Perceptually based video rate control using pre-filtering and predicted rate-distortion characteristics

In digital video coding, the rate control scheme is essential to regulate the output data rate and maintain the output quality. The control scheme defined in MPEG Test Model 5 provides a solution with very light computational overhead, but the results are not guaranteed to be good for all video sources and channel rates without resorting to a manual “tweaking” of the control parameters, with many trial-and-error encoding tests. In this paper, we propose a rate-distortion based control scheme where we use pre-filtering and block classification to achieve higher quality at low rates. In particular, we show how to efficiently include the pre-filtering parameters as part of the rate control optimization process. Our coded sequences are compatible with standard MPEG decoders and our method is suitable for channel rates lower than those normally used for MPEG-1 sequences (around 1 Mbps), but which may be more appropriate for Internet applications. In addition, our algorithm is generic and can be readily be extended for H.263/MPEG-4 encoders. Our results show significant reductions of the blockiness usually encountered at low rates, when compared to schemes such as TM5