Multichannel deconvolution based on smooth streaming prediction error filter

Deconvolution is an effective method for improving seismic resolution of imaging and reservoir prediction. Traditional deconvolution is usually implemented through trace-by-trace inversion under stationary conditions, and improved resolution cannot offset poor spatial continuity owing to the lack of spatial constraints. This paper proposes a multichannel deconvolution method based on a streaming prediction error filter, which uses temporal and spatial constraints to achieve multichannel adaptive deconvolution and improve the spatial continuity of non-stationary seismic data after deconvolution. A smoothing matrix is employed so as not to blur boundaries and geological structures, particularly complex structures. The new deconvolution method can effectively improve the vertical resolution of seismic data and reduce the workload through streaming computation, making it suitable for non-stationary big data. The processing results of synthetic data show that spatial constraints improve the spatial continuity of deconvolution, and a field data test verifies the effectiveness and practicality of this method.