Diffusion Models for Multidimensional Seismic Noise Attenuation and Super-Resolution

Seismic data quality proves pivotal to its interpretation, necessitating the reduction of noise and the enhancement of resolution. Both traditional and deep learning-based solutions have achieved varying degrees of success on low-dimensional seismic data. In this paper, we develop a deep generative solution for high-dimensional seismic data denoising and super-resolution through the innovative application of denoising diffusion probabilistic models (DDPMs), which we refer to as MD Diffusion. MD Diffusion treats degraded seismic data as a conditional prior that guides the generative process, enhancing the capability to recover data from complex noise. By iteratively training an implicit probability model, we achieve a sampling speed ten times faster than the original DDPM. Extensive training allows us to explicitly model complex seismic data distributions in synthetic datasets to transfer this knowledge to the process of recovering field data with unknown noise levels, thereby attenuating noise and enhancing resolution in an unsupervised manner. Quantitative metrics and qualitative results for 3D synthetic and field data demonstrate that MD Diffusion exhibits superior performance in high-dimensional seismic data denoising and super-resolution compared to the UNet and Seismic Super-Resolution methods, especially in enhancing thin-layer structures and preserving fault features, and shows the potential for application to higher-dimensional data.