Paper Image

Deep learning for image restoration

Published on:

22 October 2023

Primary Category:

Computer Vision and Pattern Recognition

Paper Authors:

Zhenghan Fang,

Sam Buchanan,

Jeremias Sulam


Key Details

Proposes learned proximal networks (LPN) to parameterize proximal operators with neural nets

LPN provably converges when used in plug-and-play algorithms for inverse problems

LPN can be trained unsupervised via proximal matching to learn image log-priors

Achieves state-of-the-art results for image deblurring, CT reconstruction, compressed sensing

AI generated summary

Deep learning for image restoration

This paper proposes a deep learning framework called learned proximal networks (LPN) for image restoration. The networks parameterize proximal operators, enabling convergence guarantees when used for solving inverse problems. LPN can be trained in an unsupervised way to learn the log-prior of image distributions. Experiments on image datasets demonstrate state-of-the-art performance for tasks like deblurring, CT reconstruction, and compressed sensing.

Answers from this paper


No comments yet, be the first to start the conversation...

Sign up to comment on this paper

Sign Up