SlideShare ist ein Scribd-Unternehmen logo
1 von 3
Downloaden Sie, um offline zu lesen
International Journal of Electronics Communication and Computer Technology (IJECCT)
Volume 4 Issue 3 (May 2014)
ISSN:2249-7838 IJECCT | www.ijecct.org 656
Overlap Wavelet Transform for Image Segmentation
A.S.Senthilkani, Christo Ananth
Assistant Professor,
Department of ECE
Francis Xavier Engineering College
Tirunelveli, Tamil Nadu, India
Praghash.k, Chakka Raja.M.
PG Scholar, Department of
Communication System
Francis Xavier Engineering College
Tirunelveli, Tamil Nadu, India
Jerrin John, I.Annadurai
PG Scholar,
Department of VLSI
Francis Xavier Engineering College
Tirunelveli, Tamil Nadu, India
Abstract—A new color image segmentation approach based on
OWT is presented in this work. OWT extracts wavelet features
which give a good separation of different patterns. Moreover the
proposed algorithm uses morphological operators for effective
segmentation. From the qualitative and quantitative results, it is
concluded that our proposed method has improved segmentation
quality and it is reliable, fast and can be used with reduced
computational complexity than direct applications of Histogram
Clustering. The main advantage of this method is the use of
single parameter and also very faster. While comparing with five
color spaces, segmentation scheme produces results noticeably
better in RGB color space compared to all other color spaces.
Keywords-- Overlap Wavelet transform; Histogram
Clustering;Edge Based Methods; Region Based Methods.
I. INTRODUCTION
The goal of Image Segmentation is grouping of pixels into
meaningful objects. The quality of Image segmentation
depends upon the image. The traditional K-means algorithm
has the disadvantage of the number of clusters must be applied
as a parameter. Direct applications of 2D Histogram Clustering
fail to segment all the regions. Hence, OWT based 2D
histogram Clustering is used. In the first stage, OWT is used to
introduce redundancy in the filtered images which produce
reliable result in segmentation process. The second stage 2D
histogram is used to find the peaks without any prior
knowledge. The main peak reflects the cluster centroids. The
third stage, Label Concordance algorithm is used to refine the
extracted regions based on local and global information.
II. MATERIALS AND METHODS
A. Overlap Wavelet Transform (OWT)
An adaptive window size to generate wavelet features is
proposed. For n level n, the original input image is split into
4n sub-images. Here the input image is split into 4 sub-
images. Then each image is subjected to stationary wavelet
transform (SWT). The outputs of the same kind of filter
images are interleaved to form four images of same size as the
original image. The Schematic diagram of OWT is shown in
figure.1.The final feature image is constructed using odd rows
and odd columns of the first image, odd rows and even
columns of the second image, even rows and odd columns of
the third image and even rows and even columns of the fourth
image. The maximum energy sub-bands are selected for
clustering. The advantage of this approach is the features from
different levels of resolution are combined since all input
images are of same size.
Figure 1. Schematic diagram of OWT
B. Histogram Clustering algorithm
This algorithm consists of three main stages.
1) Clustering of Color planes
2) Label Concordance Mapping
3) Majority Filtering.
By Clustering of Color Planes, the band subsets are chosen
as RG, RB, and GB pairs.2D histogram is constructed by
summing up all the intensities occurring in the plane. The main
peaks of the histogram give the cluster centroids. Due to
sparseness of the colors in the image, the histogram is noisy.
An exponential filter is applied, to remove the noise and
smoothen the histogram.
Input Image
Split into four Sub-Images
SWT
Interleav
ed output
Interleav
ed output
Interleav
ed output
Interleav
ed output
International Journal of Electronics Communication and Computer Technology (IJECCT)
Volume 4 Issue 3 (May 2014)
ISSN:2249-7838 IJECCT | www.ijecct.org 657
To speed up the cluster determination, the noiseless
histogram is down sampled by a factor 2 i.e. if the smoothed
histogram is of the size 256 X 256,it is reduced to 128 X
128.The down sampling is done by removing the neighborhood
value by the mean value of the two pixel (bin). To extract the
dominant colors, an erosion set is applied which reduces each
bin to its main colors. This method directly extracts the color
cluster centroids. The centroids are labeled and a voronoi
partitioning of the 2D histogram provides the clustering of the
histogram. The clustered histogram is finally up sampled to its
original size by replication of the pixels.
C. Label Concordance algorithm:
To unify the segmentation maps, Label matching algorithm
is used. Each image pair was segmented independently and
labeled. Label Transformation is used to match the labels of
segmentation map I to co-located segments in another map j on
the basis of maximum mutual overlap is defined as follows.
Tij(x) =y (1)
Where x denotes the source label in segmentation map i and
y denotes the target label in segmentation map j and Tij is the
label transformation. This equation shows that the region label
x in map i must be same as label y in map j on the basis of
being co-located and maximally overlapping. Thus six
transformations are formed. TRG,RB ,TRG,GB ,
TRB,RG,TRB,GB,TGB,RG,TGB,RB . Using these definitions of
transformations, bilateral matching cases are checked to find
out regions to be identically labeled. A match is defined as,
Tji(Tij(x))=x (2)
This equation means that in map i and j , there are two
segments that are each others maximally overlapping
counterparts , so that the x-labeled segment in i is mapped into
y in j , while the y-segment in j mapped to x in i. Notice that in
general if Tij(x)=y, then Tji(y)≠x.
III. EXPERIMENTAL WORK
The proposed algorithm is applied on the variety of natural
color images and it is tested on various color spaces. Figure 4
explains the entire process of the proposed algorithm. The
input color image shown in Fig.2.(a) is subjected to 2D
histogram clustering to obtain the clustered image. Initially
color image is splitted into three planes (R, G, B) and 2D
histogram of RG, RB, GB planes are calculated which are
depicted in Fig.2.(b). Then the histogram is smoothed by
Gaussian filter with standard deviation 0.625 and down-
sampled by a factor of 2. Smoothed and down-sampled
versions of the 2D histograms are illustrated in Fig.2.(c).
Morphological erosion is applied on the smoothed histogram
which directly extracts the cluster centroids i.e , dominant
peaks in the 2D histogram which is shown in Fig.2.(d). These
centroids are labeled and watershed transform of the 2D
histogram is performed which provides the clustered histogram
shown in Fig.2.(e). From the clustered histogram the
segmentation map is obtained from simple mapping. Then the
label concordance transformation is performed in order to unify
the segmentation maps which are illustrated in Fig.2.(f).
Unified segmentation maps are fused by using spatial-
chromatic majority filtering which gives the final segmented
result. Fig.2.(g) shows the segmented result.
(a)
RG RB
GB
(b)
RG RB
GB
(c)
International Journal of Electronics Communication and Computer Technology (IJECCT)
Volume 4 Issue 3 (May 2014)
ISSN:2249-7838 IJECCT | www.ijecct.org 658
RG RB
GB
(d)
(e)
Figure 2. Steps involved in proposed method (a) Input image (b)
Histograms with Labelled Centroids (c)Watershed transformed (or)
Clustered histograms (d) Segmentation maps (e) Segmented result
TABLE I. PROPOSED METHOD VS HISTOGRAM CLUSTERING
METHOD
Input
Image
OWT with Histogram
Clustering
Histogram Clustering
Noof
Regions
Color
Error
Eval
Function
Noof
Regions
Color
Error
Eval
Function
Hill 5 0.2964 0.9522 3 0.5938 1.3224
Table I shows the performance of OWT with
Histogram Clustering with respect to traditional Histogram
Clustering approach. The Performance analysis shows that the
Overlap wavelet transform outsmarts the traditional Histogram
Clustering approach in terms of Color Error and Evaluation
function
IV. CONCLUSION
A new color image segmentation approach based on OWT
is presented in this work. OWT extracts wavelet features
which give a good separation of different patterns. Moreover
the proposed algorithm uses morphological operators for
effective segmentation. From the qualitative and quantitative
results, it is concluded that our proposed method has improved
segmentation quality and it is reliable, fast and can be used
with reduced computational complexity than direct
applications of Histogram Clustering. The main advantage of
this method is the use of single parameter and also very faster.
While comparing with five color spaces, segmentation scheme
produces results noticeably better in RGB color space
compared to all other color spaces.
REFERENCES
[1] M.Celenk, “A color clustering technique for image segmentation.
Computer vision Graphics and Image Processing”, IEEE Transactions
on Image Processing, Vol no.52,Issue 4,pp-145-170, 2012
[2] J.Postaire, R.Zhang and C.Leococq-Bottle, “Cluster analysis by binary
morphology”, IEEE Transactions on pattern analysis and machine
Intelligence, Vol no.15 (2): pp-170-180,2011.
[3] Park , I. Yun, and S. Lee, “ Color image segmentation based on 3-d
clustering : morphological approach. Pattern Recognition”,Vol.no.
31(8):pp-1061-1076,2008
[4] Eduardo Akira Yonekura, and Jacques Facon, “Postal Envelope
Segmentation by 2-D Histogram Clustering through Watershed
Transform”, In proceedings of ICDAR 2010.
[5] J. Serra, “Image Analysis and mathematical morphology”, Academic
press, London, 2009.
[6] M P.Soille, “Morphological partitioning of multispectral images”,
Journal of Electronic Imaging,, Vol 18(4):pp- 252-265, 2007

Weitere ähnliche Inhalte

Was ist angesagt?

Development of stereo matching algorithm based on sum of absolute RGB color d...
Development of stereo matching algorithm based on sum of absolute RGB color d...Development of stereo matching algorithm based on sum of absolute RGB color d...
Development of stereo matching algorithm based on sum of absolute RGB color d...
IJECEIAES
 

Was ist angesagt? (20)

Segmentation of medical images using metric topology – a region growing approach
Segmentation of medical images using metric topology – a region growing approachSegmentation of medical images using metric topology – a region growing approach
Segmentation of medical images using metric topology – a region growing approach
 
An Efficient Filtering Technique for Denoising Colour Images
An Efficient Filtering Technique for Denoising Colour Images An Efficient Filtering Technique for Denoising Colour Images
An Efficient Filtering Technique for Denoising Colour Images
 
Study on Contrast Enhancement with the help of Associate Regions Histogram Eq...
Study on Contrast Enhancement with the help of Associate Regions Histogram Eq...Study on Contrast Enhancement with the help of Associate Regions Histogram Eq...
Study on Contrast Enhancement with the help of Associate Regions Histogram Eq...
 
A New Method for Indoor-outdoor Image Classification Using Color Correlated T...
A New Method for Indoor-outdoor Image Classification Using Color Correlated T...A New Method for Indoor-outdoor Image Classification Using Color Correlated T...
A New Method for Indoor-outdoor Image Classification Using Color Correlated T...
 
Multispectral Satellite Color Image Segmentation Using Fuzzy Based Innovative...
Multispectral Satellite Color Image Segmentation Using Fuzzy Based Innovative...Multispectral Satellite Color Image Segmentation Using Fuzzy Based Innovative...
Multispectral Satellite Color Image Segmentation Using Fuzzy Based Innovative...
 
BAYESIAN CLASSIFICATION OF FABRICS USING BINARY CO-OCCURRENCE MATRIX
BAYESIAN CLASSIFICATION OF FABRICS USING BINARY CO-OCCURRENCE MATRIXBAYESIAN CLASSIFICATION OF FABRICS USING BINARY CO-OCCURRENCE MATRIX
BAYESIAN CLASSIFICATION OF FABRICS USING BINARY CO-OCCURRENCE MATRIX
 
Object based image enhancement
Object based image enhancementObject based image enhancement
Object based image enhancement
 
Combining Generative And Discriminative Classifiers For Semantic Automatic Im...
Combining Generative And Discriminative Classifiers For Semantic Automatic Im...Combining Generative And Discriminative Classifiers For Semantic Automatic Im...
Combining Generative And Discriminative Classifiers For Semantic Automatic Im...
 
Feature extraction based retrieval of
Feature extraction based retrieval ofFeature extraction based retrieval of
Feature extraction based retrieval of
 
Image Inpainting
Image InpaintingImage Inpainting
Image Inpainting
 
F43053237
F43053237F43053237
F43053237
 
Av4301248253
Av4301248253Av4301248253
Av4301248253
 
Face Recognition Using Neural Network Based Fourier Gabor Filters & Random Pr...
Face Recognition Using Neural Network Based Fourier Gabor Filters & Random Pr...Face Recognition Using Neural Network Based Fourier Gabor Filters & Random Pr...
Face Recognition Using Neural Network Based Fourier Gabor Filters & Random Pr...
 
Interpolation Technique using Non Linear Partial Differential Equation with E...
Interpolation Technique using Non Linear Partial Differential Equation with E...Interpolation Technique using Non Linear Partial Differential Equation with E...
Interpolation Technique using Non Linear Partial Differential Equation with E...
 
Content Based Image Retrieval Approach Based on Top-Hat Transform And Modifie...
Content Based Image Retrieval Approach Based on Top-Hat Transform And Modifie...Content Based Image Retrieval Approach Based on Top-Hat Transform And Modifie...
Content Based Image Retrieval Approach Based on Top-Hat Transform And Modifie...
 
Development of stereo matching algorithm based on sum of absolute RGB color d...
Development of stereo matching algorithm based on sum of absolute RGB color d...Development of stereo matching algorithm based on sum of absolute RGB color d...
Development of stereo matching algorithm based on sum of absolute RGB color d...
 
A version of watershed algorithm for color image segmentation
A version of watershed algorithm for color image segmentationA version of watershed algorithm for color image segmentation
A version of watershed algorithm for color image segmentation
 
TYBSC (CS) SEM 6- DIGITAL IMAGE PROCESSING
TYBSC (CS) SEM 6- DIGITAL IMAGE PROCESSINGTYBSC (CS) SEM 6- DIGITAL IMAGE PROCESSING
TYBSC (CS) SEM 6- DIGITAL IMAGE PROCESSING
 
OBIA on Coastal Landform Based on Structure Tensor
OBIA on Coastal Landform Based on Structure Tensor OBIA on Coastal Landform Based on Structure Tensor
OBIA on Coastal Landform Based on Structure Tensor
 
Region filling and object removal by exemplar based image inpainting
Region filling and object removal by exemplar based image inpaintingRegion filling and object removal by exemplar based image inpainting
Region filling and object removal by exemplar based image inpainting
 

Andere mochten auch (7)

8 minutos
8 minutos8 minutos
8 minutos
 
Tijgerpps ws
Tijgerpps wsTijgerpps ws
Tijgerpps ws
 
2 tazas de café
2 tazas de café2 tazas de café
2 tazas de café
 
Ejercicios del seminario 10
Ejercicios del seminario 10Ejercicios del seminario 10
Ejercicios del seminario 10
 
Telecommunication Business Process - eTOM Flows
Telecommunication Business Process - eTOM FlowsTelecommunication Business Process - eTOM Flows
Telecommunication Business Process - eTOM Flows
 
Test
TestTest
Test
 
Malla de seminarios 06 julio al 11 de julio
Malla de seminarios 06 julio al 11 de julioMalla de seminarios 06 julio al 11 de julio
Malla de seminarios 06 julio al 11 de julio
 

Ähnlich wie ijecct

Research Paper v2.0
Research Paper v2.0Research Paper v2.0
Research Paper v2.0
Kapil Tiwari
 
Brain tumor segmentation using asymmetry based histogram thresholding and k m...
Brain tumor segmentation using asymmetry based histogram thresholding and k m...Brain tumor segmentation using asymmetry based histogram thresholding and k m...
Brain tumor segmentation using asymmetry based histogram thresholding and k m...
eSAT Publishing House
 
Icamme managed brightness and contrast enhancement using adapted histogram eq...
Icamme managed brightness and contrast enhancement using adapted histogram eq...Icamme managed brightness and contrast enhancement using adapted histogram eq...
Icamme managed brightness and contrast enhancement using adapted histogram eq...
Jagan Rampalli
 

Ähnlich wie ijecct (20)

Fast Segmentation of Sub-cellular Organelles
Fast Segmentation of Sub-cellular OrganellesFast Segmentation of Sub-cellular Organelles
Fast Segmentation of Sub-cellular Organelles
 
International Journal of Engineering Inventions (IJEI)
International Journal of Engineering Inventions (IJEI)International Journal of Engineering Inventions (IJEI)
International Journal of Engineering Inventions (IJEI)
 
Research Paper v2.0
Research Paper v2.0Research Paper v2.0
Research Paper v2.0
 
A New Approach for Segmentation of Fused Images using Cluster based Thresholding
A New Approach for Segmentation of Fused Images using Cluster based ThresholdingA New Approach for Segmentation of Fused Images using Cluster based Thresholding
A New Approach for Segmentation of Fused Images using Cluster based Thresholding
 
Supervised Blood Vessel Segmentation in Retinal Images Using Gray level and M...
Supervised Blood Vessel Segmentation in Retinal Images Using Gray level and M...Supervised Blood Vessel Segmentation in Retinal Images Using Gray level and M...
Supervised Blood Vessel Segmentation in Retinal Images Using Gray level and M...
 
Developing 3D Viewing Model from 2D Stereo Pair with its Occlusion Ratio
Developing 3D Viewing Model from 2D Stereo Pair with its Occlusion RatioDeveloping 3D Viewing Model from 2D Stereo Pair with its Occlusion Ratio
Developing 3D Viewing Model from 2D Stereo Pair with its Occlusion Ratio
 
Hybrid Technique for Copy-Move Forgery Detection Using L*A*B* Color Space
Hybrid Technique for Copy-Move Forgery Detection Using L*A*B* Color Space Hybrid Technique for Copy-Move Forgery Detection Using L*A*B* Color Space
Hybrid Technique for Copy-Move Forgery Detection Using L*A*B* Color Space
 
Brain tumor segmentation using asymmetry based histogram thresholding and k m...
Brain tumor segmentation using asymmetry based histogram thresholding and k m...Brain tumor segmentation using asymmetry based histogram thresholding and k m...
Brain tumor segmentation using asymmetry based histogram thresholding and k m...
 
Icamme managed brightness and contrast enhancement using adapted histogram eq...
Icamme managed brightness and contrast enhancement using adapted histogram eq...Icamme managed brightness and contrast enhancement using adapted histogram eq...
Icamme managed brightness and contrast enhancement using adapted histogram eq...
 
Two-dimensional Block of Spatial Convolution Algorithm and Simulation
Two-dimensional Block of Spatial Convolution Algorithm and SimulationTwo-dimensional Block of Spatial Convolution Algorithm and Simulation
Two-dimensional Block of Spatial Convolution Algorithm and Simulation
 
V.KARTHIKEYAN PUBLISHED ARTICLE
V.KARTHIKEYAN PUBLISHED ARTICLEV.KARTHIKEYAN PUBLISHED ARTICLE
V.KARTHIKEYAN PUBLISHED ARTICLE
 
N42018588
N42018588N42018588
N42018588
 
A comparative study on content based image retrieval methods
A comparative study on content based image retrieval methodsA comparative study on content based image retrieval methods
A comparative study on content based image retrieval methods
 
Retinal blood vessel extraction and optical disc removal
Retinal blood vessel extraction and optical disc removalRetinal blood vessel extraction and optical disc removal
Retinal blood vessel extraction and optical disc removal
 
Imagethresholding
ImagethresholdingImagethresholding
Imagethresholding
 
Multi Resolution features of Content Based Image Retrieval
Multi Resolution features of Content Based Image RetrievalMulti Resolution features of Content Based Image Retrieval
Multi Resolution features of Content Based Image Retrieval
 
B070306010
B070306010B070306010
B070306010
 
Ijetr011917
Ijetr011917Ijetr011917
Ijetr011917
 
I010135760
I010135760I010135760
I010135760
 
Content Based Image Retrieval Using 2-D Discrete Wavelet Transform
Content Based Image Retrieval Using 2-D Discrete Wavelet TransformContent Based Image Retrieval Using 2-D Discrete Wavelet Transform
Content Based Image Retrieval Using 2-D Discrete Wavelet Transform
 

ijecct

  • 1. International Journal of Electronics Communication and Computer Technology (IJECCT) Volume 4 Issue 3 (May 2014) ISSN:2249-7838 IJECCT | www.ijecct.org 656 Overlap Wavelet Transform for Image Segmentation A.S.Senthilkani, Christo Ananth Assistant Professor, Department of ECE Francis Xavier Engineering College Tirunelveli, Tamil Nadu, India Praghash.k, Chakka Raja.M. PG Scholar, Department of Communication System Francis Xavier Engineering College Tirunelveli, Tamil Nadu, India Jerrin John, I.Annadurai PG Scholar, Department of VLSI Francis Xavier Engineering College Tirunelveli, Tamil Nadu, India Abstract—A new color image segmentation approach based on OWT is presented in this work. OWT extracts wavelet features which give a good separation of different patterns. Moreover the proposed algorithm uses morphological operators for effective segmentation. From the qualitative and quantitative results, it is concluded that our proposed method has improved segmentation quality and it is reliable, fast and can be used with reduced computational complexity than direct applications of Histogram Clustering. The main advantage of this method is the use of single parameter and also very faster. While comparing with five color spaces, segmentation scheme produces results noticeably better in RGB color space compared to all other color spaces. Keywords-- Overlap Wavelet transform; Histogram Clustering;Edge Based Methods; Region Based Methods. I. INTRODUCTION The goal of Image Segmentation is grouping of pixels into meaningful objects. The quality of Image segmentation depends upon the image. The traditional K-means algorithm has the disadvantage of the number of clusters must be applied as a parameter. Direct applications of 2D Histogram Clustering fail to segment all the regions. Hence, OWT based 2D histogram Clustering is used. In the first stage, OWT is used to introduce redundancy in the filtered images which produce reliable result in segmentation process. The second stage 2D histogram is used to find the peaks without any prior knowledge. The main peak reflects the cluster centroids. The third stage, Label Concordance algorithm is used to refine the extracted regions based on local and global information. II. MATERIALS AND METHODS A. Overlap Wavelet Transform (OWT) An adaptive window size to generate wavelet features is proposed. For n level n, the original input image is split into 4n sub-images. Here the input image is split into 4 sub- images. Then each image is subjected to stationary wavelet transform (SWT). The outputs of the same kind of filter images are interleaved to form four images of same size as the original image. The Schematic diagram of OWT is shown in figure.1.The final feature image is constructed using odd rows and odd columns of the first image, odd rows and even columns of the second image, even rows and odd columns of the third image and even rows and even columns of the fourth image. The maximum energy sub-bands are selected for clustering. The advantage of this approach is the features from different levels of resolution are combined since all input images are of same size. Figure 1. Schematic diagram of OWT B. Histogram Clustering algorithm This algorithm consists of three main stages. 1) Clustering of Color planes 2) Label Concordance Mapping 3) Majority Filtering. By Clustering of Color Planes, the band subsets are chosen as RG, RB, and GB pairs.2D histogram is constructed by summing up all the intensities occurring in the plane. The main peaks of the histogram give the cluster centroids. Due to sparseness of the colors in the image, the histogram is noisy. An exponential filter is applied, to remove the noise and smoothen the histogram. Input Image Split into four Sub-Images SWT Interleav ed output Interleav ed output Interleav ed output Interleav ed output
  • 2. International Journal of Electronics Communication and Computer Technology (IJECCT) Volume 4 Issue 3 (May 2014) ISSN:2249-7838 IJECCT | www.ijecct.org 657 To speed up the cluster determination, the noiseless histogram is down sampled by a factor 2 i.e. if the smoothed histogram is of the size 256 X 256,it is reduced to 128 X 128.The down sampling is done by removing the neighborhood value by the mean value of the two pixel (bin). To extract the dominant colors, an erosion set is applied which reduces each bin to its main colors. This method directly extracts the color cluster centroids. The centroids are labeled and a voronoi partitioning of the 2D histogram provides the clustering of the histogram. The clustered histogram is finally up sampled to its original size by replication of the pixels. C. Label Concordance algorithm: To unify the segmentation maps, Label matching algorithm is used. Each image pair was segmented independently and labeled. Label Transformation is used to match the labels of segmentation map I to co-located segments in another map j on the basis of maximum mutual overlap is defined as follows. Tij(x) =y (1) Where x denotes the source label in segmentation map i and y denotes the target label in segmentation map j and Tij is the label transformation. This equation shows that the region label x in map i must be same as label y in map j on the basis of being co-located and maximally overlapping. Thus six transformations are formed. TRG,RB ,TRG,GB , TRB,RG,TRB,GB,TGB,RG,TGB,RB . Using these definitions of transformations, bilateral matching cases are checked to find out regions to be identically labeled. A match is defined as, Tji(Tij(x))=x (2) This equation means that in map i and j , there are two segments that are each others maximally overlapping counterparts , so that the x-labeled segment in i is mapped into y in j , while the y-segment in j mapped to x in i. Notice that in general if Tij(x)=y, then Tji(y)≠x. III. EXPERIMENTAL WORK The proposed algorithm is applied on the variety of natural color images and it is tested on various color spaces. Figure 4 explains the entire process of the proposed algorithm. The input color image shown in Fig.2.(a) is subjected to 2D histogram clustering to obtain the clustered image. Initially color image is splitted into three planes (R, G, B) and 2D histogram of RG, RB, GB planes are calculated which are depicted in Fig.2.(b). Then the histogram is smoothed by Gaussian filter with standard deviation 0.625 and down- sampled by a factor of 2. Smoothed and down-sampled versions of the 2D histograms are illustrated in Fig.2.(c). Morphological erosion is applied on the smoothed histogram which directly extracts the cluster centroids i.e , dominant peaks in the 2D histogram which is shown in Fig.2.(d). These centroids are labeled and watershed transform of the 2D histogram is performed which provides the clustered histogram shown in Fig.2.(e). From the clustered histogram the segmentation map is obtained from simple mapping. Then the label concordance transformation is performed in order to unify the segmentation maps which are illustrated in Fig.2.(f). Unified segmentation maps are fused by using spatial- chromatic majority filtering which gives the final segmented result. Fig.2.(g) shows the segmented result. (a) RG RB GB (b) RG RB GB (c)
  • 3. International Journal of Electronics Communication and Computer Technology (IJECCT) Volume 4 Issue 3 (May 2014) ISSN:2249-7838 IJECCT | www.ijecct.org 658 RG RB GB (d) (e) Figure 2. Steps involved in proposed method (a) Input image (b) Histograms with Labelled Centroids (c)Watershed transformed (or) Clustered histograms (d) Segmentation maps (e) Segmented result TABLE I. PROPOSED METHOD VS HISTOGRAM CLUSTERING METHOD Input Image OWT with Histogram Clustering Histogram Clustering Noof Regions Color Error Eval Function Noof Regions Color Error Eval Function Hill 5 0.2964 0.9522 3 0.5938 1.3224 Table I shows the performance of OWT with Histogram Clustering with respect to traditional Histogram Clustering approach. The Performance analysis shows that the Overlap wavelet transform outsmarts the traditional Histogram Clustering approach in terms of Color Error and Evaluation function IV. CONCLUSION A new color image segmentation approach based on OWT is presented in this work. OWT extracts wavelet features which give a good separation of different patterns. Moreover the proposed algorithm uses morphological operators for effective segmentation. From the qualitative and quantitative results, it is concluded that our proposed method has improved segmentation quality and it is reliable, fast and can be used with reduced computational complexity than direct applications of Histogram Clustering. The main advantage of this method is the use of single parameter and also very faster. While comparing with five color spaces, segmentation scheme produces results noticeably better in RGB color space compared to all other color spaces. REFERENCES [1] M.Celenk, “A color clustering technique for image segmentation. Computer vision Graphics and Image Processing”, IEEE Transactions on Image Processing, Vol no.52,Issue 4,pp-145-170, 2012 [2] J.Postaire, R.Zhang and C.Leococq-Bottle, “Cluster analysis by binary morphology”, IEEE Transactions on pattern analysis and machine Intelligence, Vol no.15 (2): pp-170-180,2011. [3] Park , I. Yun, and S. Lee, “ Color image segmentation based on 3-d clustering : morphological approach. Pattern Recognition”,Vol.no. 31(8):pp-1061-1076,2008 [4] Eduardo Akira Yonekura, and Jacques Facon, “Postal Envelope Segmentation by 2-D Histogram Clustering through Watershed Transform”, In proceedings of ICDAR 2010. [5] J. Serra, “Image Analysis and mathematical morphology”, Academic press, London, 2009. [6] M P.Soille, “Morphological partitioning of multispectral images”, Journal of Electronic Imaging,, Vol 18(4):pp- 252-265, 2007