Motion estimation for depth maps coding based on rendered view distortion

Incolour plus depth map 3D video, depth map is a luminance image that aids the virtual views generation. Since it is not the final image viewed by end user, the conventional depth map compression method, which minimizes the pixel errors of the depth map, introduces more distortions to the virtual rendered images. In this paper, a novel motion estimation scheme for depth map coding is proposed. By employing rendered view distortion in Rate-Distortion Optimization, the optimal Motion Vector is determined to improve the rendered views quality. An effective Lagrange multiplier is also derived. Experimental results illustrate that 34%-63% of bit can be saved comparing to the conventional motion estimation method.