多中心动态聚类算法及对癌症与非癌症彩色手掌图像的分类

Multi-central Dynamic Clustering Algorithm Classifying Color Palm Images for Cancer Diagnosis

Observing palm is one of diagnosis methods in Traditional Chinese Medicine and HolographicMedicine. Generally, the shape, color, ridge and line features of palm are all important for palm diagnosis. As thefirst attempt for automated palm diagnosis, the color is used and a new statistical feature of color, moment feature, isdefined in this paper. Multi-central dynamic clustering algorithm based on our new feature is proposed to recognizecancerous palm images. Applying our approach to the images in the palm database including all kinds of pathologicaland healthy palm images, the experimental results indicate that it is effective to recognize cancerous palm images andsuperior over the K-mean algorithm.