Adaptive subtraction on an f - k -partitioned wavefield
Ahmed Rushdy, Jing Wu, Zhiming James Wu, Cintia Mariela Lapilli, Vasudha Govindan · 2021
We present an f-k-partitioning-based adaptive subtraction method that can be tuned with a single meta-parameter. Historically, adaptive subtraction requires a non-linear testing process that can take weeks to reach a satisfactory result. We present a method based on f-k partitioning that can deliver accurate subtraction results with much less testing. Inspired by curvelet-domain methods, we decompose events by panels in the f-k domain. Unlike curvelet-domain methods, the subtraction is performed per panel in the t-x domain. This enables filter designs to be proportional to the t-x wavelength of each wavefield component and scaled by a single harshness parameter. The partitioned-wavefield adaptive subtraction method reduces testing time by 70% in example projects owing to filters that adapt more consistently to wavefield variations than traditional cascaded t-x subtraction.