基于自相似性和小波分析的图像增强与去噪  

Image Denoise and Enhancement Based on Structure Self-similarity in Wavelet Domain

作　　者：焦峰[1] 毕硕本[1] 赵英男[1] 耿焕同[1] 

机构地区：[1]南京信息工程大学计算机与软件学院,南京210044

出　　处：《计算机科学》2010年第10期242-245,共4页Computer Science

基　　金：国家自然科学基金项目(40471101;60702076);中国博士后基金项目(20080431114);南京信息工程大学科研基金项目(20080304;20070113)资助

摘　　要：图像的去噪是图像质量提高问题中的一个重要分支。由于噪声的种类很多,性质和表现各不相同,因此很难确定哪一种去噪算法针对当前的图像是最好的。介绍了一种针对随机噪声、利用图像的自相似性的小波域的双边滤波去噪算法。算法不仅利用了小波的时频定位和多分辨率分解特性,也利用了图像的自相似性的特点。去噪的过程分为3个步骤。首先利用图像的自相似性将其分解成多个相似的、分辨率较低的子图,并重新组合(分解变换)。然后进行小波分解,并针对小波系数进行滤波。最后用调整后的小波系数进行逆小波变换和逆分解变换,生成去除噪声和增强后的结果图。实验结果证明,算法不仅能够有效地实现去噪,还使图像的边缘和对比度也得到了增强。Image denoise is a very important problem in image quality enhancement.For the image noises are different from each other,it is difficult to select one best denoise algorithm for all images which are degraded by different kinds of noises.In the paper,a novel algorithm based on wavelet transform and structure self-similarity was introduced,which is effective to confine the random noise.Beside the properties of time frequency localization and multi-resolution of wavelets,the self-similarity of image were used.It first separated the image into several blocks equally based on the self-similarity.And then,wavelet transform coefficients were used to enhance edge information.At the same time,image noises were confined.Experiment results show the excellent effect of the algorithm.

关 键 词：图像去噪 图像增强 自相似性 小波变换 DAUBECHIES小波 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]