

Comparative Study of Image Segmentation

Algorithms. World Academy of Science, Engineering

and Technology, International Journal of Computer,

Electrical, Automation, Control and Information

Engineering, 9(8), 1966–1971.

Inderpal, S., & Dinesh, K. (2014). A Review on Different

Image Segmentation Techniques. IJAR, 4.

Ivanovici, M., Richard, N., & Paulus, D. (2013). Color

image segmentation. In Advanced Color Image

Processing and Analysis. doi:10.1007/978-1-4419-

6190-7_8

Jain, A. K., Murty, M. N., & Flynn, P. J. (1999). Data

clustering: a review. ACM Computing Surveys (CSUR),

31(3), 264–323.

Jayagar, D., & Jeyakumari, D. (2015). A Survey of Various

Image Segmentation Techniques. International Journal

of Modern Trends in Engineering and Research,

02(05), 273–278.

Jayapriya, P., & Hemalatha, S. (2019). Comparative

analysis of image segmentation techniques and its

algorithm. International Journal of Scientific and

Technology Research, 8(10), 2209–2212.

Jun, T. (2010). A color image segmentation algorithm

based on region growing. ICCET 2010 - 2010

International Conference on Computer Engineering

and Technology, Proceedings.

doi:10.1109/ICCET.2010.5486012

K., R. K., & S., S. S. (2014). Review of Image

Segmentation Techniques: A Survey. International

Journal of Advanced Research in Computer Science

and Software Engineering, 4(7), 842–845.

Kaganami, H. G., & Beiji, Z. (2009). Region-based

segmentation versus edge detection. 2009 Fifth

International Conference on Intelligent Information

Hiding and Multimedia Signal Processing, 1217–1221.

Kang, W.-X., Yang, Q.-Q., & Liang, R.-P. (2009). The

comparative research on image segmentation

algorithms. Education Technology and Computer

Science, 2009. ETCS’09. First International Workshop

On, 2, 703–707.

Karma, I. G. M. (2020). Determination and Measurement

of Color Dissimilarity. International Journal of

Engineering and Emerging Technology, 5(1), 67–71.

https://ojs.unud.ac.id/index.php/ijeet/article/view/5763

Karma, I. G. M., Putra, I. K. G. D., Sudarma, M., &

Linawati. (2020). Image Segmentation Based on Color

Dissimilarity.

Karmakar, G. C., & Dooley, L. S. (2002). A generic fuzzy

rule based image segmentation algorithm. Pattern

Recognition Letters, 23(10), 1215–1227.

Kaur, Amanpreet, & Kaur, N. (2015). Image Segmentation

Techniques. International Research Journal of

Engineering and Technology, 02(02), 944–947.

Kaur, Amritpal, & Kaur, M. A. (2014). Evaluation of

Parameters of Image Segmentation Algorithms-JSEG

& ANN. International Journal of Advanced Research

in Electronics and Communication Engineering, 3(8),

854–865.

Kaur, B., & Kaur, P. (2015). A Comparative Study on

Image Segmentation Techniques. International Journal

of Computer Science and Engineering, 3(12), 50–56.

Kaur, D., & Kaur, Y. (2014). Various Image Segmentation

Techniques: A Review. International Journal of

Computer Science and Mobile Computing.

Khan, M. W. (2014). A survey: Image segmentation

techniques. International Journal of Future Computer

and Communication, 3(2), 89.

Khan, W. (2013). Image Segmentation Techniques: A

Survey. Journal of Image and Graphics, 1(4), 166–170.

doi:10.12720/joig.1.4.166-170

Kroon, D.-J. (2008). Region Growing. MATLAB Central

File Exchange.

https://www.mathworks.com/matlabcentral/fileexchan

ge/19084-region-growing

Kumar, M. J., Kumar, D. G. V. S. R., & Reddy, R. V. K.

(2014). Review on Image Segmentation Techniques.

International Journal of Science Research Engineering

& Technology (IJSRET), 3(6), 992–997.

Kuntimad, G., & Ranganath, H. S. (1999). Perfect image

segmentation using pulse coupled neural networks.

IEEE Transactions on Neural Networks, 10(3), 591–

598. doi:10.1109/72.761716

Lalaoui, L., & Mohamadi, T. (2013). A comparative study

of image region-based segmentation algorithms.

International Journal of Advanced Computer Science

and Applications, 4.

The Color Dissimilarity based Method among Other Segmentation Methods: A Comparison

139

Li, B. N., Chui, C. K., Chang, S., & Ong, S. H. (2011).

Integrating spatial fuzzy clustering with level set

methods for automated medical image segmentation.

Computers in Biology and Medicine, 41(1), 1–10.

Luo, M., Ma, Y.-F., & Zhang, H.-J. (2003). A Spatial

Constrained K-Means Approach to Image

Segmentation. Information, Communications and

Signal Processing, 2003 and Fourth Pacific Rim

Conference on Multimedia. Proceedings of the 2003

Joint Conference of the Fourth International

Conference On, 2, 738–742.

MathWorks. (n.d.). Marker-Controlled Watershed

Segmentation. Retrieved January 29, 2018, from

https://www.mathworks.com/help/images//marker-

controlled-watershed-segmentation.html

Matta, S. (2014). Review: Various image segmentation

techniques. International Journal of Computer Science

and Information Technologies (IJCSIT), 5(6), 7536–

7539.

Maurya, P., Jadhav, A., Mahadik, S., & Pansambal, S.

(2020). Edge Detection Methods in Image

Segmentation. International Journal of Future

Generation Communication and Networking, 13(1),

252–256.

Narkhede, H. P. (2013). Review of Image Segmentation

Techniques. International Journal of Science and

Modern Engineering.

Pardhi, S., & Wanjale, M. K. H. (2016). Survey on

Techniques Involved in Image Segmentation.

International Journal of Computer Science Trends and

Technology (IJCST), 4(3), 275–280.

Phonsa, G., & Manu, K. (2019). A Survey: Image

Segmentation Techniques. In Advances in Intelligent

Systems and Computing (pp. 1123–1140).

doi:10.1007/978-981-13-0761-4_105

Qiong, P. (2015). An Image Segmentation Algorithm

Research Based on Region Growth. Journal of

Software Engineering, 9(3), 678–679.

Rahman, M. H., & Islam, M. R. (2013). Segmentation of

color image using adaptive thresholding and masking

with watershed algorithm. 2013 International

Conference on Informatics, Electronics and Vision

(ICIEV), 1–6. doi:10.1109/iciev.2013.6572557

Rajan, J. (2016). Edge Detection of Color. MATLAB

Central File Exchange.

https://www.mathworks.com/matlabcentral/fileexchan

ge/8267-edge-detection-of-color-images

Rani, V., & Sharma, D. (2012). A study of edge-detection

methods. International Journal of Science, Engineering

and Technology Research (IJSETR), 1(6), 62–65.

S, R. E., V, P. D., & Samuel, S. J. (2014). A Survey on

Region Based Image Segmentation. International

Journal of Advanced in Science Engineering and

Technology,

2(2), 3–5.

Saini, S., & Arora, K. (2014). A study analysis on the

different image segmentation techniques. International

Journal of Information & Computation Technology,

4(14), 1445–1452.

Saluja, S., Singh, A. K., & Agrawal, S. (2013). A Study of

Edge-Detection Methods. International Journal of

Advanced Research in Computer and Communication

Engineering.

Seerha, G. K., & Kaur, R. (2013). Review on recent image

segmentation techniques. International Journal on

Computer Science and Engineering, 5(2), 109–112.

Sekar, A. R. R., & Ilanchezhian, P. (2015). A Survey on

Image Segmentation Techniques. International Journal

of Science and Research (IJSR), 4(2), 2304–2306.

Senthilkumaran, N., & Rajesh, R. (2009). A study on edge

detection methods for image segmentation.

Proceedings of the International Conference on

Mathematics and Computer Science (ICMCS-2009), 1,

255–259.

Shahzad, A., Sharif, M., Raza, M., & Hussain, K. (2008).

Enhanced watershed image processing segmentation.

Journal of Information & Communication Technology

(JICT), 2(1), 9.

Shakti, S. (2013). Comparative study of various image

segmentation methods. Int. J. Multidisciplinary Acad.

Res, 2(3), 1–12.

Sivakumar, P., & Meenakshi, S. (2016). A Review on

Image Segmentation Techniques. International Journal

of Advanced Research in Computer Engineering &

Technology (IJARCET), 5(3), 641–647.

Sliž, J., & Mikulka, J. (2016). Advanced image

segmentation methods using partial differential

equations: a concise comparison. 2016 Progress in

Electromagnetic Research Symposium (PIERS), 1809–

1812.

Sofou, A., & Maragos, P. (2008). Generalized flooding and

multicue PDE-based image segmentation. IEEE

Transactions on Image Processing.

doi:10.1109/TIP.2007.916156

Sonawane, M. S., & Dhawale, C. A. (2015). A Brief Survey

on Image Segmentation Methods. National Conference

on Digital Image and Signal Processing, DISP 2015, 1.

Song, Y., & Yan, H. (2017). Image Segmentation

Techniques Overview. 2017 Asia Modelling

Symposium (AMS), 103–107.

Sravani, P., & Deepa, S. (2013). A Survey on Image

Segmentation Techniques and Clustering. International

Journal of Advance Research in Computer Science and

Management Studies, Special Issue.

Szénási, S. (2014). Distributed Region Growing Algorithm

for Medical Image Segmentation. International Journal

of Circuits, Systems and Signal Processing, 8(1), 173–

181.

Taneja, A., Ranjan, P., & Ujjlayan, A. (2015). A

performance study of image segmentation techniques.

Reliability, Infocom Technologies and Optimization

(ICRITO)(Trends and Future Directions), 2015 4th

International Conference On, 1–6.

Tang, J. (2010). A color image segmentation algorithm

based on region growing. 2010 2nd International

Conference on Computer Engineering and Technology,

6, V6-634.

Thilagamani, S., & Shanthi, N. (2011). A survey on image

segmentation through clustering. International Journal

of Research and Reviews in Information Sciences, 1(1),

14–17.

ICE-TES 2021 - International Conference on Emerging Issues in Technology, Engineering, and Science

140

Tianhao, T., & Tianzhen, W. (2005). An ANN-based

clustering analysis algorithm with dynamic data

window. Control and Automation, 2005. ICCA’05.

International Conference On, 1, 581–586.

Udupa, J. K., LeBlanc, V. R., Zhuge, Y., Imielinska, C.,

Schmidt, H., Currie, L. M., Hirsch, B. E., & Woodburn,

J. (2006). A framework for evaluating image

segmentation algorithms. Computerized Medical

Imaging and Graphics.

doi:10.1016/j.compmedimag.2005.12.001

Vartak, A. P., & Mankar, V. (2013). Colour image

segmentation-A survey. International Journal of

Emerging Technology and Advanced Engineering,

3(2), 681–688.

Vidhya, K., Revathi, S., Ashwini, S. S. S., & Vanitha, S.

(2016). Review on Digital Image Segmentation

Techniques. International Research Journal of

Engineering and Technology (IRJET), 03(02), 618–

619.

Wang, Z., & Yang, M. (2010). A fast clustering algorithm

in image segmentation. Computer Engineering and

Technology (ICCET), 2010 2nd International

Conference On, 6, V6-592-V6-594.

Wei, J., & Chan, L. (2016). An Image Segmentation

Method Based on Partial Differential Equation Models.

International Journal of Simulation--Systems, Science

& Technology, 17(36).

Weickert, J. (2001). Efficient image segmentation using

partial differential equations and morphology. Pattern

Recognition, 34(9), 1813–1824.

doi:doi:10.1016/S0031-3203(00)00109-6

Xin-Jiang, Renjie-Zhang, & Shengdong-Nie. (2009). Image

segmentation based on PDEs model: A survey. 3rd

International Conference on Bioinformatics and

Biomedical Engineering, ICBBE 2009.

doi:10.1109/ICBBE.2009.5162922

Yan, F., Zhang, H., & Kube, C. R. (2005). A multistage

adaptive thresholding method. Pattern Recognition

Letters, 26(8), 1183–1191.

Yogamangalam, R., & Karthikeyan, B. (2013).

Segmentation Techniques Comparison in Image

Processing. International Journal of Engineering and

Technology (IJET), 5(1), 307–313.

Yoruk, E., & Akgul, C. B. (2004). Color image

segmentation using PDE-based regularization and

watershed transformation. Proceedings of the IEEE

12th Signal Processing and Communications

Applications Conference, 2004., 518–521.

Yuheng, S., & Hao, Y. (2017). Image segmentation

algorithms overview. ArXiv Preprint

ArXiv:1707.02051.

Zaitoun, N. M., & Aqel, M. J. (2015). Survey on Image

Segmentation Techniques. Procedia Computer

Science. doi:10.1016/j.procs.2015.09.027

Zhan, K., Shi, J., Wang, H., Xie, Y., & Li, Q. (2017).

Computational mechanisms of pulse-coupled neural

networks: a comprehensive review. Archives of

Computational Methods in Engineering, 24(3), 573–

588.

Zhang, C., & Wang, P. (2000). A new method of color

image segmentation based on intensity and hue

clustering. Pattern Recognition, 2000. Proceedings.

15th International Conference On, 3, 613–616.

Zhengdong, L. (2015). A Garment Image Segmentation

Method Based on Salient Region and JSEG. JSW,

10(11), 1274–1282.

Zhou, B., Xiao, L. I. Y., & Liu, R. (2010). Image

segmentation with partial differential equations.

Information Technology Journal, 9(5), 1049–1052.

The Color Dissimilarity based Method among Other Segmentation Methods: A Comparison

141