![International Journal of Trend in Scientific Research and Development (IJTSRD)
Volume 3 Issue 5, August 2019 Available Online: www.ijtsrd.com e-ISSN: 2456 – 6470
@ IJTSRD | Unique Paper ID – IJTSRD27984 | Volume – 3 | Issue – 5 | July - August 2019 Page 2607
Texture Images Classification using
Secant Lines Segments Histogram
Ei Phyu Win, Mie Mie Tin, Pyae Phyo Thu
Faculty of Information Science, Myanmar Institute of Information Technology, Mandalay, Myanmar
How to cite this paper: Ei Phyu Win | Mie
Mie Tin | Pyae Phyo Thu "Texture Images
ClassificationusingSecantLines Segments
Histogram"
Published in
International
Journal of Trend in
Scientific Research
and Development
(ijtsrd), ISSN: 2456-
6470, Volume-3 |
Issue-5, August 2019, pp.2607-2609,
https://doi.org/10.31142/ijtsrd27984
Copyright © 2019 by author(s) and
International Journal ofTrend inScientific
Research and Development Journal. This
is an Open Access article distributed
under the terms of
the Creative
Commons Attribution
License (CC BY 4.0)
(http://creativecommons.org/licenses/by
/4.0)
ABSTRACT
Texture classification is the process to classify different textures from the
given images. The aim of texture classification is to classify the category of a
texture image. To design an effective algorithm for texture classification, it is
essential to find a set of texture features with good discriminatingpower.This
paper presents a texture classification system using secant lines segments
histogram and Euclidean Distance. Secant lines segments histogramis usedto
generate the features from texture images as a histogram. Thesefeaturesoffer
a better discriminating strategy for texture classification. These features are
first used for training and later on for classifying thetextureimages.Euclidean
Distance is used for distinguishing each of the known categories for
classification.
KEYWORDS: texture classification; secant lines segments histogram; Euclidean
distance.
INTRODUCTION
The visual surface characteristic and appearance of something is called the
texture. For image analysis texture characteristicis very important.Texturecan
be presented in either real or artificial data such as wood, clouds, carpet, etc.
Although texture is presented every image, there did not existed a formal
approach of texture analysis. The pattern of carpet, the grain of the wood, the
surface of the clouds, etc. is called texture.
Kavitha & Rao (2011) presented that texture contains
important information about the structural arrangement of
surfaces and their relationship to the surrounding
environment [1].
Textures are one of the important features in computer
vision for many applications. Image texture can be
considered as repeating visual patterns arranged in certain
ways. The patterns themselves are usually very similar,
giving rise to a uniform appearance.Textureis formedby the
relationship of pixel gray levels and their spatial
arrangement within a neighborhood.
Texture classification is applied in various applications such
as object recognition, medical image analysis, remote
sensing, pattern recognition, etc. Pattern is defined by the
shape, colour and texture characteristics of image as feature
arrangements. Pattern recognition is to determine the
objects into each of the known categories.Itis afundamental
component in artificial intelligence and computer vision.
Pattern recognition is used in science,engineering, business,
medicine area and etc.
Major issues in texture analysis may be summarized as
follows:
There are three major issues in the texture analysis:
1. Classify the texture in a variety of texture classes.
2. Generate a description of a texture class.
3. Determine the macro boundaries between two texture
regions of an image that have many distinct texture
areas.
Although the pattern recognition of texture feature
estimation is involved in issue 1, identification of texture is
related to issue2.
Issue 3 deals with the segmentation of texture image. The
goal of texture classification is to classify the category of
image textures.
Fig1. Simple Texture Images
Fig2. Example of Some of Texture Images
IJTSRD27984](https://image.slidesharecdn.com/505textureimagesclassificationusingsecantlinessegmentshistogram-190920081743/85/Texture-Images-Classification-using-Secant-Lines-Segments-Histogram-1-320.jpg)
![International Journal of Trend in Scientific Research and Development (IJTSRD) @ www.ijtsrd.com eISSN: 2456-6470
@ IJTSRD | Unique Paper ID – IJTSRD27984 | Volume – 3 | Issue – 5 | July - August 2019 Page 2608
The proposed system is organized as follows: texture
classification is presented in section 2. the proposed texture
classification system with result of feature extraction
method is proposed in the section 3. Finally, section 4
concludes the discussion.
TEXTURE CLASSIFICATION
The process of determining differenttextures fromthegiven
images is defined as texture classification. The classification
of textures often seems to be not meaningful, however, a
large variety of real world problems that contains the
specific textures of different images have been applied
texture classification.
Texture classification identities a giventexturetoa variety of
texture classes. There are two main classification methods:
supervised and unsupervised classification. Supervised
method needs examples of each texture class as a training
data. By using this training data, a supervised classifier is
trained to learn the characterization of each texture class.
Unsupervised method is automatically found out different
classes from input texture without using prior knowledge.
The rest method called semi-supervised is used only partial
prior knowledge being available.
Fig3. Block Diagram of Texture
Classification
There are two major stages in texture classification: feature
extraction and classification. In feature extraction stage the
most distinguishable texture features are extracted from
each texture. Then, the classifier takes these selected
features as inputs and produces texture class as outputs.
Lepisto, et al. (2003) have presented some applications in
which textured objects of surfaces are contained such as
wood species recognition, rock classification [2], fabric
classification, face detection, and etc. The targetsubjects are
viewed as a specific texture in all of the applications and
hence these systems can be performed successfully using
texture classification techniques.
A different variety of techniques have been proposed for
describing image texture. S. K. Roy, et al. (2018) has
proposed a texture classification system using a robust
descriptor, called fractal weighted local binary pattern
(FWLBP). In this system, a Gaussian Scale Space
representation of texture imageis firstlygenerated and then,
in order to get fractal dimension, differential box counting
algorithm is applied. Finally, to form a normalized feature
vector, the histogram of FWLBP are concatenatedforallLBP
elements [5].
L.Liu, et al. (2012) had proposed a very powerful texture
classification approach for large texture database
applications. In this approach a small set of random features
is extracted from local image patches. In order to do texture
classification, these random features are embedded into a
bag--of-words model. This proposed feature extraction
method is simple because the sparse nature of texture
images is leveraged [6].
Y.Q. Chen (1995) had stated thevarious textureclassification
techniques. These are categorized into five groups. They are
statistical, model-based stochastic, morphology-based,
structural and signal processing methods [3]. Among these
five techniques, statisticaland signalprocessing methods are
widely used because they can be directly used onto any
texture. The rest other methods are not widely used.
The model based stochastic methods are not easily
performed because of the complexity to estimate the
parameters. The morphology-basedmethods are not applied
because it is very newly and its process is very simple. The
structural methods require structured textures that are
naturally rare. So, these methods cannot provide a good
quality of feature textures.
PROPOSED TEXTURE CLASSIFICATION SYSTEM
The proposed texture classification system is implemented
using a new feature extraction method called secant lines
segments histogram. This system classifies texture images
such as floor coverings, wood & fiber and marble & stone.
Fig4. Process of Proposed Texture Classification System
The proposed system mainly involves three mainprocesses:
1. Feature extraction process, where texture features are
extracted from the texture image.
2. Training process, where the texture images are trained
to extract histogram of features for classification.
3. Classification process, where the probability of texture
class is determined using the extracted texture features
as inputs.
FEATURE EXTRACTION
A feature extraction process plays an important role in
content based image classification. Feature Extraction is a
crucial step to the success of classification of texture images
[7]. Therefore, it is very important to extract the most
discriminating information present in the texture image.
characterization of each texture class is yielded by feature
extraction as feature measures.
Identifying and selecting the distinguishable features that
are invariant to translation, rotation and scaling is
important. Therefore, for similar textures, the extracted
feature’s quantities measurements should be very close.
SECANT LINES SEGMENTS HISTOGRAM
The procedures of proposed feature extraction method are
as follow:](https://image.slidesharecdn.com/505textureimagesclassificationusingsecantlinessegmentshistogram-190920081743/85/Texture-Images-Classification-using-Secant-Lines-Segments-Histogram-2-320.jpg)
![International Journal of Trend in Scientific Research and Development (IJTSRD) @ www.ijtsrd.com eISSN: 2456-6470
@ IJTSRD | Unique Paper ID – IJTSRD27984 | Volume – 3 | Issue – 5 | July - August 2019 Page 2609
1. Load texture image.
2. Convert texture image to binary image.
3. Set the secant lines on the texture image as the user
like such as 5 lines, 20 lines, 100 lines, etc.
4. Generate histogram of same distance of texture
5. Image between black and white along thesecantlines as
features of texture image.
In our feature extraction method,thetextureimageisloaded
firstly. Then, this image is converted into binary texture
image to count features in the texture image. Binarization is
an important step for all image processing systems.A binary
image is one in which the pixels can only have two values, 0
and 255 that represent white and black. It can provide
information present in texture clearly for extracting the
features of texture.
Fig5. Experiment result Proposed Feature Extraction
Method
Then, the secant lines are set horizontally on the binary
texture image such as 10 lines, 20 lines, 30 lines,40lines and
50 lines. Generally, the more the secant lines, the better the
classification. The proposed texture classification system is
tested using different numbers ofsecantlines toevaluatethe
best accurate classification.
Finally, same intervals of length of textures are generated
along the secant lines as a histogram. According to our
experiments, different texture images provide the different
histograms. The proposed secant lines segments histogram,
is based on this characteristic in order to get accurate
classification.
TRAINING
In the texture training step, the texture features are
extracted from the texture images foreachoftextureclasses,
using the proposed feature extraction method.Theoutputof
feature extraction, the feature’s histograms of each texture
image, are stored in the feature database, which is used for
texture classification.
CLASSIFICATION
Mahersia & Hamrouni (2008) have presented that a texture
classification system in which the test samples oftexture are
grouped into classes, where each class involves the related
samples of some similarity criterion [4]. The aim of
classification is to determine the most appropriate category
of texture given a set of known categories. The classification
may be performed by distinguishing the probability foreach
of the known categories because perfect classification
approach is not possible.
EUCLIDEAN DISTANCE
Euclidean distance is mostly used for calculating the
similarity metrics for textures. In our system, the histogram
generated in the feature extraction stage needs a matching
metric to measure the similarity between two feature
histograms. The classification module determines how
closely the histogram matches the histogram stored in the
database. The Euclidean distance classifier is found to
perform the best to this system. Euclidean distance is
chosen as a classifier, which calculates the similarity
between the two feature histograms. Euclidean distance
between two points in p-dimensional space is a
geometrically shortest distance on the straight line passing
through both the points. For a distance between two p-
dimensional features x = (x1, x2, xp) and
y1=(y1,y2,…,yp).The Euclidean metric is defined as:
(1)
CONCLUSIONS
In this paper, a texture classification system is proposed
using a new feature extraction method called secant lines
segments histogram. Feature extraction identifies the most
prominent features for classification. This featureextraction
method provides the most discriminating information
present in the texture images as a histogram. The histogram
of each texture image is stored inthefeaturedatabase, which
is used for texture classification. Euclidean Distance is used
to determine the probability for each of the known
categories for classification.
REFERENCE
[1] C. Kavitha, B. Rao, “Image Retrieval BasedOnColorand
Texture Features of the Image Sub-blocks”,
International Journal of Computer Applications,
Volume 15– No.7, February 2011
[2] L. Lepisto, L. Kunttu, J. Autio, A. Visa, “Rock Image
Classification Using Non-Homogenous Textures and
Spectral Imaging”, WSCG, 2003..
[3] Y.Q. Chen, “Novel Techniques for Image Texture
Classification”, PhD Thesis, UniversityofSouthampton,
United Kingdom, 1995.
[4] H.Mahersia, K. Hamrouni, “New Rotation Invariant
Features for Texture Classification”, ICCCE, pp. 687-
690, 2008.
[5] S. K. Roy and B. B. Chaudhuri, N. Bhattacharya, B.
Chanda and D. K. Ghosh, “A Scale Invariant Descriptor
for Texture Classification”, August 2018.
[6] L. Liu and PW. Fieguth, “Texture Classification from
Random Features”, IEEE Trans PatternAnalMachIntll,
March 2012.
[7] E. P. Win, N. Aye, “Recognition of Human Iris Patterns
for Security Systems”, in Proceedings of the 10th
International Conference Inter-UniversityCooperation
Program, ASEAN Community KnowledgeNetworks for
the Economy, Society, Culture, and Environmental
Stability, Mandalay, Myanmar, March 29-30, 2014.](https://image.slidesharecdn.com/505textureimagesclassificationusingsecantlinessegmentshistogram-190920081743/85/Texture-Images-Classification-using-Secant-Lines-Segments-Histogram-3-320.jpg)
Texture Images Classification using Secant Lines Segments Histogram
- 1. International Journal of Trend in Scientific Research and Development (IJTSRD) Volume 3 Issue 5, August 2019 Available Online: www.ijtsrd.com e-ISSN: 2456 – 6470 @ IJTSRD | Unique Paper ID – IJTSRD27984 | Volume – 3 | Issue – 5 | July - August 2019 Page 2607 Texture Images Classification using Secant Lines Segments Histogram Ei Phyu Win, Mie Mie Tin, Pyae Phyo Thu Faculty of Information Science, Myanmar Institute of Information Technology, Mandalay, Myanmar How to cite this paper: Ei Phyu Win | Mie Mie Tin | Pyae Phyo Thu "Texture Images ClassificationusingSecantLines Segments Histogram" Published in International Journal of Trend in Scientific Research and Development (ijtsrd), ISSN: 2456- 6470, Volume-3 | Issue-5, August 2019, pp.2607-2609, https://doi.org/10.31142/ijtsrd27984 Copyright © 2019 by author(s) and International Journal ofTrend inScientific Research and Development Journal. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0) (http://creativecommons.org/licenses/by /4.0) ABSTRACT Texture classification is the process to classify different textures from the given images. The aim of texture classification is to classify the category of a texture image. To design an effective algorithm for texture classification, it is essential to find a set of texture features with good discriminatingpower.This paper presents a texture classification system using secant lines segments histogram and Euclidean Distance. Secant lines segments histogramis usedto generate the features from texture images as a histogram. Thesefeaturesoffer a better discriminating strategy for texture classification. These features are first used for training and later on for classifying thetextureimages.Euclidean Distance is used for distinguishing each of the known categories for classification. KEYWORDS: texture classification; secant lines segments histogram; Euclidean distance. INTRODUCTION The visual surface characteristic and appearance of something is called the texture. For image analysis texture characteristicis very important.Texturecan be presented in either real or artificial data such as wood, clouds, carpet, etc. Although texture is presented every image, there did not existed a formal approach of texture analysis. The pattern of carpet, the grain of the wood, the surface of the clouds, etc. is called texture. Kavitha & Rao (2011) presented that texture contains important information about the structural arrangement of surfaces and their relationship to the surrounding environment [1]. Textures are one of the important features in computer vision for many applications. Image texture can be considered as repeating visual patterns arranged in certain ways. The patterns themselves are usually very similar, giving rise to a uniform appearance.Textureis formedby the relationship of pixel gray levels and their spatial arrangement within a neighborhood. Texture classification is applied in various applications such as object recognition, medical image analysis, remote sensing, pattern recognition, etc. Pattern is defined by the shape, colour and texture characteristics of image as feature arrangements. Pattern recognition is to determine the objects into each of the known categories.Itis afundamental component in artificial intelligence and computer vision. Pattern recognition is used in science,engineering, business, medicine area and etc. Major issues in texture analysis may be summarized as follows: There are three major issues in the texture analysis: 1. Classify the texture in a variety of texture classes. 2. Generate a description of a texture class. 3. Determine the macro boundaries between two texture regions of an image that have many distinct texture areas. Although the pattern recognition of texture feature estimation is involved in issue 1, identification of texture is related to issue2. Issue 3 deals with the segmentation of texture image. The goal of texture classification is to classify the category of image textures. Fig1. Simple Texture Images Fig2. Example of Some of Texture Images IJTSRD27984
- 2. International Journal of Trend in Scientific Research and Development (IJTSRD) @ www.ijtsrd.com eISSN: 2456-6470 @ IJTSRD | Unique Paper ID – IJTSRD27984 | Volume – 3 | Issue – 5 | July - August 2019 Page 2608 The proposed system is organized as follows: texture classification is presented in section 2. the proposed texture classification system with result of feature extraction method is proposed in the section 3. Finally, section 4 concludes the discussion. TEXTURE CLASSIFICATION The process of determining differenttextures fromthegiven images is defined as texture classification. The classification of textures often seems to be not meaningful, however, a large variety of real world problems that contains the specific textures of different images have been applied texture classification. Texture classification identities a giventexturetoa variety of texture classes. There are two main classification methods: supervised and unsupervised classification. Supervised method needs examples of each texture class as a training data. By using this training data, a supervised classifier is trained to learn the characterization of each texture class. Unsupervised method is automatically found out different classes from input texture without using prior knowledge. The rest method called semi-supervised is used only partial prior knowledge being available. Fig3. Block Diagram of Texture Classification There are two major stages in texture classification: feature extraction and classification. In feature extraction stage the most distinguishable texture features are extracted from each texture. Then, the classifier takes these selected features as inputs and produces texture class as outputs. Lepisto, et al. (2003) have presented some applications in which textured objects of surfaces are contained such as wood species recognition, rock classification [2], fabric classification, face detection, and etc. The targetsubjects are viewed as a specific texture in all of the applications and hence these systems can be performed successfully using texture classification techniques. A different variety of techniques have been proposed for describing image texture. S. K. Roy, et al. (2018) has proposed a texture classification system using a robust descriptor, called fractal weighted local binary pattern (FWLBP). In this system, a Gaussian Scale Space representation of texture imageis firstlygenerated and then, in order to get fractal dimension, differential box counting algorithm is applied. Finally, to form a normalized feature vector, the histogram of FWLBP are concatenatedforallLBP elements [5]. L.Liu, et al. (2012) had proposed a very powerful texture classification approach for large texture database applications. In this approach a small set of random features is extracted from local image patches. In order to do texture classification, these random features are embedded into a bag--of-words model. This proposed feature extraction method is simple because the sparse nature of texture images is leveraged [6]. Y.Q. Chen (1995) had stated thevarious textureclassification techniques. These are categorized into five groups. They are statistical, model-based stochastic, morphology-based, structural and signal processing methods [3]. Among these five techniques, statisticaland signalprocessing methods are widely used because they can be directly used onto any texture. The rest other methods are not widely used. The model based stochastic methods are not easily performed because of the complexity to estimate the parameters. The morphology-basedmethods are not applied because it is very newly and its process is very simple. The structural methods require structured textures that are naturally rare. So, these methods cannot provide a good quality of feature textures. PROPOSED TEXTURE CLASSIFICATION SYSTEM The proposed texture classification system is implemented using a new feature extraction method called secant lines segments histogram. This system classifies texture images such as floor coverings, wood & fiber and marble & stone. Fig4. Process of Proposed Texture Classification System The proposed system mainly involves three mainprocesses: 1. Feature extraction process, where texture features are extracted from the texture image. 2. Training process, where the texture images are trained to extract histogram of features for classification. 3. Classification process, where the probability of texture class is determined using the extracted texture features as inputs. FEATURE EXTRACTION A feature extraction process plays an important role in content based image classification. Feature Extraction is a crucial step to the success of classification of texture images [7]. Therefore, it is very important to extract the most discriminating information present in the texture image. characterization of each texture class is yielded by feature extraction as feature measures. Identifying and selecting the distinguishable features that are invariant to translation, rotation and scaling is important. Therefore, for similar textures, the extracted feature’s quantities measurements should be very close. SECANT LINES SEGMENTS HISTOGRAM The procedures of proposed feature extraction method are as follow:
- 3. International Journal of Trend in Scientific Research and Development (IJTSRD) @ www.ijtsrd.com eISSN: 2456-6470 @ IJTSRD | Unique Paper ID – IJTSRD27984 | Volume – 3 | Issue – 5 | July - August 2019 Page 2609 1. Load texture image. 2. Convert texture image to binary image. 3. Set the secant lines on the texture image as the user like such as 5 lines, 20 lines, 100 lines, etc. 4. Generate histogram of same distance of texture 5. Image between black and white along thesecantlines as features of texture image. In our feature extraction method,thetextureimageisloaded firstly. Then, this image is converted into binary texture image to count features in the texture image. Binarization is an important step for all image processing systems.A binary image is one in which the pixels can only have two values, 0 and 255 that represent white and black. It can provide information present in texture clearly for extracting the features of texture. Fig5. Experiment result Proposed Feature Extraction Method Then, the secant lines are set horizontally on the binary texture image such as 10 lines, 20 lines, 30 lines,40lines and 50 lines. Generally, the more the secant lines, the better the classification. The proposed texture classification system is tested using different numbers ofsecantlines toevaluatethe best accurate classification. Finally, same intervals of length of textures are generated along the secant lines as a histogram. According to our experiments, different texture images provide the different histograms. The proposed secant lines segments histogram, is based on this characteristic in order to get accurate classification. TRAINING In the texture training step, the texture features are extracted from the texture images foreachoftextureclasses, using the proposed feature extraction method.Theoutputof feature extraction, the feature’s histograms of each texture image, are stored in the feature database, which is used for texture classification. CLASSIFICATION Mahersia & Hamrouni (2008) have presented that a texture classification system in which the test samples oftexture are grouped into classes, where each class involves the related samples of some similarity criterion [4]. The aim of classification is to determine the most appropriate category of texture given a set of known categories. The classification may be performed by distinguishing the probability foreach of the known categories because perfect classification approach is not possible. EUCLIDEAN DISTANCE Euclidean distance is mostly used for calculating the similarity metrics for textures. In our system, the histogram generated in the feature extraction stage needs a matching metric to measure the similarity between two feature histograms. The classification module determines how closely the histogram matches the histogram stored in the database. The Euclidean distance classifier is found to perform the best to this system. Euclidean distance is chosen as a classifier, which calculates the similarity between the two feature histograms. Euclidean distance between two points in p-dimensional space is a geometrically shortest distance on the straight line passing through both the points. For a distance between two p- dimensional features x = (x1, x2, xp) and y1=(y1,y2,…,yp).The Euclidean metric is defined as: (1) CONCLUSIONS In this paper, a texture classification system is proposed using a new feature extraction method called secant lines segments histogram. Feature extraction identifies the most prominent features for classification. This featureextraction method provides the most discriminating information present in the texture images as a histogram. The histogram of each texture image is stored inthefeaturedatabase, which is used for texture classification. Euclidean Distance is used to determine the probability for each of the known categories for classification. REFERENCE [1] C. Kavitha, B. Rao, “Image Retrieval BasedOnColorand Texture Features of the Image Sub-blocks”, International Journal of Computer Applications, Volume 15– No.7, February 2011 [2] L. Lepisto, L. Kunttu, J. Autio, A. Visa, “Rock Image Classification Using Non-Homogenous Textures and Spectral Imaging”, WSCG, 2003.. [3] Y.Q. Chen, “Novel Techniques for Image Texture Classification”, PhD Thesis, UniversityofSouthampton, United Kingdom, 1995. [4] H.Mahersia, K. Hamrouni, “New Rotation Invariant Features for Texture Classification”, ICCCE, pp. 687- 690, 2008. [5] S. K. Roy and B. B. Chaudhuri, N. Bhattacharya, B. Chanda and D. K. Ghosh, “A Scale Invariant Descriptor for Texture Classification”, August 2018. [6] L. Liu and PW. Fieguth, “Texture Classification from Random Features”, IEEE Trans PatternAnalMachIntll, March 2012. [7] E. P. Win, N. Aye, “Recognition of Human Iris Patterns for Security Systems”, in Proceedings of the 10th International Conference Inter-UniversityCooperation Program, ASEAN Community KnowledgeNetworks for the Economy, Society, Culture, and Environmental Stability, Mandalay, Myanmar, March 29-30, 2014.