ABSTRACT
Nowadays, skin cancer has become common among people. Skin cancer is the uncontrolled growth of cells within the skin and they can even spread to the other parts of the body and become fatal. The three broad types of skin cancer are squamous cell carcinoma, basal cell carcinoma, and melanoma. Mortality due to melanoma are maximum among patients and hence melanoma which is kind of skin cancer is dangerous when it grows outside the epidermis. This seems to be threatening in comparison to squamous cell carcinoma, basal cell carcinoma. Skin cancer can be detected during the primitive stage by non-invasive computerized dermoscopy. The proposed system is to detect melanoma using image processing that involves certain procedure, first stage is preprocessing followed by segmentation, next is feature extraction, and finally classification. The algorithm proposed for each step is the Sobel process, Otsu’s method, ABCD rule, and K-means with support vector machine (SVM) classifier. These algorithms, when implemented together, give good accuracy in terms of the values of size and shape, color, and texture of the lesion. This leads to extract the region of interest, which is then utilized for computerized surgery. PH2 dataset is used for producing the results.
