基于自相似块匹配的地震数据信噪分离方法研究

传统的基于三维曲波变换的随机噪声压制方法在同相轴不连续处会损害有效信号,存在伪影,难以取得理想的去噪效果,因而不能准确检测地震数据中地质体的边界信息。研究了基于自相似块匹配的三维地震资料去噪和边缘检测方法。自相似块匹配方法将含噪数据划分成相似数据块集合,同一数据块集合中的子块波形近似。该方法充分利用了三维地震数据的自相似性和冗余性信息,通过阈值界定相似程度将三维地震数据中的相似数据块聚集,将在常规域中难以压制的噪声变换到四维域中进行衰减,然后再进行逆变换得到去噪后的剖面。自相似块匹配方法有效提升了地震数据的信噪比和保真度,去噪结果无伪影。自相似块匹配Canny边缘检测结果表明,该方法很好地保存了不连续处的信息,能有效识别地质体边界信息。理论模型和实际资料的处理结果表明,该方法能有效压制三维地震资料中的随机噪声;与传统曲波变换阈值方法相比,该方法对有效信号的保护能力更强,能够保留原始数据中的不连续性信息,使基于自相似块匹配的Canny边缘检测方法能够实现对地质体边界的准确检测。

Seismic data denoising based on self-similarity block matching

Traditional random noise suppression methods based on three-dimensional curvelet transform may exhibit artifacts and damage the effective signals in the discontinuities such as faults,and they often have difficulties in denoising and detecting the boundary information of geological bodies from the seismic data.We therefore propose a denoising approach based on self-similarity block matching.This method divides the noisy data into a series of data blocks and the sub-blocks on adjacent in-phase axes are waveform-approximate.The proposed method makes full use of self-similarity and redundancy information of the 3D seismic data when denoising.In the method,the similar blocks in 3D seismic data were gathered into the same matrixes through block matching using threshold defining similarity.Then,the noise was attenuated in a 4D transform domain.Finally,the denoised section was obtained by the inverse transformation.Test results showed that the SNR and fidelity of seismic data were improved,and no artifacts were generated with the proposed denoising method.The Canny edge detection based on self-similar block matching can better detect the boundaries of geological bodies,due to better preservation of discontinuities in data using the proposed denoising method.