Deformation Model based Segmentation of Skin Lesions for Medical Image Data

Abstract

Dermoscopy is medical imaging technique used in diagnosis of skin lesions which needs computer aided diagnosis. An algorithm based on deformable model is proposed to segment and extract the skin lesions in dermoscopic images. Initially, RGB color model is converted into gray scale so that the information details of the images can be used efficiently to differentiate normal skin and skin lesions. The deviations in the color channels are combined to define the speed and the final criterion of ending the iteration of the deformable model. Our algorithm provides an effective and flexible segmentation. Two set of images from medical databases were used to estimate the performance.