一种基于细尺度间小波系数相关性的图像去噪方法  

Image Denoise Algorithm Based on Inter Correlation of Wavelet Coefficients at Finer Scales

作　　者：傅博[1] 王相海[1,2] 

机构地区：[1]辽宁师范大学计算机与信息技术学院,大连116029 [2]南京大学计算机软件新技术国家重点实验室,南京210093

出　　处：《计算机科学》2008年第10期246-249,共4页Computer Science

基　　金：辽宁省高等学校优秀人才支持计划(RC-04-11);辽宁省自然科学基金(20072156);辽宁省教育厅科学技术研究项目(20060486);南京邮电学院图像处理与图像通信江苏省重点实验室开放基金(ZK207006)资助

摘　　要：图像去噪问题的研究一直是图像处理的热点问题。首先对噪声图像经小波分解后噪声小波系数在细尺度子带间的分布特点进行了研究,提出了一种结合尺度内和尺度间系数相关性的噪声统计模型——细尺度间噪声系数分布的"类零树结构",以及基于分块的Bayes阈值确定方法。在此基础上将二者相结合,提出了一种新的图像去噪方法。该方法首先通过Bayes阈值去噪法去除高层子带中的噪声,同时利用基于块阈值方法定位次外层子带中的噪声位置,然后利用"类零树结构"模型,估计对应的最外层子带中的噪声的分布,并进行相应的去噪处理。实验结果表明,该方法稳定、有效,去噪效果优于传统Bayes逐点阈值去噪方法,且具有较低的时间复杂度。linage denoise has been being a hotspot in the discipline of image processing. We first analyse the inter-scale distribution characteristics of the coefficients at finer scale of the noise image, and then we propose a statistical model of noise incorprating the coefficient correlation of intra-scale and inter-scale, called the zerotree-like structure of noise coef- ficient distribution inter finer-scale, and the calculational method of block-wise Bayes shrinkage threshold. Second, based on the above, a new denoise algorithm is proposed. The algorithm first filters the noise at finer-sale using Bayes shrinkage threshold,and at the same time locates the noise in the lowest level but one of wavelet pyramid by the proposed block-wise Bayes shrinkage threhold. Then using the model of zerotree-like structure, the algorithm estimates the corresponding noise distribution in the bands of the lowest level, and do the appropriate denoise process. The experiment results show that the proposed algorithm is steady and efficient, and its denoising performance is superior to the traditional point-wise Bayes shrinkage method. Moreover, the proposed algorithm has a lower time complexity.

关 键 词：图像去噪 噪声小波系数分布 Bayes阈值法 分块阈值 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] P631.443[自动化与计算机技术—计算机科学与技术]