Melanoma is a type of skin cancer that originates in melanocytes, since these grow and multiply out of control. World-wide, this disease represents approximately 1.5% of all tumors in both gender. The most usual screening technique for detection of melanoma is dermatoscopy, through the use of a manual device for subsequent evaluation of patterns in the image. This research proposes a method for the segmentation and characterization of dermoscopic images following the evaluation criteria ABCD. A set of 428 melanoma images from the International Skin Imaging Collaboration database was used and 3 stages were considered: image preprocessing: hair removal using a modification of the DullRazor algorithm, segmentation of the lesion by using thresholding methods and mathematical morphology and lastly, feature extraction. To evaluate the performance of the proposed method, manually segmented and automatically segmented images are compared, obtaining a Sørensen-Dice similarity index of 0.74, considered an important result, considering that it is only possible to make an approximate delimitation of the area of the lesion in these types of images.