Abstract:Mid-value filtering is a specific nonlinear smoothing filtering, and widely used in graphic processing, etc. Mid-value filtering before stack of seismic data can remove wild value (inconceivable particular big value)and improve signal-noise ratio. In view of big data volume before stack, computation efficiency of mid-value filtering is critical to its feasibility. The a'igorithm used here is sufficiently based on the properties of mid-value filtering, so that the computation efficiency is greatly improved. It is experimentally shown that prestack mid-value filtering can quite well eliminate wild value, abnormal traces as well as surface waves, and raise signal-noise ratio. After lateral low frequency noises are removed by high pass filtering, mid-value filtering works better.