一种基于模糊均差和小波变换的医学图像去噪方法  

A Method for Medical Image Denoising Based on Wavelet Transformation and Fuzzy Mean Error

作　　者：李均利[1] 侯艳芹[1] 魏平[2] 陈刚[1] 

机构地区：[1]宁波大学数字技术与应用软件研究所,宁波315211 [2]浙江大学计算机学院,杭州310027

出　　处：《中国生物医学工程学报》2007年第3期342-348,共7页Chinese Journal of Biomedical Engineering

基　　金：国家自然科学基金资助项目(60672072);浙江省自然科学基金资助项目(Y106505);宁波市科技攻关项目(2005B100016)。

摘　　要：小波阈值萎缩法能够有效地去除图像中的噪声,去噪阈值直接影响去噪的效果,而噪声标准差在去噪阈值的确定中起着至关重要的作用。针对医学图像的特点、基于寻找更合适的噪声标准差估计方法,本研究提出了一种新的利用模糊均差代替普通标准方差估计噪声标准差的方法。在各层小波分解的低频图像中利用模糊积分估计噪声标准差,然后确定每一层去噪阈值,进行图像去噪。试验结果表明,本研究算法在去除噪声的同时也较好地保持了图像的细节。Wavelet threshold shrinkage algorithm can effectively remove noise of images, in which denoising threshold directly affects results of denoising. The noise standard variance is an important factor in ascertainment of thresholds in wavelet denoising. According to characteristics of medical images, in order to find more suited algorithms for estimating the standard variance, this paper presented a novel algorithm to estimate noise standard variation by fuzzy mean error instead of variance. We utilized fuzzy integral to estimate the noise standard variation in approximate coefficients of every wavelet decomposes. Determined the denoising thresholds of every wavelet decompose layer, and wiped off the noise using the soft threshold function. The experiment results show that the method can efficiently wipe off the noise and keep details of images.

关 键 词：模糊均差 小波去噪 医学图像 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]