基于对偶树复小波域图像融合的SAR图像去噪  

SAR Image De-Noising Algorithm Based on Dual-Tree Complex Wavelet Domain Image Fusion

作　　者：高佳城[1] 林岳松[1] 陈华杰[1] 

机构地区：[1]杭州电子科技大学信息与控制研究所,浙江杭州310037

出　　处：《计算机仿真》2009年第7期232-235,共4页Computer Simulation

基　　金：国家重点项目(60434020);十一五国防预研基金项目

摘　　要：边缘信息是图像重要的细节信息,保护图像的边缘信息对提高图像质量非常重要。但是在图像去噪的过程中,往往会破坏图像的边缘信息。针对去除噪声和保护边缘信息的双重考虑,提出一种基于对偶树复小波域图像融合的SAR图像阈值去噪。考虑到局部硬阈值和软阈值各自的特点,利用对偶树复小波变换的优点和图像融合的特点,首先在自然对数域对SAR图像进行对偶树复小波分解,然后对小波系数分别执行局部硬阈值去噪和局部软阈值去噪,最后依次通过图像融合,对偶树复小波反变换,指数变换得到去噪以后的图像。实验结果表明,算法融合了两种阈值去噪方法的优点,在明显去噪的同时,更好地保护了图像的边缘信息。Edge information is the important detail information of image, and protecting the edge information is very important for improving image quality. Unfortunately, image de - noising is often followed by destroying the edge information. To remove noise and protect edge information, a new threshold de - noising algorithm combining the local hard - thresholding with soft - thresholding, based on dual tree complex wavelet transform （ DT - CWT） and image fusion is presented. Considering the characteristic of the local hard - thresholding and soft - thresholding and making use of the advantages of DT - CWT and image fusion, the first step is to carry out natural logarithm transform which is followed by DT - CWT. After DT - CWT, the coefficients are processed by local hard - threshotding and soft -thresholding, respectively. At last, the noise free image is obtained by implementing image fusion, dual tree com- plex wavelet inverse transform and exponent transform. Experimental results show that this algorithm can fuse the ad- vantages of the two methods and provide significant improvement over conventional image de - noising methods in terms of the ability of de - noising and protecting edge information.

关 键 词：保护边缘 图像融合 对偶树复小波 硬阈值去噪 软阈值去噪 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]