抄録
In order to classify skin lesions as benign or malignant the dermoscope (a microscope that shows the pigment structure of a skin lesion) is used. However, it is difficult for a doctor to classify a skin lesion just by a dermoscopic image. Therefore dermoscopic images are analyzed by the computer. It is desirable to segment areas of the same color in order to distinguish between healthy skin and the skin lesion and to analyze the pigmented structures. In this research the RGB data are transformed into various color spaces. The color vectors are clustered by a self-organizing map. After learning the similarities between the clusters are minimized by fusion of neighboring neurons.
