In this paper, we propose a successful approach for automatic segmentation and classification of skin lesions in two classes: melanoma and non-melanoma. Initially, skin images obtained in a clinical environment are pre-processed to remove unwanted hair and reduce noise. A region growing segmentation technique using automatic initialization of seed points is then applied, in order to isolate the lesion areas for further processing. Subsequently, the extracted lesion areas are represented by a set of color and texture features. Using a Support Vector Machine (SVM) classifier, the features extracted from each segmented lesion are organized in a feature vector that is further used to discriminate the lesions between melanoma and non-melanoma. The classifier performance was evaluated using a stratified holdout approach to measure its sensitivity, specificity and accuracy. The best results for accuracy were obtained with the Gaussian kernel and are in the order of 85%. The results are promising compared with similar works [1].