一种基于能量分布特性的小波去噪算法  

A New De-Noising Algorithm of Wavelet Based on Characteristic of Energy Distribution

作　　者：王忠华[1] 汪胜前[2] 

机构地区：[1]南昌航空工业学院电子工程系 [2]江西科技师范学院应用物理系,江西南昌330013

出　　处：《江西师范大学学报（自然科学版）》2005年第3期251-254,共4页Journal of Jiangxi Normal University(Natural Science Edition)

基　　金：江西省自然科学基金资助项目(0412008).

摘　　要：目前对于保持图像细节、滤除噪声,普遍采用空间域、频率域滤波.在空间域滤波,尽管能够有效地限制噪声,但是同时模糊了图像细节.因此,在频率域滤波的方法越来越引起关注.在小波频率域中,我们常常采用Donoho阈值方法处理小波系数来以此去除噪声,保留图像细节,然而该方法同时也一定程度上模糊了图像细节.小波变换具有良好的时、频局部化性能,图像经过多级小波变换得到不同分辨率的子图个数,各高频子图上的小波系数具有相似的能量统计分布特性.也就是说随着分解层数的增加,分辨率最低子图的小波系数范围最大,而高分辨率子图上大部分数值接近于0.因此,该文提出了一种新的基于能量分布特性的小波去噪算法(WCED).<Abstrcat> For keeping image detail and constraining image noise, traditional filters are mostly those in space domain or frequency domain. In space domain, we can more effectively constrain noise while it blurs image details. So, the filters in frequency domain have attracted more and more attention. In wavelet frequency domain, image frequency can be effectively decomposed and then noise can be restricted. Traditionally, we make use of Donoho's threshold to de-noise and preserve image details with regard to wavelet coefficients. However, which also results in blur image details etc. It is known that wavelet transformation has good performance in local time domain and frequency domain. Sub-images are acquired by multilevel wavelet transformations, then we can find that wavelet coefficients own similarity of energy distribution in high frequency sub-images, that is to say, wavetlet coefficients distribute much wider through increasing scale of decomposition. Sub-images of lower resolution whose wavelet coefficients own wider range, sub-images of higher resolution whose wavelet coefficients own narrower range. Therefore, we present a new wavelet de-noising algorithm based on characteristic of energy distribution.

关 键 词：去噪算法 能量 频率域滤波 图像细节 小波系数 统计分布特性 小波变换 高分辨率 空间域 子图 噪声 局部化 阈值 接近 数值 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TN911.7[自动化与计算机技术—计算机科学与技术]