The motion vector reuse algorithm to improve dual-stream video encoder

Because of the limitation of bandwidth, video coder needs one stream of high-resolution for local storage and one steam of low-resolution for network transmission. Reusing high- resolution stream is the very point when designing for improving efficiency. Based on the analysis of the existing methods for motion vector synthesis with using four standard testing sequences, we proposed a Double Predictor Refinement method for the fine treatment. It is restricted in a window that has only four pixels. The re-fine treatment is combined with the decision on mode of macroblock method. The result shows that the PSNR is a little lower than re-motion estimation, but the search times are much fewer.