Melanoma is the most fatal form of all skin cancer types. An early screening of melanoma can greatly contribute to successful treatment, hence reliable early detection systems are highly demanded. In this paper, we propose a novel melanoma detection method based on Mahalanobis distance learning and constrained graph regularized nonnegative matrix factorization. The proposed method allows supervised learning for feature dimensionality reduction by incorporating both global geometry and local manifold, so as to enhance the discriminability of the classification performance. The proposed method is evaluated on PH2 Dermoscopy Image Dataset and Edinburgh Dermofit Image Library, with comparison against four alternative classification methods. Our method demonstrates the best performance, with 94:43% sensitivity and 81:01% specificity on PH2 dataset and 99:50% sensitivity and 93:68% specificity on Edinburgh Library.
展开▼
