改进的矢量分解压噪方法

An improved vector resolution noise removal approach

矢量分解压噪方法利用噪声偏离有效信号的夹角压制噪声,是一种经典的角度域去噪技术。文中在简要分析矢量分解压噪方法及其现有改进措施的基础上,针对该方法的理论误差来源和算法特点,提出了应用"曲边时窗"进一步改进的技术思路和方案。实施的关键在于根据时窗内各道与时窗中间道达到最大相关时的最佳时移值,建立在纵向上按道分割、在横向上按最佳时移值进行曲线分割的曲边时窗。利用数值模拟和实际数据对比了方法改进前、后的去噪效果,结果表明,应用曲边时窗的矢量分解压噪方法在提高去噪效果的同时,能更好地保护有效信号。针对实际数据设计了曲边时窗矢量分解压噪方法与典型频率域去噪方法(带通滤波、小波阈值去噪)的对比实验,去噪结果表明,对于噪声和有效信号时频特征接近的地震数据,利用本文提出的方法从角度域入手往往能取得更好的去噪效果。

Vector resolution,a typical angle-domain denoising approach,uses the deviation angle between signal and noise to remove noise.We review briefly the vector resolution method and existing improved measures,and propose a new measure which uses curved time-window scheme according to its errors.The key point is to build a curved time-window based on the best time-shift values for maximum correlation coefficients between each trace and the center trace in the time window.Numerical simulation and real data de-noising test results show that the proposed method can better remove noise.We also perform real data tests with the proposed approach and frequency-domain denoising methods such as pass-band filtering and wavelet threshold de-noising.The proposed approach gets better results event for seismic data in which signal and noise have very close frequency.