基于内存优化的小波零块嵌入图像编码算法

小波嵌入零块编码算法(Set Partitioned Embedded bloCK,SPECK)是一种高效的具有渐进传输特性的图像编码算法,但其在编解码过程中的巨大内存占用致使编解码速度慢且不利于硬件实现.本文提出了一种基于内存优化的小波零块嵌入图像编码算法,采用标志状态图和块深度优先搜索策略完成嵌入编码中的排序和细化过程.两张标志状态图分别标识编解码过程中的重要系数和不重要集合,同时结合块深度优先搜索策略检索块结构中的不重要集合,代替排序和细化过程中的重要系数链表和不重要集合链表,较大地节省了内存占用.实验结果表明,与SPECK算法相比,在保持相当信噪比的情况下,内存占用仅为原来的1/12,与另一种低内存零树编码算法LZC(Listless Zerotree Coding)相比,在内存略有增加的情况下,信噪比至少提高1.1dB,为硬件实现小波零块编码算法探讨了一条有效途径.

A New Low Memory Set Partitioned Embedded Block Coder

Said and Pearlman's Set Partitioned Embedded block(SPECK) algorithm is computationally simple and efficient for image compression and progressive transmission,however there is a main drawback of high memory requirement for hardware implementation.We propose a new low memory embedded block coder that keeps those desirable features,and the novelties are twofold.A compact form of the flag maps is introduced to store the significant coefficients and sorting information instead of using lists.A new block depth-finding strategy is developed for searching insignificant sets at sorting stage.Experimental results show that the obtained PSNR values for the decoded images are very close to those of SPECK algorithm and the memory consumption is reduced by 12 times.Compared with another low memory coder(Listless Zerotree Coder,LZC),the PSNR of our algorithm improves at least 1.1dB with a minor increase of memory requirement.This provides an efficient way for hardware implementation of wavelet embedded block coding.