基于L2范数的相关叠加去噪方法

本文在研究传统相干算法的基础上,提出了一种利用邻域相关性的加权叠加压制地震随机噪声的方法,即利用L2范数估计邻域间的相似度,并以邻域间欧氏距离的高斯函数作为计算加权系数的策略,实现邻域点的相关叠加,从区域识别信号到局部保真压噪。模型及实际资料的处理计算结果表明:该方法能够有效消除随机干扰,且信号的保真度高,是低信噪比条件下资料处理的可行方法。

Denoising by correlative stacking via L2 norm

This paper concentrates on a method of correlative stacking to suppress seismic random noise.In examining the basis of the traditional coherence algorithm,we proposed a denoising method by weighted stacking via neighborhood correlation(WS-NC),namely using L2 norm which also called Euclidean distance to estimate the similarity between two neighborhoods.Then we apply the Gauss function on Euclid distance as the weighting coefficients when stacking neighborhood points.So the random noise can be identified from region and suppressed in the local,while the amplitude of effective reflected wave can be well maintained.Applications on both synthetic data and actual seismic data show that the new method eliminates the noise portion more efficiently and retains a greater amount of geologic data,and it may be an effective method for seismic data with low S/N ratio.