Fast motion re-estimation for arbitrary downsizing video transcoding using H.264/AVC standard

Downsizing video transcoding is a technique for adapting the bit rate or frame size of a preceded video to suit better the constraints of transmission networks and target applications. To sustain good video quality at reduced bit rates, we have previously proposed H.263-based transcoding techniques for downsizing preceded videos by an arbitrary factor. To fully exploit the advantages of video transcoding, we investigate and propose in this paper an effective and fast motion re-estimation method for arbitrary downsizing video transcoding using the newly developed H.264/AVC standard. Specifically, as H.264/AVC allows more macroblock partitions and reference frames for inter-frame motion-compensated prediction, we propose an area-weighted vector median filter to estimate the new required motion vectors from the motion vectors existing in a preceded video. Experimental results show that the proposed method can achieve a notable improvement in both subjective and objective video quality for transcoding preceded H.263 or H.264 videos at reduced bit rates and frame sizes.