A new motion and disparity vector prediction technique for H.264 based stereoscopic video coding

Abstract

Owing to the provision of a more natural and realistic reconstruction of a scene in the form of two views presented to the left and right eyes, stereoscopic imaging/video technology offers a more attractive and effective solution for imaging/video production. However due to the fact that more information has to be transmitted to reproduce the stereoscopic image pair, the transmission bit rate might be considerably increased if the available redundancies are not properly exploited. This paper presents a novel motion and disparity vector prediction technique for the H.264 based stereoscopic video codec that can be used for future consumer video applications such as 3D TV and 3D mobile communication in order to generate the real 3D perception in the human psycho-visual system. The proposed technique predicts both motion and disparity vectors more intelligently using the previously coded motion and disparity vectors of the surrounding macroblocks. Results show that the proposed algorithm can reduce the average bit budget by about 12% compared to the conventional motion and disparity prediction algorithm employed by the H.264 [1] based stereoscopic codec[2]. Furthermore, results suggest that the objective quality of the sequences is also improved. © 2006 IEEE.