Binocular stereopsis of traditional Chinese paintings

This paper presents an interactive technique for generating stereoscopic images from traditional Chinese paintings. The technique exploits a traditional way of moving focus based spatial composition used by classical Chinese painters. According to the moving focus rules of depicting objects of a painting, each non-ground pixel has the same depth with its corresponding pixel on the ground. Therefore, the depth map computation of the painting is decomposed into two steps. Firstly, the depth map of the ground is computed based on the moving focus rules of expressing spatial depth variation, with line-shape cues given through a user interface. Secondly, in order to calculate the depth of pixels belonging to the non-ground objects, the user interface provides a tool to sketch the objects and their occupying regions in the ground part. After that, each pixel of the non-ground objects is automatically matched to a pixel in the occupying ground regions by linear interpolation and takes the depth of the ground pixel as its depth. Finally, an anaglyph is computed by the obtained depth map. Experimental results demonstrate that the method presented in this paper can generate convincing binocular stereo images with easy user interaction.