A colour image processing method for melanoma detection

Abstract

In this paper, we propose a method dedicated to classification between benign and malignant lesions in Dermatology in the aim to help the clinicians for melanoma diagnosis'. The proposed methodology reduces the very numerous informations contained in the digitized images to a finite set of parameters giving a description of the colour and the shape of the lesions. The whole process is shared in three steps: preprocessing, segmentation and classification of the lesions. The proposed method was applied on a data base of 38 lesions (20 benign lesions and 18 malignant lesions) in the aim to assess the feasability of the proposed method. The good classification rate obtained with the method is discussed and later tests to engage are underlined.