This paper presents a segmentation approach of melanoma by color morphology. The images of melanoma are filtered by morphological tools using a lexicographic order onto HSI color space. A thresholding technique is applied to segment the melanoma Region of Interest (ROI). Binary morphological techniques are used to filter the ROI. The approach was tested on two benign and malignant image databases, both containing 100 images, and the results were compared to ground-truth segmentation and Fuzzy CMeans one. By performing twelve metrics, the results have shown the promising aspects of this approach to segment benign and malignant lesions.
展开▼
