Segmentation of Skin Cancer Images Based on Multistep Region Growing

An automatic method for segmentation of skin cancer images is presented. Firstly the algorithm automatically determines the compounding colors of the lesion, and builds a number of distance images equal to the number of main colors of the lesion (reference colors). These images represent the similarity between reference colors and the other colors present in the image and they are built computing the CIEDE2000 distance in the Lab color space. Texture information is also taken into account extracting the energy of some statistical moments of the L component of the image. The method has an adaptative, N-dimensional structure where N is the number of reference colors. The segmentation is performed by a texture-controlled multi-step region growing process. The growth tolerance parameter changes with step size and depends on the variance on each distance image for the actual grown region. Contrast is also introduced to decide the optimum value of the tolerance parameter, choosing the one which provides the region with the highest mean contrast in relation to the background. The method has been tested with a database of 20 images obtaining excellent results