Learning to Recognize the Art Style of Paintings Using Multi-cues

Visual characteristics of paintings display high-level semantic concept: art style to the viewers. Classification of art style depends mainly on human knowledge and experience, which remains a big challenge for computer vision. In this paper, based on careful studies on art literature, we propose a simple but effective method to automatically identify the art style between the Chinese wash painting and the foreign art painting. The efficiency of our method lies on that three cues: color contrast, blank-leaving and uniformity of illumination are utilized to recognize the art style of one image. Experiments results show that, our method outperforms state-of-the-art approaches, yielding higher precision while requiring less computation time. Using the cues presented in this paper, our method can successfully identify whether one painting belongs to Chinese or foreign art style.