New stereo video coding in redundant wavelet domain

The 3D technology has been one of the fastest growing technologies in the recent years. This paper proposed a high performance 3D stereoscopic video coding system in redundant wavelet domain utilizing a multiband variable size block motion compensation technique. Our approach is based on structure from motion techniques and uses Depth-based rendering to reconstruct the desired stereo views for each video frame. The depth image has a low energy and does not have sharp boundaries; therefore, it is not an easy task to obtain an accurate motion vector. Fortunately, the redundant wavelet domain provides a good solution by retaining all the phase information and provides a multiple prediction possibilities for motion techniques.