Coordinate-Based Seismic Interpolation in Irregular Land Survey: A Deep Internal Learning Approach

Physical and budget constraints often result in irregular sampling, which complicates accurate subsurface imaging. Pre-processing approaches, such as missing trace or shot interpolation, are typically employed to enhance seismic data cases. Recently, deep learning has been used address the interpolation problem at expense of large amounts training adequately represent typical events. Nonetheless, most research this area focused on reconstruction, with little attention having devoted interpolation. Furthermore, existing methods assume regularly spaced receivers/sources failing approximating from real (irregular) surveys. This work presents a novel gather approach uses continuous coordinate-based representation acquired wavefield parameterized by neural network. The proposed unsupervised approach, we call (CoBSI), enables prediction specific characteristics land surveys without using external during network training. Experimental results synthetic 3D validate ability method estimate smooth events time-space frequency-wavenumber domains, improving sparsity low-rank-based methods.