Pollination based optimization for color image segmentation
Published on: Mar 4, 2016
Transcripts - Pollination based optimization for color image segmentation
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976 – 6367(Print),ISSN 0976 – 6375(Online) Volume 3, Issue 2, July- September (2012), © IAEME407POLLINATION BASED OPTIMIZATION FOR COLOR IMAGESEGMENTATIONGaganpreet Kaur1, Dr. Dheerendra Singh21Assistant Professor, Department of CSE, Sri Guru Granth Sahib World University,Fatehgarh Sahib, India2Professor & Head, Department of Computer Science & Engineering, SUSCET,Tangori, Mohali, IndiaE-mail: email@example.com, firstname.lastname@example.orgABSTRACTColor image segmentation is a process of partitioning an image into disjoint regions, i.e. into subsets ofconnected pixels which share similar color properties. Region extraction in color images is a difficultprocess. I have proposed a new optimization method Pollination Based Optimization (PBO) to select bestoptimal clusters in color images. The methodology consisted of four steps: color space conversion,generation of candidate color cluster centers using Fuzzy K Means, pollination based optimizationmethod to select optimum color cluster centers, image segmentation. Pollination in flowers is used forselecting optimal clusters in colored image. The optimization method worked well on images used. Thetotal elapsed time used to compute segmentation also reduced considerably.Keywords: Segmentation, Clustering, Pollination Based Optimization (PBO)INTRODUCTIONSegmentation involves partitioning an image into a set of homogeneous and meaningful regions,such that the pixels in each partitioned region possess an identical set of properties . Imagesegmentation is one of the most challenging tasks in image processing and is a very important pre-processing step in the problems in the area of image analysis, computer vision, and pattern recognition. In many applications, the quality of final object classification and scene interpretation depends largelyon the quality of the segmented output . In segmentation, an image is partitioned into different non-overlapping homogeneous regions, where the homogeneity of a region may be composed based ondifferent criteria such as gray level, color or texture. . Image segmentation is a complex and hard taskin color images, but also it is the one of the important and curial problems in color image analysis.Because high performing segmentation algorithms lead effective image recognition and retrieval systems. In this paper, PBO approach is used for color image segmentation by using clustering is used. Thepaper is organized as follows. In section 2, pollination based Optimization (PBO) is discussed. In section3, the design and implementation of PBO is discussed. Experimental results on images are presented insection 4. Finally, in Section 5, some conclusions and directions for future work are discussed.INTERNATIONAL JOURNAL OF COMPUTER ENGINEERING &TECHNOLOGY (IJCET)ISSN 0976 – 6367(Print)ISSN 0976 – 6375(Online)Volume 3, Issue 2, July- September (2012), pp. 407-414© IAEME: www.iaeme.com/ijcet.htmlJournal Impact Factor (2012): 3.9580 (Calculated by GISI)www.jifactor.comIJCET© I A E M E
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976 – 6367(Print),ISSN 0976 – 6375(Online) Volume 3, Issue 2, July- September (2012), © IAEME408A. Image SegmentationThe segmentation procedures analyze the colors of pixels in order to distinguish the different objectswhich constitute the scene observed by a color sensor or camera. It is a process of partitioning an imageinto disjoint regions, i.e. into subsets of connected pixels which share similar color properties .Segmentation schemes can be divided into two main approaches with respect to the used predicate .The first one assumes that adjacent regions representing different objects present local discontinuities ofcolors at their boundaries. The second one assumes that a region is a subset of connected pixels whichshare similar color properties. The methods associated with this assumption are called region constructionmethods and look for subsets of connected pixels whose colors are homogeneous . These techniquescan be categorized into two main classes, whether the distribution of the pixel colors is analyzed either inthe image plane or in the color space .B. Region construction based on a color space analysisThe color of each pixel can be represented in a color space, it is also possible to analyze the distributionof pixel colors rather than examining the image plane. In the (R,G,B) color space, a color point is definedby the color component levels of the corresponding pixel, namely red (R), green (G) and blue (B). It isgenerally assumed that homogeneous regions in the image plane give rise to clusters of color points in thecolor space, each cluster corresponding to a class of pixels which share similar color properties . Theclasses of pixels are constructed by means of a cluster identification scheme which is performed either byan analysis of the color histogram or by a cluster analysis procedure. When the classes are constructed,the pixels are assigned to one of them by means of a decision rule and are mapped back to the originalimage plane to produce the segmentation. The regions of the segmented image are composed ofconnected pixels which are assigned to the same classes. When the distribution of color points is analyzedin the color space, the procedures generally lead to a noisy segmentation with small regions scatteredthrough the image. Usually, a spatial-based post-processing is performed to reconstruct the actual regionsin the image [6, 7].C. Clustering-based segmentationClustering is the process of identifying natural groupings or clusters, within multidimensional data,based on some similarity measure (e.g. Euclidean distance) , . Clustering algorithms are used inmany applications, such as data mining, compression, image segmentation, machine learning etc. Acluster is usually identified by a cluster center (or centroid) . Data clustering is a difficult problem asthe clusters in data may have different shapes and sizes . Most clustering algorithms are based on twopopular techniques known as hierarchical and partitional clustering . In hierarchical clustering, theoutput is "a tree showing a sequence of clustering with each clustering being a partition of the data set”.Partitional clustering aims to optimize cluster centers, as well as the number of clusters . Mostclustering algorithms require the number of clusters to be specified in advance . Finding the"optimum" number of clusters in a data set is usually a challenge since it requires a priori knowledge,and/or ground truth about the data, which is not always available. The problem of finding the optimumnumber of clusters in a data set has been the subject of several research efforts , however, despite theamount of research in this area, the outcome is still unsatisfactory.II. POLLINATION BASED OPTIMIZATIONOptimization is a natural process embedded in the living beings .Pollination is a process of transferof pollen from male parts of flower called anther to the female part called stigma of a flower. Someflowers will develop seeds as a result of self-pollination, when pollen and pistil are from the same plant,often (but not always) from the same flower. Other plants require cross-pollination: pollen and pistil mustbe from different plants. Plants benefit from pollinators because the movement of pollen allows them to
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976 – 6367(Print),ISSN 0976 – 6375(Online) Volume 3, Issue 2, July- September (2012), © IAEME409reproduce by setting seeds. However, pollinators dont know or care that the plant benefits. They pollinateto get nectar and/or pollen from flowers to meet their energy requirements and to produce offspring. Inthe economy of nature, the pollinators provide an important service to flowering plants, while the plantspay with food for the pollinators and their offspring. The floral display, fragrance and nectar lurepollinators and leads to pollination. Some species of plants optimize their nectar, display and fragranceproducing resources. If pollination process is proceeding smoothly the plants spend average resources. Ifpollination process is above normal the plants reduce expenditure on resources for producing nectar,floral display and fragrance in the flowers. If the pollination success goes below normal, plants increasethe resource expenditure such that more floral display, fragrance and nectar to attract pollinator. As morepollinators and their number of visits increase the pollination success rate increases .I have used the model suggested by Thakar et al. . The model suggests that the reproductivesuccess for every plant can be modelled by the following expression.R= ሺൈሺןାൈሻሻ ഀןశಲൈವሻൈேುುାேುെ ܥሺܰ ܦሻ (1)Whereߙ =variable denoting average display at a given average nectar content. (a = optimum D = 1.2)A = Average Investment in Nectar Content of a species (A= optimum N = 0.9. its range is 0.8 -1.4)D = individual investment in display (0 - 1.2 typical 1.2)N = individual investment in nectar (0.8 - 1.5 typical 1.2 at A = 0.9)P = parameter related to pollinators learning efficiency P = m x a+c here m and c areConstants. (Range 0.1 – 25, typical value.2)C =proportionality constant relating investment to reproductive cost (1)A. PBO ALGORITHMInitialize a=1.2, A=0.9, D=1.2, N41.9, P=2,number of_plants = 6;number of weeks =6;number_of_seasons =20 (number of iterations)pollination_weekly_goal = [0.10 0.25 0.50 0.75 0.90 1.00]Randomly generate Investment Vector (IV)*For season = 1 to number o[seasons (iterations)For week = 1: number of weeksFor k = 1: number of_plantsEvaluate R using equation 1Based upon R, update IVEvaluate Error = Goal - RBased upon error update N, D, AEndExit, if Error acceptableEndEndIII. DESIGN AND IMPLEMENTATION OF POLLINATION BASED OPTIMIZATION FORCOLOR IMAGE SEGMENTATIONFrom the literature, it has been found that there is a need to develop a new optimization for selectingoptimum clusters in colored image for image segmentation. The proposed model focuses on followingsteps:a) New optimization method based on pollination in flowers to select best clusters.
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976 – 6367(Print),ISSN 0976 – 6375(Online) Volume 3, Issue 2, July- September (2012), © IAEME410b) The total time used to compute segmentation is reduced.This is practically implemented using MATLAB 7.11.0 environment.A. System Level DesignThe work is implemented on three images-1. Hestain.png2. Fabric.png3. Lion.jpg.Figure 1 shows the original test images used in the work.B. Algorithm Level DesignTo segment image into regions, our method operates some successive tasks step by step. First, segmentthe image into clusters using fuzzy K means clustering method .Then, pollination based optimizationalgorithm is run to select optimum cluster points over candidate cluster center points set. Finally, pixelsare classified according to their closest cluster center point, and image is segmented into homogeneousregions.The Image Segmentation using PBO algorithm can be described with the following algorithm.Step1) Take an image and convert it into Lab image.Step2) Segment the image into clusters using Fuzzy K Means. Step3) Select the optimum cluster pointscall PBO algorithm discussed in 2.1. This loop can be terminated after a predefined number ofgenerations or after an acceptable problem solution has been found.Step 4) Segmented image is obtained.Fig.1 Original Test Images (a) “Hestain.png” (b) “Fabric.png” (c) “Lion.jpg”IV. RESULTSThe implementation was done on original images of hestain.png, fabric.png and lion.jpg for colorimage segmentation using PBO approach shown in Figure 1. Clusters of the objects obtained from theoriginal images are shown in Figure 2, Figure 4 and Figure 6. Finally segmented homogeneous regions ofthe original image are shown in Figure 3, Figure 5 and Figure 7.The implementation was done in Matlab7.11.0. The time elapsed by new algorithm to identify the clusters and further segment the regions ofinterest is in seconds as shown in Table1.
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976 – 6367(Print),ISSN 0976 – 6375(Online) Volume 3, Issue 2, July- September (2012), © IAEME411Fig. 2 (a) Objects in Clusters 1 (b) Objects in Clusters 2 (c) Objects in Clusters 3 for image 1 (a)Fig. 3 (a) Segmented Red Colored Region obtained from Image 1(a) (b) Segmented Purple ColoredRegion (c) Segmented Magenta Colored RegionFig. 4 (a) Objects in Clusters 1 (b) Objects in Clusters 2 (c) Objects in Clusters 3 for image 1 (b)Fig. 5 (a) Segmented Red Colored Region obtained from Image 1(b) (b) Segmented Purple ColoredRegion (c) Segmented Magenta Colored Region
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976 – 6367(Print),ISSN 0976 – 6375(Online) Volume 3, Issue 2, July- September (2012), © IAEME412Fig. 6 (a) Objects in Clusters 1 (b) Objects in Clusters 2 (c) Objects in Clusters 3 for image 1 (c)Fig. 7 (a) Segmented Red Colored Region obtained from Image 1(c) (b) Segmented Purple ColoredRegion (c) Segmented Magenta Colored Region (d) Segmented Green Colored Region (e) SegmentedYellow Colored RegionTABLE I: TIME TAKEN BY PBO BASED COLOR IMAGE SEGMENTATIONOptimization Techniqueused/ Time taken tosegment an image( in seconds)POLLINATION BASEDOPTIMIZATION(PBO)Hestain.png 1.34Fabric.png 1.59Lion.jpg 3.66
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976 – 6367(Print),ISSN 0976 – 6375(Online) Volume 3, Issue 2, July- September (2012), © IAEME413V. CONCLUSIONThis paper derives new optimization technique based on pollination of plants to segment the colorimages. This new optimization algorithm has been implemented in Matlab for extraction the optimalclusters from the colored images. The new algorithm worked accurately on the colored images and thetime required to segment the images into distinct homogeneous regions was considerably reduced ascompared to other optimization algorithms.REFERENCES K. Bhoyar, O. Kakde, “Color Image Segmentation Based On Jnd Color Histogram”, InternationalJournal of Image Processing (IJIP) Vol. 3, Issue 6, pp. 283-292. H. D. Cheng, X. H. Jiang, Y. Sun, J. Wang, “Color image segmentation: advances and prospects”,Pattern Recognition, pp. 2259–2281, 2001. A.W. Liew, H. Yan, N.F. Law, “Image segmentation based on adaptive cluster prototypeestimation”, IEEE Trans. Fuzzy Syst. Vol. no.13 (4), pp. 444–453, 2005. D. Aydin, A. Ugur, “Extraction of flower regions in color images using ant colony optimization”,Procedia Computer Science, pp. 530–536, 2011. L. Busin et. al, “Color spaces and image segmentation”, PhD Thesis, 2007. H. Cheng, X. Jiang, J. Wang, “Color image segmentation based on homogram thresholding andregion merging”, Pattern Recognition, pp. 373–393,2002. D. Nikolaev, P. Nikolayev, “Linear color segmentation and its implementation”, ComputerVision and Image Understanding, pp. 115–139, 2004. A.K. Jain, M.N. Murty, P.J. Flynn, “Data Clustering: A Review”, ACM Computing Surveys, Vol.31(3), pp. 264-323, 1999. A.K. Jain, R. Duin, J. Mao, “Statistical Pattern Recognition: A Review”, IEEE Transactions onPattern Analysis and Machine Intellgence, Vol. 22 (1), pp.4-37, 2000. Mahamed G.H. et. al., “Dynamic Clustering using Particle Swarm Optimization with Applicationin Unsupervised Image Classification”, World Academy of Science, Engineering andTechnology, 2005. G. Hamerly, C. Elkan, Learning the K in K-means, 7th Annual Conference on Neural InformationProcessing Systems, 2003. S. Kumar & Singh Ajay, “ Pollination based optimization”, presented at 6th International MultiConference on Intelligent Systems, Sustainable, New and Renewable Energy Technology andNanotechnology (IISN2012), March 16-18,2012,pp. 269-273. J.D.Thakar, K. Kunte, A.K. Chauhan, A.V. Watve, M.G. Watve, “Nectarless flowers: ecologicalcorrelates and evolutionary stability”,Oecologia. 136, pp. 565-570, 2003. G. J. Klinker, S. A. Shafer, T. Kanade,, “A physical approach to color image understanding”,International Journal of Computer Vision, 7–30, 1990. S. Bansal, D. Aggarwal, “Color Image Segmentation using CIELab Color Space using AntColony Optimization”, International Journal of Computer Applications, pp. 28-34, 2011. M.E. Lee1 et. al., “Segmentation of Brain MR Images using an Ant Colony OptimizationAlgorithm”, IEEE, 2009. P. Thakare et. al., “A Study of Image Segmentation and Edge Detection Techniques”,International Journal on Computer Science and Engineering (IJCSE), Vol. 3 No. 2, Feb 2011. A.V. Baterina et al., “Image Edge Detection Using Ant Colony Optimization”, InternationalJournal Of Circuits, Systems And Signal Processing, Volume 4, Issue 4, 2010. A. Z. Chitade et. al., “Colour Based Image Segmentation Using K-Means Clustering”,International Journal of Engineering Science and Technology Vol. 2(10), pp. 5319-5325, 2010.
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976 – 6367(Print),ISSN 0976 – 6375(Online) Volume 3, Issue 2, July- September (2012), © IAEME414 P. Singh et. al., “Satellite Image Classification by Hybridization of FPAB Algorithm andBacterial Chemotaxis”, International Journal of Computer Technology and ElectronicsEngineering (IJCTEE), Volume 1, Issue 3, pp. 21-27, 2011. C.Y. Lee, E.K. Antonsson, “Dynamic Partitional Clustering Using Evolution Strategies” , In TheThird Asia-Pacific Conference on Simulated Evolution and Learning, 2000. H. Frigui and R. Krishnapuram, “A Robust Competitive Clustering Algorithm with Applicationsin Computer Vision”, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol.21(5), pp. 450-465, 1999. Y. Leung, J. Zhang, Z. Xu, “Clustering by Space-Space Filtering”, IEEE Transactions on PatternAnalysis and Machine Intelligence, Vol. 22(12), pp. 1396-1410, 2000. C. Rosenberger and K. Chehdi, “Unsupervised Clustering Method with Optimal Estimation of theNumber of Clusters: Application to Image Segmentation”, International Conference on PatternRecognition (ICPR00), Vol. 1, pp.1656-1659, 2000.