Decoding Study Of HEVC Video Based On Parellella Multi-Core Processor

HEVC as a new generation of video coding standard,relative to the H.264/AVC standard,HE VC's compression rate is increased by about 50%without reducing the quality of video,but its high computing complexity has become a significant hurdle for application.So in order to improve the efficiency of decoding,we designed a parallel decoding algorithm for HEVC decoder based on Parallella multi-core processor platform.This paper will take Parallella the multi-ecore processor as the hardware platform,combining with OpenCL parallel coding framework to parallel processing for the key modules of HEVC decoder,so as to realize the high efficiency decoding.In this paper,the main work is as follows:(1)Analysis bottlenecks of H.265 decoding computational performance.By analysising HM of HEVC official test model,In HEVC Intra Prediction taked the most time in the process of decoding,about 29%of the decoding time,followed by IQ/IT module,about 25%of the decoding time,loop filter is about 16.7%.(2)H.265 partitioning of decoding task.First to partition task in accordance with the decoding process,then parallelly process these modules with a significant amount of calculation in the process of decoding.The NAL stream parsing and CABAC entropy decoding will be serially processed on dual-core ARM CPU of Parallella the multi-core processor.For Intra_Prediction module,IQ/IT module and deblocking filter of the loop filter module will be parallelly processed combine OpenCL heterogeneous framework on Epiphany-16.(3)Mapping and optimization on Parallel la platform,evaluating performance of parallel decoder.Through the assessment of HE VC parallel decoder designed in this paper,the speed ratio can reach 5.0,decoding time relative to the HM serial decoding algorithm reduces about 80%.