一种高效存储多级二维9/7离散小波变换结构
A Multi？？Level 2-D 9/7 DWT Architecture with Efficient Memory

针对JPEG2000图像压缩编码硬件实现中多级离散小波变换模块(DWT)与优化截断嵌入式模块(EBCOT)之间大量小波系数存储带来的缓存问题,提出了一种新型高效存储的多级二维9/7离散小波变换结构。首先,基于9/7小波变换提升算法,设计了可满足并行结构要求的基本处理单元,并在此基础上搭建了二维变换模块;然后采用一种新型的数据扫描方式,通过对图像分块并进行块间组合扫描,大幅降低了DWT模块与EBCOT模块间的小波系数缓存;最后根据数据扫描特点完成了非折叠结构多级二维小波变换模块的构建。仿真实验结果表明,三级9/7离散小波变换结构处理一幅大小为512×512像素的图像,对比已有硬件结构可节约存储资源40％以上。
A novel multi？？level two？？dimensional 9/7 discrete wavelet transform (DWT) architecture with efficient memory is proposed to solve the problem that a large number of wavelet coefficients between the multi？？level 2-D 9/7 DWT (2-D 9/7 DWT) module and the embedded block coding with optimized truncation (EBCOT) module need to be cached in the hardware implementation of JPEG2000 image compression and to decrease the cost of cache. A parallelizable basic processing unit is firstly designed based on the 9/7 DWT lifting scheme, and then a 2-D transform module is built. A novel block？？based data scanning method is then adopted to scan combined blocks of an image so that the storage of wavelet coefficients between the modules is dramatically reduced. Finally, a unfolded multi？？level 2-D DWT module is constructed according to the characteristics of data scanning. Experimental results and comparisons with existing hardware structures show that the proposed 3？？level 2-D 9/7 DWT architecture saves at least 40％ of storage resources for an image with the size of 512×512 pixels