Seismic random noise removal by delay-compensation time-frequency peak filtering
Seismic random noise removal by delay-compensation time-frequency peak filtering
复制标题
DOI：
10.1088/1742-2140/aa6495
复制
发表时间：
2017-06
影响因子：
1.4
通讯作者：
Pengjun Yu;Yue Li;Hongbo Lin;N. Wu
中科院分区：
文献类型：
--
作者：
Pengjun Yu;Yue Li;Hongbo Lin;N. Wu
在过去的十年中，由于时频峰值滤波（TFPF）在抑制非平稳和强地震随机噪声方面的突出性能，它越来越受到人们的关注。传统的基于时间加窗的方法实现了局部线性，满足无偏估计。然而，传统的TFPF算法（包括变窗长的改进算法）很难缓解去噪与恢复地震信号之间的矛盾，这种情况在波峰和波谷处更为明显，即使是变窗长（WL）算法也是如此。为了提高算法的效率，在时间-空间域中，如Radon域和径向道域中应用了以下TFPF。时-空变换获得降低的频率输入以减小TFPF误差并且沿某一方向沿着拉伸期望信号，因此时-空展开通过增强反射同相轴和衰减噪声带来改善。但由于其方向应匹配为直线或二次曲线，因此在实际应用中仍然受到限制。因此，在处理复杂地层记录时，可能出现波形畸变和假同相轴。本文的重点是时空TFPF的应用扩展。延迟补偿TFPF中的重建信号是根据反射同相轴之间的相似性生成的，克服了方向曲线拟合的局限性。而且，重构信号恰好满足TFPF线性无偏估计，并将信号保留和噪声衰减结合起来。通过对合成模型和实测数据的仿真实验表明，时延补偿TFPF滤波算法比传统滤波算法具有更好的性能。
Over the past decade, there has been an increasing awareness of time-frequency peak filtering (TFPF) due to its outstanding performance in suppressing non-stationary and strong seismic random noise. The traditional approach based on time-windowing achieves local linearity and meets the unbiased estimation. However, the traditional TFPF (including the improved algorithms with alterable window lengths) could hardly relieve the contradiction between removing noise and recovering the seismic signal, and this situation is more obvious in wave crests and troughs, even for alterable window lengths (WL). To improve the efficiency of the algorithm, the following TFPF in the time–space domain is applied, such as in the Radon domain and radial trace domain. The time–space transforms obtain a reduced-frequency input to reduce the TFPF error and stretch the desired signal along a certain direction, therefore the time–space development brings an improvement by both enhancing reflection events and attenuating noise. It still proves limited in application because the direction should be matched as a straight line or quadratic curve. As a result, waveform distortion and false seismic events may appear when processing the complex stratum record. The main emphasis in this article is placed on the time–space TFPF applicable expansion. The reconstructed signal in delay-compensation TFPF, which is generated according to the similarity among the reflection events, overcomes the limitation of the direction curve fitting. Moreover, the reconstructed signal just meets the TFPF linearity unbiased estimation and integrates signal reservation with noise attenuation. Experiments on both the synthetic model and field data indicate that delay-compensation TFPF has a better performance over the conventional filtering algorithms.