基于曲波变换-联合双边滤波的二维地震数据随机噪声去除方法研究

在地震数据去噪处理中, 曲波变换硬阈值法容易导致弱同相轴模糊, 为此, 提出了一种基于曲波变换-联合双边滤波的地震随机噪声去除方法。曲波系数尺度分析结果表明, 粗尺度信号主要表征地震数据构造特征; 中尺度信号通常包含大量的弱同相轴的信息; 细尺度信号通常包含噪声信息。曲波变换-联合双边滤波方法对粗尺度信号进行双边滤波处理, 对中间尺度信号进行联合双边滤波处理, 对细尺度信号应用硬阈值处理。曲波变换可以较好地处理多方向的线状变化特征, 而联合双边滤波具备引导图, 可以修复有效信息的关键特征, 提高弱同相轴的连续性。叠前和叠后数值模拟测试和实际地震数据处理结果表明, 曲波变换-联合双边滤波方法的去噪结果优于曲波变换硬阈值法和小波变换方法, 具有更高的信噪比和峰值信噪比。该方法克服了曲波变换硬阈值法的局限性, 不仅能够较好地去除随机噪声, 而且增强了弱同相轴的能量, 提高了弱同相轴的连续性。

The curvelet transform method for seismic data denoising with a hard threshold is prone to weak event blurring when used to denoise seismic data because of the discontinuity at the threshold. This study proposed a method for removing random noise based on curvelet transform-joint bilateral filtering. First, the curvelet transform was applied to the original seismic data, and the scale information of the original curvelet coefficients was analyzed. Coarse-scale signals primarily characterize seismic data structures. To process the coarse-scale curvelet coefficients, the proposed method used bilateral filtering. Intermediate-scale signals typically contain a large amount of weak event information. Joint bilateral filtering was performed on the intermediate-scale curvelet coefficients. The curvelet transform handles multi-directional linear changes more effectively, and joint bilateral filtering has a guide map that can repair the key features of effective information and improve the continuity of weak events. Fine-scale signals typically contain noise. The hard-thresholding method was applied to deal with fine-scale curvelet coefficients. The processed coarse-, intermediate-, and fine-scale curvelet coefficients were then used to compose new curvelet coefficient sets. The denoised seismic data was obtained by performing an inverse curvelet transform on the new curvelet coefficient sets. Simulated and real pre- and post-stack seismic data were used to verify the effectiveness of the proposed method. When compared with the curvelet transform method with a hard threshold and the wavelet transform method, the proposed method can remove random noise and obtain a higher signal-to-noise ratio and peak signal-to-noise ratio. The energy of weak events was enhanced, and their continuity was improved. The curvelet transform-joint bilateral filtering method can be extended to seismic data with low signal-to-noise ratios to provide high-quality seismic data to modeling and interpretation engineers.