Accurate bit prediction for intra-only rate control

  • Authors:

  • Ling Tian;Xin Yin;Yu Sun;Shixin Sun

  • Affiliations:

  • College of Computer Science & Engineering, University of Electronic Sci. & Tech. of China;Dept. of Computer Science, University of Central Arkansas;Dept. of Computer Science, University of Central Arkansas;College of Computer Science & Engineering, University of Electronic Sci. & Tech. of China

  • Venue:
  • ICME'09 Proceedings of the 2009 IEEE international conference on Multimedia and Expo
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

Rate control plays a crucial role for video communication applications. It ensures that the generated compressed bit streams satisfy bandwidth and buffer constraints. Rate control algorithms recommended by H.264/AVC adopt rate-distortion (R-D) models for inter-frames to determine quantization parameters (QPs) but not for intra-frames. Instead, they directly compute QPs without any considerations of bitrates and coding complexities for intraframes. In order to obtain more accurate target bit prediction for intra-frames, we first introduce the geometry gradient information as a new complexity measure to accurately represent the complexities for intra-frames. Then, we propose a novel R-D model which is an integration of a linear rate-complexity model and an exponential rate-quantization model. Finally, we develop an accurate and robust intra-only rate control algorithm for H.264/AVC. Experimental results demonstrate that, compared with JVT- W042, the proposed algorithm achieves higher precise bit estimation, provides more robust buffer control, and also improves coding quality.