Automatic Categorization of Traditional Chinese Painting Images with Statistical Gabor Feature and Color Feature

This paper presents an automatic statistical approach to categorize traditional Chinese painting (TCP) images according to subject matter into three major classes: figure paintings, landscapes, and flowerand-bird paintings. A simple statistical Gabor feature is presented to describe the local spatial configuration of the image, which is then integrated with color histogram that represents the global visual characteristic to build the feature subspace. A relative-distance based voting rule is proposed for final classification decision. The effectiveness of the proposed scheme is demonstrated by the comparable experimental results.