Robust motion vector prediction algorithms with application to very low bit rate image sequence coding

摘要

Abstract: In this paper a new motion compensated (MC) predictivecoding method for image sequences at very low bit ratesis presented. This method utilizes a prediction of thedisplacement vector field (DVF) in order to produce anMC prediction error which is coded and transmitted.Assuming that the two previous frames are availableboth at the transmitter and the receiver the DVFcorresponding to the previous frame is estimated. Then,based on this estimate a temporal prediction of the DVFat the current frame is obtained. Using the assumptionthat the motion is constant along its trajectory, anauto-regressive (AR) model based method for performingsuch a prediction is proposed. According to theproposed scheme there is no need to transmit the DVF.Since the transmission of the DVF represents a sizeableoverhead in very low bit rate coding of video signals,this method constitutes a major contribution to bitrate reduction and quality improvement. Using thepredicted DVF, the MC prediction error is obtained,which is then transform and entropy coded. The proposedalgorithm is experimentally tested on standardvideo-conferencing image sequences and compared topreviously reported methods which transmit the motionvectors. Significantly improved results are obtainedcompared to the methods that transmit the motionvectors in terms of reduced bit- rate and improvedquality of the reconstructed image sequence. !24