Prescreening of Skin Lesions from Digital Images

Melanoma is one of the deadliest form of skin cancer. Incidence rates of melanoma have been increasing, especially white males and females, but survival rates are high if detected early. The automated system assess a patients risk of melanoma using images of their skin lesions which is taken using a standard digital camera. Locating the skin lesion in the digital image is a big challenge. A set of texture distributions were learned from an illumination-corrected photograph. Texture distinctiveness metric is calculated for each texture distribution. Later regions in the image are classified as normal or cancerous based on representative texture distributions. A texture filter based skin lesion segmentation is proposed. Based on this method the region of lesion is segmented without noise. Features are extracted from the segment and fed to classifier. The proposed work has higher segmentation accuracy compared to all other tested algorithms.