A new bilateral filter for post-removing the noise of synthesis view in 3D video

In depth-based 3D video, the common noises in texture and depth videos of reference views may not be removed before view synthesis. Then they are transferred to generated synthesis view. We find that the noise additionally confuses the edge profiles in the synthesis view compared to that of natural images, for which the famous bilateral filter (BF) does not work. In this paper, we develop a new explicit filter for the synthesis view, which has a similar expression with the BF but not the same weight terms. On one hand, the spatial term in the classical BF is directly adopted due to its robustness to the noise, which gives high weights to spatially closed pixels of the pixel to be filtered. On the other hand, a reliability term is designed, which gives high weights to pixels that are not likely affected by the noise. It is inspired by the finding that not all pixels are severely affected in the synthesis view. In this way, true edge profiles are protected in the filtering process. Experiments are made and compared to two explicit filters, which verifies its effectiveness in removing the noise and protecting the edge profiles.