Computer vision systems are increasingly used for the early detection of skin cancers. Recognizing the first sign of melanoma is very important because if melanoma is found and treated in its primary stage the chances for long-term survival are excellent. On the contrary, as it progresses its treatment becomes increasingly harder and it has worse outcome. The various proposals of computer vision systems are characterized by some fundamental common phases: image acquisition, pre-processing, segmentation, features extraction and finally classification. Feature extraction aims at extracting the features from the lesion image in order to characterize the melanoma and feed the classifier. The recent research provided many different feature extraction algorithms for melanoma diagnosis from dermoscopy images from the simplest to the most sophisticated. Features are typically extracted using digital image processing methods (i.e., segmentation, edge detection and color and structure processing), and an open discussion about the meaning of these features and the objective ways of measuring them is ongoing. This paper is a contribution to the feature extraction phase as it describes the most frequently used features in the elaboration of computer vision systems and reports a description of recent works for feature extraction and classification.