Auto classification of skin symptom based on Mahalanobis distance

A scheme for auto classification of skin symptom is introduced in this paper. It classifies different skin symptoms based on the principle of least Mahalanobis distance. Skin images with symptom to be identified will be preprocessed at first. Basic operations of preprocessing includes color space transformation, image segmentation based on threshold by self-adapting method, image post-processing by mathematical morphology, edge detection, contour tracing, seed filling and so on. After that, twenty-six characteristic parameters are extracted from symptom areas of the processed image. Calculate the distance between these parameters and the preset parameters of standard symptoms (chloasma, comedo, blackhead and ephelis), and we can classify the symptoms to certain category in accordance with their Mahalanobis distance in terms of the least difference principle.