De-aliasing using the U-Net image segmentation algorithm
Madhav Vyas, Qingqing Liao · 2020
It is fairly common to encounter aliasing as part of seismic data processing. Generally, it is the waves that travel slowly that tend to be aliased. Almost all existing methods that address aliasing are adhoc best-practice workflows as opposed to accurate Physics-based solutions. Therefore, we do believe that this problem is ideally suited for machine-learning algorithms. Here we propose a method, that utilizes deep learning based image segmentation to de-alias seismic data. This concept can also be easily integrated as part of interpolation schemes such as POCS and Anti-Leakage Fourier Transforms. Presentation Date: Tuesday, October 13, 2020 Session Start Time: 8:30 AM Presentation Time: 8:30 AM Location: 351F Presentation Type: Oral