Training an Active Random Field for Real-Time Image Denoising

TLDR

Computer vision problems are often framed as Bayesian models using Markov or conditional random fields, which are typically learned independently of the inference algorithm. The authors propose an active random field that jointly trains a Fields‑of‑Experts MRF and a 1–4 iteration gradient‑descent inference algorithm on paired image data by optimizing a loss function. Joint training yields a 1000–3000× speedup and improved benchmark performance, allowing real‑time denoising at 8 fps on a single CPU with near state‑of‑the‑art accuracy.

Abstract

Many computer vision problems can be formulated in a Bayesian framework based on Markov random fields (MRF) or conditional random fields (CRF). Generally, the MRF/CRF model is learned independently of the inference algorithm that is used to obtain the final result. In this paper, we observe considerable gains in speed and accuracy by training the MRF/CRF model together with a fast and suboptimal inference algorithm. An active random field (ARF) is defined as a combination of a MRF/CRF based model and a fast inference algorithm for the MRF/CRF model. This combination is trained through an optimization of a loss function and a training set consisting of pairs of input images and desired outputs. We apply the ARF concept to image denoising, using the Fields of Experts MRF together with a 1-4 iteration gradient descent algorithm for inference. Experimental validation on unseen data shows that the ARF approach obtains an improved benchmark performance as well as a 1000-3000 times speedup compared to the Fields of Experts MRF. Using the ARF approach, image denoising can be performed in real-time, at 8fps on a single CPU for a 256 x 256 image sequence, with close to state-of-the-art accuracy.