一种无链表零树熵编码算法及其FPGA实现  

Listless Zerotree Entropy Coding Algorithm and Its FPGA Implementation

作　　者：梁春锋[1] 唐艳发[2] 孙红亮[2] 张利[1] 

机构地区：[1]清华大学电子工程系,北京100084 [2]清华大学微电子所,北京100084

出　　处：《计算机工程》2012年第4期202-204,211,共4页Computer Engineering

基　　金：国家自然科学基金资助项目(60972022)

摘　　要：传统熵编码算法复杂度高,且硬件实现较难。为此,提出一种易于FPGA硬件实现的无链表零树图像编码改进算法,该算法结合LZC和SPIHT算法的特点,增加了对孙子节点的判断,克服SPIHT算法中内存消耗大、复杂度高以及LZC算法中图像重建质量下降等缺点。采用该算法的编码器在对任意尺寸图像进行5层小波变换时,所需缓存固定为1 536 bit,远小于SPIHT算法。在相同比特率的情况下,图像重建质量PSRN值比LZC算法平均提高约0.7 dB。在Xilinx Virtex4 LX100硬件平台上进行实验,结果表明,当时钟频率为100 MHz、压缩率为1/40时,该算法处理1 600×1 200×24 bit图像的速度可达20 f/s。Since the traditional entropy coding algorithm is hard to realize using hardware method and with high complexity, a Listless Zerotree image Coding（LZC） algorithm is proposed and implemented in Field Programmable Gate Array（FPGA） platform. The algorithm adopts the features from LZC and Set Partitioning in Hierarchical Trees（SPIHT）, and adds the judgment for grandchildren nodes. With the improvement, the algorithm overcomes the high memory overhead and high complexity of SPIHT, and enhances the coding quality of LZC. For any size of images, the memory overhead for this coder implemented in FPGA is fixed, 1 536 bit, far less than that of SPIHT. The PSNR of decoded image is enhanced about 0.7 dB than LZC under the same compression bitrate. When the algorithm is implemented in Xilinx Virtex4 LX100, running at 100 MHz, it performs a frame rate of 20 f/s for 1 600x 1 200x24 bit images with a compression ratio of 1/40.

关 键 词：熵编码 零树 FPGA实现 图像压缩 小波变换 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]