Multiscale approach for skin lesion analysis and classification

Abstract

Skin cancer is one of the most deadly cancers in the world. If not diagnosed in early stages it might be hard to cure. This paper suggests a new approach for automatic segmentation and classification of skin lesion for dermoscopic images. The segmentation is based on a pre-processing; using the color structure-texture image decomposition to decompose a textured image into texture and geometrical components. Geometrical component is used in the lesion segmentation and the texture component is used to extract the lesion texture features. Feature classification is performed using the Support Vector Machine (SVM) classifier. The efficiency and the performance of the proposed approach are evaluated in comparison with recent and robust dermoscopic approaches from literature.