Comparison between blur transfer and blur re-generation in depth image based rendering
Norishige Fukushima, Naoki Kodera, Yutaka Ishibashi, Masayuki Tanimoto · 2014
An issue of object boundary reconstruction is important in depth image based rendering. For natural rendering, pixel values in an object boundary should be gradually changed from the background color to the foreground color. There are two approaches for this problem; one is blur regeneration by using alpha matting, and the other is blur transfer by using foreground biased dilation. The former requires three additional steps, i.e. foreground, background texture synthesis, and alpha channel estimation. The latter requires only one step, which is dilation for a depth map. To validate the effectiveness of these methods, we compare the matting method with the dilation method. In addition, we modify the dilation method to improve the quality of the synthesis view. Experimental results show that the improved method reaches the state-of-the-arts of the matting method, and its additional computational cost within 4 ms.