基于椭圆方向窗的非下采样Contourlet域图像去噪算法.pdfVIP

第41卷第4期 河 海 大 学 学 报 ( 自然 科 学 版 ) Vol.41No.4 2013年7月 Journal of Hohai University(Natural Sciences) Jul. 2013 DOI:10.3876/ j.issn.1000 1980.2013.04.016 基于椭圆方向窗的非下采样Contourlet域图像去噪算法 金彩虹 (南京晓庄学院物理与电子工程学院,江苏 南京摇 211171) 摘要:利用非下采样Contourlet 变换的平移不变性和多方向选择性,考虑非下采样Contourlet 域内 子带能量聚集的各异性,在不同尺度、不同方向的子带内,采用不同的椭圆方向窗对窗中心系数进 行基于NeighShrink算法的估计。 估计时各子带所用椭圆方向窗的主轴方向与图像能量聚集方向 一致,椭圆方向窗的长、短轴之比与该尺度下非下采样Contourlet基函数的支撑区间相匹配,椭圆方 向窗的最佳大小和该窗内的最佳阈值通过最小化Stein 的无偏风险估计获得。 试验结果表明,用该 椭圆方向窗在能量聚集处能够实现对高频子带系数的最佳估计,在去除噪声的同时尽可能多地保 留图像的边缘细节,提高恢复图像的信噪比。 关键词:非下采样Contourlet 变换;椭圆方向窗;方向相关性;Stein无偏风险估计 中图分类号:TN911郾73摇 摇 摇 文献标志码:A摇 摇 摇 文章编号:1000 1980(2013)04 0371 06 An image de鄄noising algorithm based on elliptic directional windows in non鄄subsampled contourlet domain JIN Caihong (School of Physics and Electronic Engineering,Nanjing Xiaozhuang University,Nanjing 211171,China) Abstract: Based on translation invariance and multi鄄directional selectivity of the non鄄subsampled contourlet transform and anisotropic energy gathering in sub鄄bands of non鄄subsampled contourlet domains,different elliptic directional windows were used to estimate the central coefficients in sub鄄bands of different scales and directions using the NeighShrink algorithm. During the estimation process,the major axis direction of the elliptic directional window was kept consistent with that of images爷 energy gathering,the long axis鄄to鄄short axis ratio of the elliptic directional window matched the supporting domains of the non鄄subsampled contourlet function at the same scale, and the minimized Stein爷s unbiased risk estimate (SURE)wasused to obtain the most suitable size of the elliptic directional window aswellasitsoptimalthresh