Том: 2021-September , Год издания: 2021

Land seismic data are challenging for reliable seismic imaging. Even after extensive processing, desired reflections may remain weak. We present a new approach combining the seismic time-frequency masking with the minimum statistic approach to extract weak reflections from noisy data. Initial signal estimation is provided by nonlinear beamforming. We employ trace-by-trace phase and amplitude time-frequency masking to original data to refine signal estimation, correct distorted phase, and achieve noise suppression. Phase masks are responsible for the coherency of desired events. Amplitude masks aim to suppress amplitude noise. To construct amplitude masks, we adopt the Minimum Statistics approach from speech processing. New methods allow preserving the full frequency content of reflected signals and extracting reliable amplitude behavior.