基于旋转取反压缩码书的研究

A Study of the Code Book for Rotation Decompression

矢量量化是一种高压缩算法，码书设计在VQ技术中起着非常重要的作用．提出了一种基于旋转取反压缩的码书生成方法，该方法将1024尺寸码书压缩为128尺寸，分别采用全搜索（FS）和等和值块扩展最近邻搜索算法（EBNNS）对其仿真，结果表明，图像的峰值信噪比（PSNR）比1024尺寸码书略有提高．研究选取10幅标准测试图测试，与使用1024码书相比，全搜索的平均PSNR仅下降0．029％，EBNNS算法的平均PSNR提高了2．72％．但使用128尺寸码书可以减少87．5％码书存储空间，大大减少运算量和码书存储面积．

Vector quantization is a kind of compression algorithm, the design for the code book plays an important role in vector quantization technique. This paper discusses a code book generating method, based on. This method is to compress the 1024 sized code book into 128, by adopting full search and equal--sum block--extending nearest neighbor search algorithm （EBNNS） respectively for simulation. The result shows that peak signal to noise ratio （PSNR） of the picture is slightly higher than that of 1024 sized code book. 10 standard test charts are tested, compared with 128 sized code book, the average PSNR falls only by 0. 029M, while the average PSNR based on EBNNS increases by 2. 72%, however, the 128 sized code book can decrease by 87. 5M of its storage space, greatly reducing the operand and storage space.