AN IMPROVED RATE-QUANTIZATION MODEL FOR RATE CONTROL IN REAL-TIME VIDEO ENCODING (WA-P5)

Author(s) : Bo Xie (PacketVideo Corporation, USA) Wenjun Zeng (Univ. of Missouri-Columbia, USA)

Abstract : There is a constant need to improve the rate-quantization (R-Q) model to achieve accurate rate control for real-time video encoding. In the literature, considerable efforts have been focused on developing all kinds of R-Q models. In this paper, we extend our previous work in [1] and propose a new algorithm for R-Q model parameter estimation, as well as a new approach to address the model failure issue by bounding the quantization parameter (QP) determined by the R-Q model without introducing additional complexity. Experimental results show that the new R-Q model parameter estimation algorithm improves over the one we previously proposed in [1], and the QP bounding approach can address the model failure issue well. With both contributions, the proposed QP determination solution can achieve very close performance to the benchmark solution discussed in [1] that achieves the best performance by using an operational re-quantization approach.

Menu