Rate-Distortion-Complexity Optimization of H.264/AVC for Multiple Programs on a Multi-Core Processor

摘要

In this paper we analyse the encoding of multiple video programs when optimizing rate, distortion, and complexity jointly. We assume that the processing platform allows allocating the processing resources dynamically between the encoders, such as e.g. possible when multiple encoder instances run on a multi-core processor. The goal is to allocate processing resources to the program which currently benefits most from it. For this purpose, the complexity-distortion trade-off is evaluated for each Group of Picture (GOP) and resources are allocated dynamically on a GOP time frame. Besides dynamic allocation of processing resources we also consider the dynamic allocation of bit rate under consideration of the rate-distortion trade-off. The analysis of the potential gain with respect to several system parameters (duration of GOP, processing resources, bit rate) is the topic of this paper. Simulation results are presented for an ensemble of 8 programs encoded with H.264/AVC at various bit rates and complexity levels. The simulation results indicate that an average gain of 3.2 dB PSNR can be achieved where most of the gain can be attributed to rate-distortion optimization rather than complexity distortion optimization.