An automatic depth map generation method by image classification

This paper presents an automatic algorithm which generates depth map for 2D to 3D conversion. The depth map is generated by fusing global depth gradient and local depth refinement. First, the proposed algorithm judges the scene type of the input image. Then, it generates global depth gradient of the image according to its scene type. At last, local depth refinement is implemented by using warm/cool color theory. Experimental results show that the algorithm has advantages in both adaptability and processing time.