An improved segmentation approach for skin lesion classification

Abstract

Skin cancer is considered as one of the dangerous types of cancer with a high proportion of deaths. This cancer can be categorized into two main types; Melanoma, which is the deadliest form, and Non-Melanoma. Early melanoma detection and diagnosis allows more treatment options and decreases significantly the number of deaths. Many researchers proposed to use image processing for skin lesion detection. The process can be divided into three main stages: lesion identification based on image segmentation, features extraction and lesion classification. Segmentation and features extraction are the key-steps and significantly influence the outcome of the classification results. In this paper, an improved segmentation approach for skin lesion detection and classification has been proposed. The proposed approach consists on a pre-processing based on multiscale decomposition thats separate the input image into two components. The geometrical component will be used in the segmentation stage and the texture component in features extraction. The efficiency and the performance of the proposed approach has been evaluated in comparison with recent and robust dermoscopic approaches from literature.