Abstract: Lateral medium filtering is a general method in the de-noising of seismic data.As the samples have more stringent requirements in the median filtering,de-noising effect is sometimes not ideal.Pre-stack seismic data acquired by vibroseis meets the application conditions of median filtering method (namely,the same energy,unified waveform and time alignment).However,the boundary effect of conventional median filtering is more prominent when the samples for participating median filtering is few.Therefore,the donut,end to end median filtering was proposed.By using this method,samples of participating median filtering can be supplemented by the neighboring data when they are not enough for median filtering.Aiming at the problem that strong noise exists continuously in more traces in seismic data acquired by vibroseis,donut-asymmetric median filtering method was proposed,which gives the method for recognizing the samples with strong noise and their replacement method.Based on theoretical model and real seismic data acquired by vibroseis,the validity,practicability and applicable scope of the donut-asymmetric median filtering technology is discussed.