Seismic data interpolation based on diffusion model deep learning

In this article, we propose a new deep learning network training method based on the diffusion model, which can be applied to seismic data interpolation. The core of this method lies in applying the diffusion model to generate labeled data with different levels of Gaussian random noise, and then using existing network models such as U-net or ResNet for training. The benefit of this approach is its simple implementation and excellent data interpolation performance. Random missing and continuous missing tests show that the proposed algorithm can effectively reconstruct the missing seismic data.