Frequency down-extrapolation with TV norm minimization

摘要

In this work, we present a total-variation (TV)-norm minimization method to perform model-free low-frequency data extrapolation for the purpose of assisting full-waveform inversion. Low-frequency extrapolation is the process of extending reliable frequency bands of the raw data towards the lower end of the spectrum. To this end, we propose to solve a TV-norm based convex optimization problem that has a global minimum and is equipped with a fast solver. The approach takes into account both the sparsity of the reflectivity series associated with a single trace, as well as the inter-trace correlations. A favorable byproduct of this approach is that it allows one to work with coarsely sampled trace along time (as coarse as 0.02s per sample), hence substantially reduces the size of the proposed optimization problem. Last, we show the effectiveness of the method for frequency-domain FWI on the Marmousi model.