Efficiency analysis of multihypothesis motion-compensated prediction for video coding

Overlapped block motion compensation or B-frames are examples of multihypot hesis motion compensation where several motion-compensated signals are supe rimposed to reduce the bit-rate of a video codec, This paper extends the wi de-sense stationary theory of motion-compensated prediction (MCP) for hybri d video codecs to multihypothesis motion compensation. The power spectrum o f the prediction error is related to the displacement error probability den sity functions (pdf's) of an arbitrary number of hypotheses in a closed-for m expression. We then study the influence of motion compensation accuracy o n the efficiency of multihypothesis motion compensation as well as the infl uence of the residual noise level and the gain from optimal combination of N hypotheses. For the noise-free limiting case, doubling the number of (equ ally good) hypotheses can yield a gain of up to 1/2 bits/sample, while doub ling the accuracy of motion compensation (such as going from integer-pal to 1/2-pel accuracy) can additionally reduce the bit-rate by up to 1 bit/samp le independent of N. For realistic noise levels, it is shown that the intro duction of B-frames or overlapped block motion compensation can provide lar ger gains than doubling motion compensation accuracy. Spatial filtering of the motion-compensated candidate signals becomes less important if more hyp otheses are combined. The critical accuracy beyond which the gain due to mo re accurate motion compensation is small moves to larger displacement error variances with increasing noise and increasing number of hypotheses N. Hen ce, sub-pet accurate motion compensation becomes less important with multih ypothesis MCP. The theoretical insights are confirmed by experimental resul ts for overlapped block motion compensation, B-frames, and multiframe motio n-compensated prediction with up to eight hypotheses from ten previous fram es.