Research On Computer-aided Classification, Authentication And System Of Chinese Traditional Painting

Traditional Chinese painting (TCP) is precious cultural heritage of our country and is striding into the collection market of the world. But many imitations are badly affecting the internationalization of TCP. The traditional authentication method, which is strongly dependent on the experts' empirical knowledge , is usually lack of objective and convincing proofs. The computer-aided authentication of TCP based on recognition technology will help improve the reliability and scientificity. This thesis addresses itself to the authentication of TCP problem. The main contributions are as follows:Firstly, we propose an automatic classification algorithm of traditional Chinese painting based on statistical Gabor feature and color feature. The method combines the Gabor texture feature and color histogram feature to express an image, and then use the vote classifier to implement the automatic classification. At last the sample selection is used to improve the accuracy of the classification.Secondly, authentication method of TCP based on MRF is proposed. The method builds a three-level authentication model for the objects in TCP and uses belief propagation algorithm to infer the true or false state. It provides certain proofs for the quantitative analysis of authentication of TCP. The experimental results demonstrated that our method is promising.Thirdly, we provide a practical system of authentication of TCP based on the image recognition technology. The system classifies and retrievals the objects in TCP and help the expert to authenticate the TCP.