F-X域经验模态分解去噪方法在处理非稳态地震数据时存在两个局限,一是单纯剔除第一个固有模态分量将导致有效信号缺失及去噪能力偏弱问题,二是分解复信号时对实部和虚部分别分解存在分解数目不一致的风险。本文对上述两个方面进行了改进,提出了一种新的F-X域投影法复数经验模态分解预测滤波方法,首先采用基于空间投影的复数经验模态分解将F-X域地震数据直接分解为不同的复固有模态分量,然后再对这些分量分别进行F-X域预测滤波。合成记录及实际资料测试表明,本文的新方法能更好地衰减随机噪声,更有效地保持地震信号。 There are two limitations in the FX domain empirical mode decomposition de-noising method when dealing with unsteady seismic data. One is that simply removing the first natural mode component will result in the loss of effective signal and the weakening of the denoising ability. The second is the decomposition of complex Signal when the real and imaginary part of the existence of decomposition of the number of decomposition of the risk of inconsistencies. In this paper, we improve the above two aspects and propose a new method of empirical EMF prediction filtering in FX domain projection. Firstly, we use the empirical EMF based on spatial projection to decompose the FX domain seismic data directly into different Complex modal components, and then these components separately FX domain prediction filter. The synthetic records and actual data tests show that the new method in this paper can better attenuate the random noise and maintain the seismic signal more effectively.