Region-based adaptive bilateral filter in depth map coding

In this paper, we present a new depth map coding scheme to improve coding efficiency. For this purpose, we developed an adaptive bilateral filter that restores the coded depth maps in accordance with the estimated noise variance. More specifically, each pixel is classified in such a way that the distortion cost is minimized and is restored depending on local noise level. The restored picture is used for the motion-compensated prediction of future pictures. This can improve coding efficiency. In particular, we propose classifying each pixel based on the optimal threshold and selecting the optimal fall-off parameter, which makes the bilateral filter more powerful. Finally, video coding that performs the new approach was implemented and this resulted in 11.4% (rate-distortion sense) and 12.0% (rate-quality sense) BD-bitrate reduction on average in FRExt Profile when compared with H.264/MVC.