易于硬件实现的内嵌图像编码算法  

Embedded image coding algorithm for hardware implementation

作　　者：杜伟娜[1] 周磊[1] 孙军[1] 

机构地区：[1]上海交通大学图像通信与信息处理研究所,上海200030

出　　处：《红外与激光工程》2005年第3期352-355,共4页Infrared and Laser Engineering

摘　　要：EZBC算法综合利用了子带内和子带间系数的相关性,把零树/零块结构和基于上下文编码的优点有机结合在一起,获得了比SPHIT算法更好的压缩性能,比EBCOT更高的压缩效率。但是EZBC算法编码中的两个排序链表需要很大且非固定的存储空间,这使得EZBC算法的硬件实现非常困难。在EZBC算法的基础上提出了一种易于硬件实现、低存储量、高压缩性能的内嵌零块图像编码算法。该算法利用比特平面节点重要性状态表和上下文查找表来完成整个编码过程和形成上下文。实验结果表明,所提出的算法具有与EZBC算法基本相同的高压缩性能,但所需存储空间约为EZBC算法的四分之一,所以该算法更易于硬件实现。EZBC algorithm combines the advantages of zeroblock/zerotree coding and context modeling of the subband/wavelet coefficients by utilizing the correlation of inter-band and intra-band. In EZBC, the sophisticated context models were designed for coding quadtree nodes at different levels and subbands. Thus,EZBC outperforms SPHIT and can be competitive with EBCOT in compression efficiency. But a large amount of memory is required to maintain two lists that are used to store the coordinates of the quadtree nodes needed to be coded, also a great amount of operations to read and write the memory are required in each coding pass. These become drawbacks for a hardware implementation. An improved EZBC algorithm based on zeroblock and quadtree, with low complexity and high performance is presented in this paper. The improved algorithm utilizes the significance state table of bitplane nodes and the context look-up table to complete the coding passes and form the context, the comparison reveals that the PSNR results of the proposed algorithm are nearly the same performance as that of EZBC, furthermore, the algorithm requires low memory and reduces the implementation complexity.

关 键 词：零块编码 四叉树 内嵌编码 

分 类 号：TN919.81[电子电信—通信与信息系统]