An integrated color image segmentation with multi-class SVM followed by SRFCM

Abstract

In existing the segmentation of a color image is mostly depends on the features color, texture or on both color and texture but proposed method for color image segmentation is based on both color and texture with multi-class SVM (Support Vector Machine).For color feature extraction we used homogeneity model and for textural features we used PLD (Power Law Descriptor). With the help of SR-FCM (Soft Rough Fuzzy-C-Means) clustering. Membership functions based on the fuzzy set are facing the major problem of cluster overlapping. The rough set concepts can help us to get correct data from incomplete data, uncertainty of data. For defining the soft set theory there is no any requirement of parameterization tools. To get improved results of proposed algorithm the combination of aspects of fuzzy sets, rough sets as well as soft sets are used. The feature extraction for textural feature is done by using spatial domain which helps to reduce the run time complexity. Proposed method provides better performance which is compared with all the state of art techniques which is developed and analyzed using MATLAB.