Integration of Clustering, Optimization and Partial Differential Equation Method for Improved Image Segmentation

Abstract

Image segmentation generally refers to the process that partitions an image into mutually exclusive regions that cover the image. Among the various image segmentation techniques, traditional image segmentation methods like edge detection, region based, watershed transformation etc. are widely used but have certain drawbacks, which cannot be used for the accurate result. In this paper clustering based techniques is employed on images which results into segmentation of images. The performance of Fuzzy C-means (FCM) integrated with the Particle Swarm optimization (PSO) technique and its variations are analyzed in different application fields. To analyze and grade the performance, computational and time complexity of techniques in different fields several metrics are used namely global consistency error, probabilistic rand index and variation of information are used. This experimental performance analysis shows that FCM along with fractional order Darwinian PSO give better performance in terms of classification accuracy, as compared to other variation of other techniques used. The integrated algorithm tested on images proves to give better results visually as well as objectively. Finally, it is concluded that fractional order Darwinian PSO along with neighborhood Fuzzy C-means and partial differential equation based level set method is an effective image segmentation technique to study the intricate contours provided the time complexity should be as small as possible to make it more real time compatible.