Refined Diffusion Posterior Sampling for Seismic Denoising and Interpolation

Abstract Seismic data is vital in the petroleum industry, revealing subsurface structures and reservoir potential. However, this data frequently suffers from degradation due to noise and missing traces, necessitating denoising and interpolation tasks. Recently, diffusion models have gained prominence, showing remarkable results in diverse image enhancement tasks. In this paper, we explore several cutting-edged diffusion-based techniques for addressing seismic denoising and interpolation challenges. Furthermore, we propose a novel method, which addresses the limitations of existing approaches, establishing a new state-of-the-art in the field, offering enhanced accuracy and robustness in seismic data processing.