地震图像随机噪声的非局部均值去噪法  

Denoising algorithm of random noise with seismic image based on nonlocal means

作　　者：黄英[1] 文晓涛[1] 贺振华[1] 

机构地区：[1]成都理工大学地球物理学院,四川成都610059

出　　处：《断块油气田》2013年第6期730-732,共3页Fault-Block Oil & Gas Field

基　　金：国家自然科学基金项目"裂缝性储层地震识别机理及相应方法研究"(41174115);国家自然科学基金青年基金项目"基于复杂弱信号检测的礁滩相储层预测及油气检测技术研究"(40904034);国家科技重大专项课题"南方海相碳酸盐岩层系大中型油气田形成规律与勘探评价(二期)"(2011ZX05005-003)

摘　　要：随机噪声的存在往往会影响地震图像分析的准确度。为了提高图像分析质量,提出了一种基于非局部均值滤波(Non Local Means)抑制地震图像随机噪声的新方法。在对滤波像素点去噪时,该方法分配给每个相似像素点的权重不依赖于2个像素点的空间距离,而是依赖以该像素点为中心的图像子块与以当前像素点为中心的子块之间的相似性,且滤波参数h的选取对滤波效果起到至关重要的作用。结合实例,对地震资料进行了具体分析。结果表明,与传统方法(如中值滤波、高斯滤波)相比,采用非局部均值滤波方法合成地震记录和实际数据时,既能有效地抑制地震随机噪声,又能较好地保留地震同相轴陡变处或同相轴弯曲处的边缘细节信息,具有实用性和有效性。The presence of random noise tends to affect the accuracy of the image analysis. To improve the quality of seismic data, this paper puts up a new algorithm of random noise with seismic data based on nonlocal means. This algorithm denoises each sample or pixel within an image by utilizing other similar samples or pixels regardless of their spatial proximity, making the process nonlocal. Filtering parameter h is importance for denoising random noise. Combined with examples, the seismic data are analyzed. The results indicate that the tests with synthetic and real data sets demonstrate that the nonlocal means algorithm does not smear seismic energy across sharp discontinuities or curved events when compared to traditional methods such as median filter, Gaussian filter, which shows that the nonlocal means algorithm is a practical and effective method.

关 键 词：地震数据 图像处理 非局部均值滤波 随机噪声 

分 类 号：TE132[石油与天然气工程—油气勘探]