局部奇异值分解法压制随机干扰

对于低信噪比资料，压制随机噪声，增强有效信号是地震资料处理的首要任务。而传统的奇异值分解去噪算法，在有效信号横向相干性较强时，去噪效果明显，但当有效信号同相轴呈倾斜、弯曲或孤立状态时，其在压制随机噪声的同时，存在滤除部分有效信号的弊端，为此通过对不同时窗内的地震数据进行拉平、奇异值分解数据重构与反拉平等处理方法，对常规奇异值分解算法进行改进，以克服其对包含非水平连续信号资料去噪效果差的局限。理论数据和实际资料的去噪结果表明，改进后的算法去噪效果明显优于常规奇异值分解法，能在保证有效波不被滤除的前提下有效提高地震资料的信噪比。

Local Singular Value Decomposition for Random Noise Suppressing

For low signal to noise （S/N） seismic data, the primary task of data processing is to suppress the random noise and improve the S/N. The paper presents a modified approach that is based on traditional singular value decomposition （SVD）, which can overcome the limitations that traditional SVD can＇t effectively cope with discontinuous and dipping events. Data within a local window are first extracted. Any laterally coherent events in windows are flattened and then SVD is applied for suppressing random noise. Finally, the output data are shifted back to their original positions. The results of synthetic and field data processing show that the approach can effectively improve the signal-to-noise ratio.