An efficient complexity-scalable video transcoder with mode refinement

MPEG-4 AVC (H.264) has been recently approved and is rapidly being adopted for applications including HD-DVD and satellite broadcast. To facilitate inter connectivity between different applications using MPEG-4 AVC, transcoding will be a key factor. When requantizing a bitstream, the incoming coding decisions are usually kept unchanged in order to reduce complexity. However, this can have a major impact on the coding efficiency. This paper proposes a novel mode refinement algorithm for inter prediction when requantizing MPEG-4 AVC bitstreams. The proposed approach produces a quality comparable to a full search with only 10% of its complexity, by exploiting the statistical properties of the mode distribution to enable restricted motion vector refinement. Moreover, it enables a trade off between complexity and quality by adapting the algorithm to the computational resources available.