A CBR Rate Control Algorithm Based on IQM

摘要

This paper proposes a rate control algorithm based on image quality model (IQM). Although classical CBR rate control algorithms can satisfy the need for constant bit rate, they rarely protect the image quality enough in case of scene transitions. As a result the image quality degenerates acutely. These algorithms generally have quite complex control models, and they must compute quantization parameter for each macro block at macro block level. The algorithm proposed by this paper sets up a rate control model strictly based on image quality at frame level. In terms of the image quality parameter, we can predict the optimum quantization parameter for the next frame directly, so we can obtain steady image quality of reconstructed frames under the whole restriction for bit rates. Experimental results show that this algorithm can achieve a higher objective and subjective quality at sudden and gradual scene transitions than TM5, MPEG-4 VM18 SRC and Divx rate control algorithms and assure minimal rate control error at the same time. In addition, this algorithm is much easier and faster than other algorithms, so it can be used in real time applications