Generation of Stereo Oil Paintings from RGBD Images

Painterly (or stroke-based) rendering is one of the major approaches for the creation of non-photorealistic rendering images. In this paper, a fully automatic algorithm is proposed, which is able to transform a photograph into a pair of oil-style stereo-viewable paintings. The developed stroke-based painting simulation algorithm generates a unique brush stroke each time it paints during the painting process according to the characteristics of the local image region. This special feature is able to reduce the undesirable systematic impression, which appears to be a common artifact of painterly rendering. The mobile imaging devices nowadays are often equipped with depth sensors, which capture depth information while taking a picture. By incorporating with the depth information, the program paints the left and the right images simultaneously. The resulting stereo paintings can be viewed by any type of stereo visualization technique.