Melanoma is the most aggressive type of skin cancer. The mortality degree and the treatment costs associated with this disease can be drastically reduced by its early detection. In this context, we investigate the correlation dimension of the attractor associated to a skin lesion for its relevance in the context of melanoma identification. The analysis is performed on non-standardized color images taken with a mobile device (smartphone) camera. The grey scale image and each color channel are separately investigated. It can be concluded that the correlation dimension calculated for the red image channel and the grey scale image represent relevant texture descriptors, with high standalone sensitivity and specificity and they should be included, along with other (textural, color, geometric) descriptors, for melanoma early detection.