Error concealment for the transmission of H.264/AVC-compressed 3D video in color plus depth format

• Recovery of lost information of one kind (color or depth) with the aid of the other kind.
• Classification of blocks into different kinds for different MV refinement and extrapolation.
• Shape-adaptive motion compensation.
• Joint matching based on color and depth information.

An error concealment method for the transmission of 3D video in color/depth dual stream format (i.e., separately encoded) is proposed. We consider frame losses of color or depth along error-prone channels. Based on motion vector extrapolation from previous decoded frames, lost frames can be reconstructed. Specifically, the received information of one kind (color or depth) is intensively used when recovering a lost frame of the other kind. Our contributions come from classification of blocks into different kinds for different processing, joint color/depth matching, and shape-adaptive motion compensation. According to experiments, our proposed method, against traditional MV extrapolation (for color) and MV sharing (for depth) schemes, has shown distinctive gains in both color and depth. Our algorithm is effective in retaining integrity and smoothness of object boundaries, which is important to view synthesis for 3D display.