Pattern classification of nevus with texture analysis.

The purpose of this research is to classify the pattern on the surface of the nevus. The digital image that contains one nevus is classified into three kinds of patterns of homogeneous pattern, globular pattern, and reticular pattern by the texture analysis. The tumor part in the image is specified first, and the specified tumor part is divided into some sub-images. Afterwards, the amount of the texture features of each sub-image was calculated. The pattern was classified by the discriminant analysis based on the amount of the texture features. As a result, the patterns could be classified correctly into three categories at the ratio of 94%.