FISTA框架下基于构造导向空变中值滤波的含异常噪声地震数据重建

Simultaneous erratic noise attenuation and seismic data reconstruction based on structure-oriented space-varying median filter under the FISTA framework

受实际环境、采集成本等因素的制约,野外观测的地震数据经常出现不规则缺失的现象.为了给后续地震资料处理和成像提供完整的数据,势必要对缺失地震道进行插值重建.基于稀疏变换的插值方法被广泛用于地震数据重建,但传统的基于稀疏变换的插值方法对异常噪声非常敏感,当采集数据中存在异常噪声时会严重影响该方法的重建精度.为了克服这一问题,本文在快速迭代阈值收缩算法(FISTA)框架中引入构造导向中值滤波技术,提出了一种能同时压制异常噪声并对缺失地震道进行高精度重建的插值方法.该方法在FISTA的每次迭代中都对地震信号沿其局部倾角方向进行中值滤波以衰减异常噪声并最大限度地保护有效信号不被损害.得益于此,新方法可对含异常噪声的欠采样地震数据进行高精度重建.模型和实际数据测试均表明,当采集数据存在异常噪声时,新方法在插值精度和稳定性方面均优于传统稀疏促进插值方法.

Limited by the physical environment,acquisition cost,and other factors,field data always suffer from irregularly missing traces.In order to provide complete data for subsequent seismic data processing and imaging,it is inevitable to reconstruct the missing seismic traces.The interpolation method based on sparse transform is widely used for seismic data reconstruction,but the traditional interpolation method based on sparse transform is very sensitive to erratic noise,and the reconstruction accuracy of this method is seriously affected when erratic noise exists in the seismic data.To overcome this problem,we propose a interpolation method that simultaneously suppresses erratic noise and reconstructs missing seismic traces with high accuracy by introducing the structure-oriented median filter technique in the framework of the Fast Iterative Shrinkage-Thresholding Algorithm(FISTA).The method performs median filtering of the seismic signal along its local dip direction in each iteration of FISTA to attenuate the erratic noise and maximize the protection of the effective signal from damage.Benefit from this,the proposed method can perform high-precision reconstruction on under-sampled seismic data with erratic noise.Both model and field data tests show that the proposed method is superior to the conventional sparsity-promoting interpolation method in terms of interpolation accuracy and stability when erratic noise is present in the seismic data.