摘要
传统的JPEG2000MQ编码器串行编码效率低下,同时现有的多上下文并行编码的MQ编码器占用资源过大.本文对MQ编码算法中的运算流程,索引值和概率估计值的求解函数,条件交换和重归一化算法等四个方面进行了优化,减弱了上下文之间的依赖性,简化了条件交换和重归一化算法的复杂度.依据该算法,本文提出了一种高速的MQ编码器VLSI结构,实验结果表明,本文提出的MQ编码器VLSI结构能够工作在532.91MHz,吞吐率为532.91Msymbols/sec,相比Dyer提出的Brute force with modified结构,工作频率提高1倍,吞吐量提高近27%,且面积仅为其四分之一.
MQ-encoder,which is a key bottleneck in the JPEG2000 image compression system,presents challenges for realization of a high-speed and low cost VLSI architecture.In this paper,some optimization work has been done to the MQ-encoder arithmetic about the process flow,index and probably estimation function,condition change and renormalization.These optimization schemes are highly effective in simplifying the relation between contexts,reducing the complexity of the condition change and renormalization process,and cutting down the critical path delay of the pipeline.Based on this improved arithmetic,a VLSI architecture for an MQ-encoder with high speed is proposed.Synthesis result shows that the processing speed of the MQ-coder could reach as high as 532.91MHz with a throughput of 532.91 Msymbols/sec.Compared with Dyer＇s architecture,the architecture with an improved throughput as high as 27% presented in this paper can get a speed two times greater than the former one while its area is only a quarter of the former one.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2012年第11期2158-2164,共7页
Acta Electronica Sinica
基金
中央高校基本科研业务费专项资金(No.K50511010017)
