Automated Segmentation of Skin Lesions Using Seed Points and Scale-Invariant Semantic Mathematic Model

A color image-based segmentation method for segmenting skin lesions is proposed in this paper. This proposed methodology mainly includes two parts: First, a combination of scale-invariant and semantic mathematic model is utilized to classify different pixels. Second, a strategy based on skeleton corner point's extraction is proposed in order to extract the seed points for the skin lesion image. By this method, the skin slices are processed in series automatically. As a result, the lesions present in the skin can be segmented clearly and accurately. The proposed algorithm is trained and tested for 360 skin slices in order to evaluate the accuracy of segmentation. Overall accuracy of the proposed method is compared with existing conventional techniques. An average missing pixel rate of 3.02 % and faulting pixel rate or 2.36 % has been obtained for segmenting the skin lesion images.