A Bag-of-Features Approach for the Classification of Melanomas in Dermoscopy Images: The Role of Color and Texture Descriptors

The identification of melanomas in dermoscopy images is still an up to date challenge. Several Computer Aided-Diagnosis Systems for the early diagnosis of melanomas have been proposed in the last two decades. This chapter presents an approach to diagnose melanomas using Bag-of-features, a classification method based on a local description of the image in small patches. Moreover, a comparison between color and texture descriptors is performed in order to assess their discriminative power. The presented results show that local descriptors allow an accurate representation of dermoscopy images and achieve good classification scores: Sensitivity = 93% and Specificity = 88%. Furthermore it shows that color descriptors perform better than texture ones in the detection of melanomas.