基于Curvelet变换和全变差的图像去噪方法  

Image Denoising Method Based on Curvelet Transform and Total Variation

作　　者：倪雪[1] 李庆武[1] 孟凡[1] 石丹[1] 范新南[1] 

机构地区：[1]河海大学计算机及信息工程学院,江苏常州213022

出　　处：《光学学报》2009年第9期2390-2394,共5页Acta Optica Sinica

基　　金：国家863计划(2007AA11Z227);江苏省社会发展科技项目(BS2007058)资助课题

摘　　要：Curvelet变换用于图像去噪可以较好地保留图像的细节信息,但在边缘处会产生"划痕"现象。采用全变差法进行去噪能保持边缘形状不变,但也会丢失图像的纹理等细节信息。为了充分利用两种方法的优点,将Curvelet变换和全变差相结合提出了一种有效的图像去噪方法。首先,对含噪图像分别进行Curvelet阈值去噪和全变差去噪。然后,将两幅去噪图像进行Curvelet融合,对于低频系数和高频系数分别采用加权平均和绝对值取大的融合算法。最后,将融合后的低频系数和各尺度高频系数进行Curvelet反变换得到融合后的去噪图像。实验表明,该方法能有效地降低图像噪声,又尽可能地保留图像的细节,其去噪效果明显优于单一Curvelet阈值法和全变差法。Curvelet transform can preserve more details for image denoising, but it always has the ‘warp-around＇ artifacts in image edges. Total variation, another effective image denoising method, can preserve edges better, but image texture information will be also smoothed. An efficient image denoising method based on combination of curvelet transform and total variation is proposed. Firstly, the image is denoised by curvelet thresholding method and total variation method. Then, the two denoised images are fused using curvelet transform. Here the weighted average algorithm and maximizing absolute value algorithm are used respectively to process the low-frequency coefficients and the high-frequency coefficients. Finally, the denoised image is reconstructed by the inverse curvelet transform. Experimental results show that the new method is effective in removing white noise, and the detail of the image is kept well. It has better denoising effect than single curvelet thresholding method and total variation method.

关 键 词：图像处理 CURVELET变换 全变差 图像去噪 图像融合 

分 类 号：TN911.73[电子电信—通信与信息系统]