Neda Razazzadeh - M.S. Student, Dept. Computer and Informatics Engineering, Payame Noor University, Qeshm, Iran
Mehdi Khalili - Assistant professor, Dept. Computer and Informatics Engineering, Payame Noor University, Tehran, Iran

Differentiation of benign and malignant (melanoma) of the pigmented skin lesions is difficult even for the dermatologists thus in this paper a new analysis of the dermatoscopic images have been proposed. Segmentation, feature extraction and classification are the major steps of images analysis. In Segmentation step we use an improved FFCM based segmentation method (our previous work) to achieve to binary segmented image. In feature extraction step, the shape features are extracted from the binary segmented image. After normalizing of the features, in classification step, the feature vectors are classified into two groups (benign and malignant) by SVM classifier. The classification result for the accuracy is 71.39%, specificity is 85.95%, and it has the satisfactory results in sensitivity metrics.

Dermoscopic images, Segmentation, Shape features, SVM classifier