Abstract: There are two problems in the random noise attenuation in f-x domain:(1) Reflected waves as hyperbolic events in pre-stack CMP gather or shot gathers, de-noising will damage the effective wave; (2) Seismic signals are complex non-stationary signals, requiring that the de-noising method has the adaptivity. Aiming at these two problems, a random noise suppression method based on f-x EEMD was proposed. The method utilizes the reflected waves which are horizontal events in common offset gathers, satisfying the f-x domain de-noising assumption, and the good adaptability of EEMD algorithm to non-stationary signals. For each frequency slices in the f-x domain, signal is decomposed into a series of IMFs by EEMD, and the first IMF component which is noise dominant is removed. Finally, f-x domain data is inversely transformed back to t-x domain to realize noise separation. The theoretical trial and practical application indicate that the proposed method can suppress the random noise while maintaining the desired signal.