基于小波变换和SVM的图像压缩仿真研究  

Approach of Image Compression Based on Wavelet Transform and SVM

作　　者：赵楠楠[1] 孙红星[1] 徐心和[1] 

机构地区：[1]东北大学人工智能与机器人研究所,辽宁沈阳110004

出　　处：《系统仿真学报》2006年第11期3034-3037,共4页Journal of System Simulation

基　　金：国家教委博士点基金资助项目(20040145012)。

摘　　要：为了在较高的压缩比上获得很好的压缩性能,提出了一种基于小波变换和支持向量机(SVM)的图像压缩方法。压缩过程分三个步骤:首先对图像进行四级提升小波变换,这里采用提升格式是因为它比采用传统的Mallat算法的计算速度快;其次对变换后的小波系数用SPIHT的继承树进行重新排序;然后用回归支持向量机提取支持向量;最后对压缩后的数据进行算术编码。图像的解压缩过程是上面4个步骤的逆过程。实验结果表明,所提出的方法与常用的JPEG2000相比,当压缩比较高时有很好的性能。In order to achieve better compressive performance in higher compression ratio, a novel algorithm of image compression combining support vector machines （SVM） and wavelet transform was developed. This algorithm consists of the following steps： Firstly, an image was transformed by 4-level wavelet transform. Here, second-generation wavelet i.e. lifting scheme was used for its higher computational than traditional Mallat＇s algorithm. Secondly, the wavelet coefficients were rearranged by SPIHT tree structure. Third, the SVM extracted support vectors （SVs） of each tree. Finally, arithmetical coding was applied to code compressive data. The decompressing procedure is a reverse process corresponding to the steps above. The experimental results show that the proposed algorithm performs better than JPEG2000 in high compression ratio.

关 键 词：图像压缩 小波变换 提升格式 SVM 

分 类 号：TN919.81[电子电信—通信与信息系统]