A study on iterative video coding rate-distortion optimization and its rate control

摘要

Encoding a video clip with as large a compression ratio as possible within a framework of established standards is a key technology for digital TV broadcasting, content creation, distribution, and storage. Optimization, encoding a video clip at the highest possible quality within a given bit rate, is expected to become increasingly important in the future. As this process repeats encoding many times, rate control should converge quickly. This paper proposes an iterative encoding (N-pass) optimization strategy based on the Lagrange multiplier method. It focuses on local coding units and knowledge about optimal multipliers within an MPEG-2 framework. We also propose and demonstrate an affiliated accurate rate control algorithm. Simulation results show the PSNR gain for this optimization method is 1.0 to 1.9 dB higher than that offered by the TM5 method, and is significantly higher than the 0.03 to 0.7 dB gain improvement over TM5 we achieved with quantization-parameter-based optimization. The rate control framework is highly accurate; the error is less than 1% within one or two rate control iterations.