一种小波去噪方法的几点改进  

Several improvements for a wavelet denoising method

作　　者：谢杰成[1] 张大力[1] 徐文立[1] 

机构地区：[1]清华大学自动化系,北京100084

出　　处：《清华大学学报（自然科学版）》2002年第9期1269-1272,共4页Journal of Tsinghua University(Science and Technology)

基　　金：清华大学"九八五"资助项目

摘　　要：Mihcak M.K,IEEE Signal Processing L etters,1996 ,6 (12 ) :30 0 30 3提出了小波去噪方法 L AWML(L AWMAP) ,并得到了非常好的去噪效果。该文针对其过保留小系数的特点提出了引入阈值处理以减少重建毛刺和提高去噪后信号的压缩率 ;根据小波系数子带能量分布的特点 ,提出对不同尺度不同子带选择不同邻域来提高方差估计的准确性 ;根据子带的方向选择性 ,提出用长方形代替正方形邻域的改进方案。通过实验证明了这些改进可以有效地弥补原先 L AWML (L AWMAP)方法的不足 ,得到更好的去噪效果.The recently proposed wavelet denoising method, locally adaptive window -based denoising using maximum likelihood (LAWML), can be further improved using three strategies proposed here. First, a threshold is used to eliminate small coefficients, which leads to smoother reconstruction and higher compression. Secondly, the size of the neighbor window was varied based on the energy distribution in different levels and different classes of subbands to provide more accurate estimates of the variance. Finally, a rectangular neighbor window was used instead of the square window. Experiments showed that using these strategies, the modified LAWML provided better performance than the original.

关 键 词：图像信息处理 阈值处理 LAWMAP方法 小波去噪 小波萎缩 小波变换 重建误差 压缩率 

分 类 号：TN911.73[电子电信—通信与信息系统]