Аннотация:Low frequency noise attenuation of high quality data is a critical step prior to the application of many modern seismic processing algorithms. While great progress has been made in recent years through the adoption of techniques from fields related to seismic imaging such as compressive sensing, not all approaches necessarily honour the physics of the problem. Here, we present an improved apex shifted hyperbolic Radon transform based on frequency-wavenumber and phase-shift extrapolation operators that take into account vertical velocity variations. The use of these migration and modelling operator pairs enables us to inject our knowledge about the physics of seismic wave propagation into the noise attenuation problem itself. This in turn allows effective protection of primary signal critical for seismic imaging. To demonstrate the versatility of our improved transform, we first explore the separation of low frequency signal and noise using a synthetic example. We then demonstrate that the use of our algorithm greatly improves the continuity and interpretability of a deep crustal reflector observed offshore Namibia.