常规的ｆ－ｘ域预测滤波不仅理论上的假设比较苛刻，而且在分离随机噪音和有效信号时会引起信号畸变，降低输出剖面的保真效果和分辨率。本文对预测滤波器的各种误差来源和稳定性进行了分析，并提出了相应的改进措施。理论模型试算和实际资料处理结果表明，改进的ｆ－ｘ域预测滤波可以克服以上问题，使压制随机噪音的效果变好 Conventional f-x domain prediction filtering not only theoretically assumes harsh conditions, but also disturbs signal distortion and reduces the fidelity and resolution of the output profile when separating random noise and valid signals. In this paper, various error sources and stability of the prediction filter are analyzed, and the corresponding improvement measures are put forward. The theoretical model trial and actual data processing results show that the improved f-x domain prediction filter can overcome the above problems and make the effect of suppressing random noise better