结合近似最优比特分配的改进SPIHT算法  

Improved SPIHT Algorithm Combined with Nearly Optimal Bit Allocation

作　　者：潘志刚[1] 张文超[1] 王晓晨[2] 

机构地区：[1]中国科学院电子学研究所,北京100080 [2]合肥工业大学电气与自动化工程学院,合肥230009

出　　处：《计算机工程》2007年第15期46-48,共3页Computer Engineering

摘　　要：SPIHT算法是一种实用、高效的小波零树图像编码算法。针对SPIHT算法存储空间需求大、运算复杂度较高等缺点,提出了一种改进的快速、低存储SPIHT算法,该算法将小波变换所形成的水平、垂直、对角和低频4个子带分成4个处理单元,对每个处理单元分别进行量化编码,并在各单元之间采取近似最优比特分配以提高量化性能。实验结果表明,改进算法在提高峰值信噪比等性能指标的同时,有效地减少了算法的存储需求及运算时间。SPIHT（set partitioning in hierarchical tree） algorithm is a wavelet based zerotree image encoding algorithm known for its high efficiency. However, its high memory requirement and long execute time consumption are obstacles to implement real-time compression. This paper presents a new fast and low memory image zerotree encoding algorithm. This algorithm processes wavelet coefficients of horizontal, vertical, diagonal and low frequency subband respectively, and a nearly optimal bit allocation stage is applied in the above subband to get better compression performance. Experimental results show this algorithm outperforms the original SPIHT algorithm in PSNR and efficiently reduces both the memory requirement and the time consumption.

关 键 词：图像编码 小波变换 零树量化 SPIHT 比特分配 

分 类 号：TP919.81[自动化与计算机技术]