Early diagnosis in the analysis of suspected pigmented skin lesions is very important in order to excise precociously malignant lesions and to avoid unnecessary removal of the benign ones. Decisions are based on colors, shapes and irregularities of the lesions and the differences of these characteristics in repeated observations. An objective measure of the boundary irregularity can be obtained estimating fractal dimension, DF, of the lesion contour. Since previous attempts have given contradictory results, we propose a method based on chromatic image characteristics, instead of the more common black and white. The image acquisition was performed by means of a high resolution apparatus, the SkinLab system and fractal dimension of processed image contour has been estimated by the box-counting method implementing the algorithm on Matlab software. The method could help in discriminating between nevi and melanomas and in following lesion evolution with time.