cs.CVJul 1, 2024

An Expectation-Maximization Algorithm for Training Clean Diffusion Models from Corrupted Observations

Authors: Weimin BaiYifei WangWenzheng ChenHe Sun

Organizations: Academy for Advanced Interdisciplinary Studies, Peking University · College of Future Technology, Peking University · National Biomedical Imaging Center, Peking University · Yuanpei College, Peking University · Wangxuan Institue of Computer Technology, Peking University · State Key Laboratory of Multimedia Information Processing, Peking University, Beijing, China

Abstract

Diffusion models excel in solving imaging inverse problems due to their ability to model complex image priors. However, their reliance on large, clean datasets for training limits their practical use where clean data is scarce. In this paper, we propose EMDiffusion, an expectation-maximization (EM) approach to train diffusion models from corrupted observations. Our method alternates between reconstructing clean images from corrupted data using a known diffusion model (E-step) and refining diffusion model weights based on these reconstructions (M-step). This iterative process leads the learned diffusion model to gradually converge to the true clean data distribution. We validate our method through extensive experiments on diverse computational imaging tasks, including random inpainting, denoising, and deblurring, achieving new state-of-the-art performance.

Explore similar work

CardsList
  1. Learning Normalized Energy Models for Linear Inverse Problems

    May 15, 2026Nicolas Zilberstein, Santiago Segarra, Eero Simoncelli +1Ill-Posed Inverse ProblemInverse Problem

  2. Latent Diffusion for Missing Data

    May 27, 2026Alberte Heering Estad, Ignacio Peis, Jes FrellsenLatent Diffusion ModelLatent Space