A novel approach of black skin lesion images segmentation based on MLP Neural Network

Black skin lesion images generally show low contrast between lesion parts and safe ones. Segmentation of such images is a challenging task. In this paper, we propose a novel method that first select safe likely skin region using a Multi-Layer Perceptron (MLP) Neural Network. Then, color of these selected regions is transformed, in order to increase the contrast with lesion regions. The obtained image is therefore segmented using clustering algorithms. The proposed method has been applied to many black skin lesion images and produced good results.