Adaptive wavelet coding of the depth map for stereoscopic view synthesis

Abstract

Multi-view video and 3D television are emerging applications raising the problem of efficient encoding of a depth map, in addition to classical texture images. This paper investigates depth image coding via an adaptive wavelet lifting scheme. Switching between long filters in homogeneous areas and short filters over the edges of the depth map is decided based on the contours detected in the texture image. The method takes thus into consideration the correlation existing between the edges in the texture and in the depth image, leading to an improved encoding of the latter one.