第二代曲波变换的图像降噪新算法  

New image denoising algorithm based on the second generation curvelet transform

作　　者：都伊林[1] 白直灿[2] 

机构地区：[1]浙江警官职业学院,杭州310009 [2]浙江树人大学信息科技学院,杭州310015

出　　处：《声学技术》2010年第3期331-335,共5页Technical Acoustics

摘　　要：Curvelet变换表示曲线奇异函数的异向性及图像边缘时,具有比小波变换更优的表示特性。针对小波图像降噪存在的不足,分析基于wrapping算法的快速离散曲波变换的特点,提出结合循环平移、厄尔迭代方法和蒙特卡洛阈值规则的新消噪方法。该算法充分利用曲波系数的相关性,消除了因Curvelet变换缺乏平移不变性引起的图像"划痕"失真和"振铃"效应。实验结果表明,该算法与传统的小波消噪、二代小波消噪、小波包消噪和曲波硬阈值消噪相比,得到降噪图像的峰值信噪比更高,视觉效果更好。Compared to the wavelet transform the curvelet transform is more superior in expressing curve singular func-tion anisotropy and image edge.In view of the insufficiency of wavelet image noise reduction,the characteristics of the fast discrete curvelet transform based on the wrapping algorithm is analyzed,and a new denoising method combining with cycle shift,Ell iterative and Monte Carlo methods’ threshold rule is proposed.This algorithm fully uses the corre-lation of curvelet coefficient to eliminate the scratch distortion and the ringing effect caused by lack of image translation invariability of the Curvelet transform.Experimental results show that the algorithm yields denoised images with higher PSNR and better visual effects in comparison with traditional wavelet denoising,second-generation wavelet denoising,wavelet packet denoising and hard threshold denosing based on curvelet transform.

关 键 词：快速离散曲波变换 图像降噪 wrapping算法 循环平移 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]