Categorizing Traditional Chinese Painting Images

Traditional Chinese painting (”Guohua”) is the gem of Chinese traditional arts. More and more Guohua images are digitized and exhibited on the Internet. Effectively browsing and retrieving them is an important problem need to be addressed. This paper proposes a method to categorize them into Gongbi and Xieyi schools, which are two basic types of traditional Chinese paintings. A new low-level feature called edge-size histogram is proposed and used to achieve such a high level classification. Autocorrelation texture feature is also used. Our method based on SVM classifier achieves a classification accuracy of over 94% on a 3688 traditional Chinese painting database.