Efficient one-pass frame level rate control for H.264/AVC

  • Authors:

  • Xiang Li;Andreas Hutter;André Kaup

  • Affiliations:

  • Chair of Multimedia Communications and Signal Processing, University of Erlangen-Nuremberg, Erlangen, Germany and Siemens Corporate Technology, Information & Communications, Munich, Germany;Siemens Corporate Technology, Information & Communications, Munich, Germany;Chair of Multimedia Communications and Signal Processing, University of Erlangen-Nuremberg, Erlangen, Germany

  • Venue:
  • Journal of Visual Communication and Image Representation
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

In this paper, an efficient one-pass frame level rate control algorithm is proposed for H.264/AVC, where the two essential problems in rate control, i.e., the budget allocation (BA) and the quantization parameter determination (QPD) are both considered. First, an efficient BA scheme is designed with special consideration of the inter-frame dependency. Accordingly, the error propagation caused by improper QP assignment in the motion compensation process is reduced and the total distortion is kept at a low level. Second, a better QPD method is developed based on an accurate rate model and a second feedback mechanism so that a high rate accuracy is guaranteed. Simulations verified the performance of the proposed algorithm. Compared with the fixed QP coding and the two recommended rate control algorithms (G012-MB and G012-Frame) in H.264/AVC reference software, up to 1.50, 1.12, and 0.94dB were achieved with a higher rate accuracy in the coding efficiency test defined by ITU-T VCEG. Particularly, a more significant improvement was observed in slow movement scenarios: compared with G012-MB, a gain of 0.80dB on average and 1.43dB at maximum was obtained with 66.0% reduction of average rate mismatch. Compared with G012-Frame, the average and maximum gains are 0.34 and 1.06dB, respectively. While at the same time, the average rate mismatch was reduced by 90.4%. Considering the low computational cost, the proposed algorithm is quite appealing to real-time video applications.