International Conference on Advances in Computing, Communication Control and Networking (ICACCCN2018)

Use of Machine Learning in the Area of Image

Analysis and Processing
Youddha Beer Singh Aditya Dev Mishra Parma Nand
Computer Science & Engineering Computer Science & Engineering Computer Science & Engineering
GCET GCET Sharda University, Greater Noida,
Greater Noida, India Greater Noida, India India
[email protected] [email protected] [email protected]

Abstract: This report tries to investigate the assumed the group pixels that will identified that
effective use of the machine learning in the area of object was known. Here, we will focus on the
image analysis and processing. The field of methods that will find the particular pixels that
healthcare taking a large datasets of image like X- make an object. Large varieties of segmentation
ray, Ultrasound, MRI, CT scan, Echo- method are proposed in the past decades, and
cardiograpgy with the known classification, Take some categorization is necessary to present the
images as input and then process the images, in methods properly here. A disjoint categorization
case of process the images following steps are does not seem to be possible though, because
involved formatting of images if not suitable as even two very different segmentation approaches
may share properties that defy singular
per the required format and then Segmentation,
categorization1. The categorization presented in
Feature extraction and Classification and then
this chapter is therefore rather a categorization
train the model with given data of images with
regarding the emphasis of an approach than a
known classification (that is called data sets) after strict division” [2].
that model can predict the classification of the
given new images that is called test data. And then There are the following types of segmentation are
compare the result with the already given model used
result with different-different machine learning A. Threshold based segmentation
algorithm. And try to improve the result with
existing result and target is to achieve 100 % In this type of segmentation technique
accuracy. To improve the result is depend on defining the Threshold value and based on that
many things like in image processing it depend on value slicing the image. And it will be used to
how well had done segmentation and then feature the given images, It can also combine with
extraction is again main factor to improve the preprocessing and postprocessing techniques
result, how many feature taken during developing [2].
model and then classification, it also depend on B. Edge based segmentation
machine learning algorithm used as per the given
data sets. In the technique edge based identify the edges
in the given image in a way that they are
Keywords: Segmentation, Feature Extraction, representing the object boundaries[2].
Classification. C. Region based segmentation.
In the technique edge based identify the edges
I. Introduction
in the given image in a way that they are
“The division of the given image into meaningful representing the object and by filling that
format, image segmentation, is mostly an object in the specified region based on the
important step in image analysis, the object segmentation technique[2].
representation/visualization, and many different
other image processing. Mostly we focus that how
to analyze and how to represent an object, but

ISBN: 978-1-5386-4119-4/18/$31.00 ©2018 IEEE 117

Authorized licensed use limited to: REVA UNIVERSITY. Downloaded on June 19,2025 at 10:23:29 UTC from IEEE Xplore. Restrictions apply.
 International Conference on Advances in Computing, Communication Control and Networking (ICACCCN2018)

Area of image analysis, the classification

D. Clustering techniques processes is considered as the type of the
supervised learning algorithms, that is the
In this technique segmentation is done based learningconcept where training feature vector
on the cluster or example as synonym and this datasets with known classification is provided.
type of segmentation technique are used in the And the algorithm that will be used for
grouping something like objects [2]. classification is known as the classifier. And In
E. Matching Classification process there is two known
Matching segmentation is used when we have important steps are there training and then the
some idea about the object which we want to testing. And these steps be further divided into
go identify from the given image look like. down into the sub-steps [2][3][4].
And this information we can use to locate the
object in the given image [2]. Table. 1-First Training
Process Description
II. Feature Extraction Pre-processing “Processes the data so it is in a
Feature extraction is used to reduce the large suitable form”.
amount of data into the meaning full and non
redundant data. We compute feature like “number Feature “Reduce the amount of data by
of pixels”, Centroid “center of mass”, “Elliptical extraction extracting relevant
parameters”, Skewness, Kurtosis, “High order information. Usually results in
moments”[2][3]. a vector of scalar values. (We
also need to NORMALIZE the
When input data provided to a machine learning
features for distance
very large amount of data sets then there is
measurements)”
possibility that data may be not meaning full and
there is chance of possibility that redundant data Model “By using the finite set of
is there, so in all these cases feature extraction is Estimation feature vectors, need to
very important. In feature extraction there are two estimate the model for each
steps are involved first one is Feature detection class of the training data”.
and second one is Feature selection [2].
And then Testing
A. Feature detection
As we know that in feature extraction there are Process Description
two steps are involved in which one of them is Pre-processing Same as training
feature detection in the area of image analysis
feature detection is the concept that is used to Feature Same as training
identify the correct object from the given extraction
image[3].
Classification “Compare feature vectors to
B. Feature selection
the various models and find
Feature selection is also one of the important
the closest match. One can
step that is involved in the feature extraction,
use a distance measure”.
to identify the feature from the given image
we should have the idea about the application
or model which we are going to develop,
otherwise it is very difficult to the feature III. Application of Machine Learning
selection. Always done the feature selection In the field of computer science, machine learning
after the feature detection that give the idea is very important area which can be applicable to
about the object [2][3]. the almost everywhere in the branch of computer
science.
C. Classification Process There are some areas where machine learning

118

Authorized licensed use limited to: REVA UNIVERSITY. Downloaded on June 19,2025 at 10:23:29 UTC from IEEE Xplore. Restrictions apply.
 International Conference on Advances in Computing, Communication Control and Networking (ICACCCN2018)

algorithm are used[5][6][7]. image processing and image analysis are used in
1. Machine learning used in Fare many different-different work like Bank note
Detection authentication and healthcare.
2. Machine learning used in application
Face Recognition. V. Objectives and Research Methodology
3. Application of Image classification used A. Objectives
machine learning Several machines learning algorithm are
4. Machine learning is used Speech used for image processing and image
Recognition analysis that is proposed. And algorithms
5. Used in Anti-virus to detect virus that have certain problems each algorithm
6. Detecting spam will behaves in differently way under the
7. In Genetics different conditions. And the any issues in
8. In Weather forecast machine learning the application there will be the particular
used. algorithm should be applied. The objective
of the proposed research is expected to be.
IV. Literature Review  Survey and review of the literature on in
Discussed about the different machine learning the Area of Image Analysis and image
algorithms for image processing and image
 Processing 
analysis and presented a comparative study.  Improvement in the existing proposed
Research includes study of machine learning model of a Machine learning algorithm. 
techniques like the Rule based, contents based, 
 Verification and validation of the
SVM support vector machine and the kernel proposed Model. 
based algorithms for image processing and the
image analysis. And present a comparative B. Research Methodology
analysis for the different image processing Achieving our objectives, we have to follow
techniques and advantages and limitation.
the following methodology:
Nearly in years 2009 the suggested method are
Bayesian algorithm for analysis. And KNN 1. Basic Method
algorithm, decision tree, SVM, and the
 Analyzing issues which used in machine
improved bayesian algorithm are used for learning in the field of image processing and
classifying problem. KNN algorithm is very analysis. 
simple and the accurate method for the image 
 Comparing and analyzing the previous
processing and image analysis by using the
model designed for image processing and
KNN. And SVM is also one of the algorithms image analysis. 
used for Image processing and find plane to 
classify. And it works with the smaller training 2. Research Points
data sets. Decision tree also is used for faster
 Optimizing the previous approaches or
and the very simple classification algorithm
constructing a new approach for Related
which give the higher accuracy of the judgment. problem encountered in application,
The Bayesian algorithm is base and very simple technologies and theories. 
classification method algorithm to classify 
[6][7][8].  Proposing new Machine learning Algorithm
model based on previous Algorithm for
Related Work: There are lot of related work image processing and image analysis. 
have done in the area of image analysis and 
 Use of the mathematical background to
processing. And they used many machine prove the proposed Machine learning
learning algorithm for this task and this work of algorithm. 

119

Authorized licensed use limited to: REVA UNIVERSITY. Downloaded on June 19,2025 at 10:23:29 UTC from IEEE Xplore. Restrictions apply.
 International Conference on Advances in Computing, Communication Control and Networking (ICACCCN2018)

 Learning Techniques to Diagnose Breast

3. Implementation Cancer from Image-Processed Nuclear
Simulate the purposed Model by one of the Features of Fine Needle Aspirates,”
existing tools can be used MATLAB, Python, Cancer Letters, vol. 77, pp. 163-171.
Weka, R system. [3] Wernick, Yang, Brankov, Yourganov
and
Conclusion: In this repot mainly describe the Strother, Machine Learning in Medical
machine learning algorithms in the introduction Imaging, IEEE Signal Processing
parts and the discuss the machine learning Magazine, vol. 27, no. 4, July 2010, pp.
applications. In Machine learning application 25–38.
discuss the image processing and image [4] A.K. Jain, R.P.W. Duin, and J. Mao,
analysis. Steps involve in the Image processing “Statistical Pattern Recognition: A
and image analysis mainly Segmentation and Review,” IEEE Transactions on Pattern
Feature extraction and the classification. These Analysis and Machine Intelligence, Vol.
are the general discussion and I have specify that 22, No. 1, pp. 4-37, January, 2000.
the [5] Supposedly paraphrased from: Samuel,
particular application like bank not Arthur (1959) "Some Studies in Machine
authentication and the healthcare. I have to plan Learning Using the Game of Checkers".
to work in the machine learning algorithm in IBM Journal of Research and
image processing and image analysis for the Development.3(3)
application healthcare like image taken as MRI, .doi:10.1147/rd.33.0210.
X-Ray, Ultra Sound and many other. After [6] Cover, T. M., and Heart, P. E. 1967.
develop the model compare the result with the Nearest neighbor pattern classification.
existing results for different-different applicable IEEE Transactions on Information
machine learning algorithm. Theory, 13, 21-27.
[7] Ripley, R.M. (1998), Neural Networks
References: for Breast Cancer Prognosis, Ph.D.
[1] Machine Learning by Tom Mitchell, Thesis, Department of Engineering
McGraw Hill, 1997.414 pages. ISBN Science, University of Oxford.
0070428077. Cover, T. M., and Heart, P. [8] Aha, D. W., Kibler, D., and Albert, M.
E. 1967. K. 1991. Instance-based learning
[2] Wolberg, W.H., Street, W.N. and algorithms. Machine Learning, 6, 37-66.
Mangasarian, O.L. (1994), “Machine

120

Authorized licensed use limited to: REVA UNIVERSITY. Downloaded on June 19,2025 at 10:23:29 UTC from IEEE Xplore. Restrictions apply.