基于非下采样四元数轮廓波变换域的图像去噪  

Image denoising based on nonsubsampled quaternion contourlet transform domain

作　　者：殷明[1] 吴江敏[1] 于立萍[1] 孔冉冉[1] 

机构地区：[1]合肥工业大学数学学院,合肥230009

出　　处：《计算机应用研究》2014年第12期3902-3906,共5页Application Research of Computers

基　　金：安徽省自然科学基金资助项目(1308085MA09);安徽省教育厅基金资助项目(2013AJZR0039)

摘　　要：为了克服传统小波变换的不足,提出一种将非下采样四元数轮廓波域对称正态逆高斯分布模型与非局部均值滤波相结合的图像去噪新方法。非下采样四元数轮廓波变换是级联四元数小波变换和非下采样方向滤波器组得到的,具有高度的方向选择性、各向异性及平移不变性。仿真实验表明:该方法在峰值信噪比、平均结构相似性及视觉效果上均优于一些经典的去噪算法。In order to overcome the shortages of the traditional wavelet transform,this paper presented new image denoising algorithm based on the combination of symmetrical normal inverse Gaussian model in nonsubsampled quaternion contourlet transform（ NSQCT） domain and non-local means filter（ NLMF） in spatial domain. It constructed NSQCT by combining the quaternion wavelet transform（ QWT） and nonsubsampled directional filter banks（ NSDFB）,which had highly directional selectivity,multi-direction and approximate shift invariant. Simulation experiment shows that the algorithm can obtain better performances than those classical denoising algorithms in terms of peak signal-to-noise ratio（ PSNR）,mean structural similarity（ MSSIM） as well as visual quality.

关 键 词：图像去噪 非下采样四元数轮廓波 对称正态逆高斯分布模型 非局部均值滤波 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]