Image-Domain Least-Squares Migration in Rapidly-Varying Media: Practical Considerations

Least-squares migration implemented in the image domain by approximating the blurring operator through a set of point spread functions has proven to be a viable method to mitigate the blurring effects present in depth migrated data. Yet, rapidly-changing media can challenge the accuracy/cost trade-off required for practical application of the approach, especially if the lithology justifies the inclusion of high contrasts in the velocity model (e.g. at salt boundaries). In this paper we discuss several methods to mitigate inaccuracies arising from such media, whilst keeping the cost affordable. We propose to work with irregularly-sampled PSFs and to rely on a masked linear interpolation that prevents interpolating PSFs across high velocity contrasts. Furthermore, when two-way wavefield propagators are employed, we propose two complementary approaches to mitigate high-order scattering artefacts arising at high velocity contrasts and distorting point spread functions close to these contrasts. We show that the combination of these approaches can help approximate the blurring operator more reliably, leading to improved inversion results.