基于BEMD与双树复小波-局部Wiener滤波的毫米波图像去噪  

Millimeter wave image de-noising based on BEMD and dual-tree complex wavelet transform integrated with local Wiener filtering

作　　者：戴桂平[1] 林洪彬[2] 

机构地区：[1]苏州市职业大学电子信息工程系,江苏苏州215104 [2]燕山大学电气工程学院,河北秦皇岛066004

出　　处：《机械设计》2011年第9期12-16,共5页Journal of Machine Design

基　　金：国家自然科学基金资助项目(60970058);河北省科学技术研究与发展计划资助项目(10212152)

摘　　要：为了有效地去除毫米波图像中含有的噪声,提高目标识别的精度,提出一种将二维经验模式分解(BEMD,Bidimensional Empirical Mode Decomposition)与基于双树复小波变换(DTCWT)的加窗局部Wiener滤波相结合的图像去噪算法。首先,对毫米波图像进行BEMD分解,得到不同特征尺度的本征模函数(IMF,Intrinsic Mode Function)子图像集;其次,利用双树复小波变换对中高频IMF子图像进行多尺度、多方向分解,并结合带有椭圆方向窗的局部Wiener滤波算法对各个高频方向子带进行去噪;最后通过DTCWT逆变换重构得到去噪后的IMF,并与残差图像相加进行BEMD重构。实验结果表明,该融合算法与单独的BEMD,DTCWT-Wiener滤波及离散小波变换-Wiener滤波算法相比,去噪后图像的视觉效果更好,提取的目标的边缘及细节特征更清晰,因而峰值信噪比最高。In order to remove the millimeter wave image noise effectively and improve the accuracy of target identification,a novel image de-noising method based on BEMD（Bi-dimensional Empirical Mode Decomposition） and Dual-tree Complex Wavelet Transform（DTCWT） integrated with Wiener filter is proposed.Firstly,BEMD was carried out to decompose the millimeter wave image including noise into a group of Intrinsic Mode Functions（IMF） sub images with different intrinsic time scales.Secondly,the first several IMF sub images corresponding to high frequency information and noise were decomposed with different scales and directions by means of dual-tree complex wavelet transform,and then the Wiener filter with elliptic direction windows was used to de-noise each high-frequency direction sub-band.Finally,the image was reconstructed through adding the processed IMF and the residual component.Simulation results show that,compared with the single BEMD,DTCWT-Wiener filter and discrete wavelet transform-Wiener filter,this proposed method can not only have advantages of more sufficiently retaining edge and detail information while de-noising,but also achieve superior PSNR（peak signal-noise-ratio） of the reconstructed image.

关 键 词：二维经验模式分解 双树复小波变换 局部Wiener滤波 毫米波图像去噪 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]