基于非局部滤波算法的真实相干激光雷达强度像去噪方法比较  

Comparison of Intensity Image Denoising Methods Based on Non-Local Filtering Algorithms for Real Coherent Ladar

作　　者：李琦[1] 夏志伟[1] 刘正君[1] 王骐[1] 

机构地区：[1]哈尔滨工业大学可调谐激光技术国家级重点实验室,黑龙江哈尔滨150080

出　　处：《光学学报》2013年第B12期68-74,共7页Acta Optica Sinica

摘　　要：基于非局部滤波的去噪算法较之传统的基于局部统计特性的算法可以获得更好的图像复原质量。由于相干激光雷达强度像主要受散斑噪声的影响，采用结合同态变换的同态非局部均值（H-NLM）算法、同态多级非局部均值（HM-NLM）、同态块非局部均值（HB-NLM）和同态块匹配三维滤波（H-BM3D）等基于非局部滤波的算法对真实相干激光雷达仿真强度像进行了去噪处理，并对去噪结果进行了对比分析。实验结果表明，H-BM3D算法对单帧强度像、HM-NLM对6帧平均强度像具有较好的复原效果，复原图像目标灰度分布均匀性好，边缘比较清晰，能够反映目标的几何结构特征。The denoising algorithms based on non-local filtering can obtain higher quality of restored images than the classical ones based on local statistic characteristics. Due to the influence of speckle noise on intensity images of coherent ladar, the non-local filtering algorithms are integrated with homomorphism transform in this paper, i.e., homomorphism non-local means （H-NLM） algorithms, homomorphism multilevel non-local means （HM-NLM） algorithms, homomorphism non-local means by blocks （HB-NLM） algorithms, and homomorphism block-matching and three-dimensional filtering （H-BM3D）. Real intensity images of coherent ladar are denoised by these algorithms, and the results are compared and analyzed. It is shown show that single frame intensity image denoised by H-BM3D and 6-frame averaging intensity image denoised by HM-NLM can achieve better restoration. The restored images have better gray uniformity in target area and better clarity of edges, and can reflect the geometry structure of target.

关 键 词：图像处理 相干激光雷达 图像去噪 非局部滤波 散斑噪声 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]