Perceptual Segmentation of Natural Color Images by Fuzzy Split-and-Merge Algorithm and Its Application to Dermoscopy Images

Abstract

The purpose of this paper is perceptual segmentation of natural color images and its application to the segmentation of dermoscopy images, which are closeup images of pigmented skin lesions. We propose the fuzzy split-and-merge algorithm that integrates the L^*a^*b^* components as color features and the statistical geometrical features as texture features by using fuzzy inference. The proposed method is composed of four stages: simple splitting, local merging, global merging and boundary refinement. We perform the comparative experiments of the proposed method in comparison with the human segmented results to dermostrate the effectiveness of the proposed method.