| Video compression is one of the key supporting technologies of multimedia communications, and palys a important role in the application and development of multimedia technology. Due to the increasing performance requirements of decoding H.264/AVC in HDTV or larger resolutions, new approaches are necessary to enable real-time processing. With the development of multi-core architecture, parallel decoding in the multi-core processor environment becomes an efficient way. However parallel decoding algorithm is much more challengeable, due to high complexity of H.264 decoder, communication between cores and task schduling. It has become the focus of recent academic research and real application.Most of the proposed parallel decoding algorithm for H.264 just focus on the minimizing of speedup, ignoring the requirement for real-time, load balance, and scalability. The main purpose of this research is to resolve the question of how to take a comprehensive consideration of these factors, and finish the design of parallel decoding algorithm.According to the characteristics of H.264/AVC standard, this research analysed the parallelism of decoding algorithm, presented an approach of Hybrid Decomposition. An adapted MB-group level data parallelization was proposed to deal with the performance loss caused by frequent scheduling in high resolution. A dynamic task scheduling mechanism based on condition matrix and dispatcher queue was also designed for effective control of the load balancing. Theoretical analysis of speedup was given, and a serial extension was presented for breaking the bottle-neck of pipeline.We implemented parallel decoding based on the TILE64 Homogeneous Multi-Core Processor in this paper. Some local optimization was also done for seeking the optimal performance. The experiment result shows that the proposed parallel decoding algorithm is feasible and efficient with good speedup, real-timing, load-balancing and scalability. |
PDF Full Text Request