Skin Cancer Classification Using K-Means Clustering

Detection of skin cancer gives the best chance of being diagnosed early. Biopsy method for skin cancer detection is much painful. Human interpretation contains difficulty and subjectivity therefore automated analysis of skin cancer affected images has become important. This paper proposes an automatic medical image classification method to classify two major type skin cancers: Melanoma, and Non-melanoma. In this paper, we have used the color and texture features in combination which gives better results than using color or gray level information alone. We have used k-means clustering algorithm to segment the lesion. The features are extracted by six different color-texture feature extractors from the segmented images. Classification accuracy of our proposed system is evaluated on four different types of classifiers and their values are compared with one another. The results of the proposed system are computed on five different classification rate in order to perform better analysis of our proposed system.