Research Of Parallelism In H.264 Encoding Algorithm And Implementation On Multi-core DSP

In order to compress video information more effectively, MPEG and VCEG developed H.264 standard. In order to get better coding efficiency, H.264 integrates many high complexity encoding tools. It's hard for single processor to meet real-time needs. With the expanding of digital signal processors application, multi-core processors on chip become a hot spot.This paper studies H.264 encoding algorithm and the structure of YHFT-QDSP. We design three parallel encoding schemes. Main work and achievements of this thesis are listing below.1,This paper studied structure and working methods of YHFT-QDSP, especially on the methods of data transmission among multi-cores.2,We also give a deep research to H.264 standard, and analysis the parallelism of it, focusing on the data-level parallelism,3,We raised a block-based parallel algorithm based on the analysising of data parallelism. On base of this parallelism, we discuss the key points: the demarcation and distribution of images, data correlation while intra predicting, data correlation while inter predicting, data correlation while filtering, and data correlation while entropy encoding.4,In this paper, we proposed three parallel coding schemes: parallel entropy coding parallel intra codes; entropy coding parallel intra-coding and entropy coding parallel inter-frame. These schemes can parallel encode images without distortion.5,We also implement these three schemes on QDSP simulator, and transfer correlate data through SDP and Qlink. In order to get better parallel efficiency, we establish new data structure and synchronous methods between processors.6,According to simulator information, we analysis the performance of our schemes. We also analysis the performance of data transmission components. In order to get better performance, we also give some advicesIn the dual-processor parallel coding experiments, intra prediction coding speedup achieved 1.7, inter-frame prediction coding speedup reached 1.8; Parallel coding on four processors, intra-and inter-frame coding speedup are all above 3.