基于小波变换系数去冗余的图像无损压缩方法
An algorithm of lossless image compression based on wavelet transform coefficients to eliminate redundancy

本文对基于DPCM与整数小波相结合的算法在无损图像压缩中的应用进行了深入研究.研究表明, 分块图像数据经过整数小波变换后仍存在冗余, 对这些冗余的数据进行分析, 可通过DPCM编码, 对变换后的数据做进一步的压缩.由于5/3整数小波变换的特性, 其冗余信息主要分布在水平、垂直和对角线的方向.因此, 本文在大量实验数据的基础上, 提出了一种按照冗余信息的分布规律, 构造出预测模板, 再对变换后的不同的子带系数选择相应的、开销最小的模板进行残差系数冗余信息去除的方法.实验结果表明, 所提出方法的残差系数的平均码长比原始图像的平均码长减小了2.61, 总压缩率比JPEG 2000压缩率平均提升了4.63%, 相比原始图像直接进行DPCM编码压缩率平均提升了6.16%, 同时因其算法简单, 非常适合于硬件FPGA上的实现.
This paper conducts an in depth research in the applications of lossless image compression based on the combination of DPCM and integer wavelet algorithm. The former research shows that redundancy still exists in the residual coefficients after image block integer wavelet transform (IWT). Under this phenomenon, the transformed data can be compressed further if an analysis of these redundant data is carried out before the differential pulse code modulation (DPCM) coding. The redundant information is mainly distributed in the horizontal, vertical and diagonal directions because of 5/3 integer wavelet transform’s features. As a consequence, this paper proposes an algorithm of the removal of residual coefficient’s redundant information, on the basis of a large amount of experimental data. The algorithm includes constructing different prediction templates in accordance with redundant information’s distribution law, choosing the corresponding and least cost template according to the transformed sub band coefficients, removing redundant information in the residual coefficients, etc. The experimental results show that the proposed algorithm decreases the average code length of the residual coefficient at an average of 2.61 compared with that of the original image. What’s more, the compression ratio increases by 4.63% and 6.16% on average, respectively compared with the JPEG 2000 and the original image directly compressed by DPCM coding. Also it is very suitable for hardware implementation