Color Segmentation for Skin Lesions Classification

Differential diagnosis of Erythemato-Squamos diseases is considered a real problem in dermatology. They all share the clinical features of erythematic and scaling, with very little differences. This paper introduces an unsupervised color segmentation procedure applied to one disease of this group named Atopic Dermatitis. Evaluation of different color models is done to select the most appropriate model for representing skin lesions. Two steps of color segmentation were done to detect skin lesions with less computation. First is coarse segmentation with optimal threshold of the CIE color model, and second is fine segmentation with K-means clustering technique. Results of the proposed algorithm prove its success as the manual segmentation with expertise.