Diagnosis of melanoma with fractal dimensions

In this paper, we have been focused on the use of fractal concept in measuring the fuzziness of a mole. The box-counting method was used initially with two variations: fixed grid method and dynamic grid method. With the inherent difficulties in box-counting such as integral values of box sizes, we later investigated the variation method and the correlation method. However, the above methods depended on the segmentation result which separated a mole from its surrounding skin. In order to overcome the problem, we had experimented with two different methods which manipulated the intensities around the border of a mole. The first one calculated the size of the intensity surface area at different scales and the second method used the average absolute intensity difference of pixel pairs to obtain normalized fractional feature vectors. The paper reports the different ways of calculating the fractal dimensions and compares their differentiation power in the use of diagnosis of melanoma images.<>