A fast block-based approach for segmentation and classification of textural images using contourlet transform and SVM

摘要

In this paper, the problem of texture image segmentation and classification using contourlet transform and Support Vector Machine (SVM) classifier is discussed and a new block-based approach is proposed. At first, the images are split into M × M blocks. In the next step expanded contourlet coefficients of each block are applied as an input to develop feature vectors. Energy and standard deviation of the coefficients, and their combinations are used in order to obtain feature vector of each block for SVM classifier. After classification of all blocks and merging them, two refinement phases containing block refinement and non-connected islands elimination are applied to image for obtaining final segmented and classified image. Finally, the results are compared with the results of other related works. The experimental results on prototype data showed that the proposed algorithm provides a faster tool with enough accuracy that can be implemented in a parallel structure for real-time processing.