基于阵列的固定块全搜索分形编码并行计算结构  

Array-Based Parallel Computing Architecture of Fixed-Size-Block Full Search Fractal Coding

作　　者：赵伟[1] 

机构地区：[1]河南大学国际教育学院,河南开封475001

出　　处：《微电子学与计算机》2013年第9期164-168,共5页Microelectronics & Computer

摘　　要：与DCT结合的分形编码能够利用两种技术的优点,在提供较高压缩比的同时确保图像的质量.分形编码对于计算量的需求是制约其编解码速度进而影响其应用实时性的主要因素.本文针对固定块大小分形编码与DCT结合的编码结构的计算步进行分析,针对占其主要计算开销的固定块大小全搜索分形编码结构,研究其数据局部性和计算并行性,应用阵列结构,提出一种结构规整、易于扩展并能够方便实现的并行计算结构.该结构能够映射为VLSI实现,以高效的硬件结构提供编解码计算结构所要求的实时性.Abstract： The combination of fractal coding and DCT can take advantage of the benefits of both technologies. Higher compression ratio is provided, and at the same time, the quality of image is acceptable. The calculation amount of fractal coding is the main factor, which is restricting its codec speed, and affects its real- time applications. In this paper, fixed- block--size fractal coding and DCT encoding structure calculation step is analyzed. The data locality and computational parallelism of full search fractal coding structure is studied, which is the major computational overhead. A regular parallel computing structure is proposed by using array structure. This structure is scalable and can be easily mapped to VLSI implementation. The efficient hardware structure can meet the real--time requirements of the codec computing structure.

关 键 词：分形编码 固定块 阵列结构 并行计算 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]