DESIGN CHALLENGE OF HIGH-DEFINITION VIDEO ENCODER ARCHITECTURE AND EVALUATION MODEL OF MULTIPLE TARGET PERFORMANCE

Yin Haibing,Zhang Xiaoyun,Gao Zhiyong
DOI: https://doi.org/10.3969/j.issn.1000-386X.2012.10.007
2012-01-01
Abstract:FPGA and ASIC are the proper implementation platforms for high-definition video encoder.Design of video encoder architecture encounters quite a few challenges,such as high throughput burden,extremely high external memory access bandwidth,high hardware resource cost,data dependency chokes normal pipelining,and difficulties in multiple targets performance trade-off.The design of high-definition video encoder architecture is supposed to be in combination with the algorithm optimisation for realising the trade-off between the performances of multiple targets.How to evaluate the multiple target performance of hardware architecture is the problem having to be solved first.In this paper,we propose the metric method for multiple target performance parameters on the basis of analysing the encountering challenges of chips architecture design and the multiple target performance typically concerned,and make reasonable mapping and normalisation process on multiple target parameters,then we propose at the first time the evaluation model for multiple target performance.The multiple target performances of predominating high definition H.264/AVC encoder architectures are evaluated and compared with this model.Our study is of meaningful and valuable as the reference for the optimisation of H.264/AVC and next generation HEVC encoder architectures.
What problem does this paper attempt to address?