Abstract：Lithology analysis methods, such as wave impedance (ρv), ampltude variation with offset (AVO) and seismic lithologic modeling (SLIM), have been widely used in geophysical exploration along with the development of geophysical exploration. These methods need good amplitude and waveform preservations. However, amplitude and waveform variations always occur because of the surface in consistency resulting from shot and receiving conditions. In recent years, the methods such as surface-consistent amplitude compensation, surface-consistent deconvolution and statistical wavelet deconvolution have been introduced to remove these deficiencies. These methods all use surface consistency to separate out amplitude and twavelet informations, bringing good results. However, non-surface-consistent noises exist in land seismic data, and cause bad influence on above theoretical statistical wavelets. The non-surface-consistent noises in common-shot gather are low apparent-velocity surface wave, 50Hz industrial noise, culture noise, environmental noise and so on. In order to remove the influence of the non-surface-consistent noises, we perform （f-x) domain mid-value filtering of common-shot or common-receiver data, It has been shown theoretically and practically that our measure can improve deconvolution result to achieve waveform preservation.
