Paper Info
Title
Normalizing Flow as a Flexible Fidelity Objective for Photo-Realistic Super-resolution
Abstract
Super-resolution is an ill-posed problem, where a ground-truth high-resolution image represents only one possibility in the space of plausible solutions. Yet, the dominant paradigm is to employ pixel-wise losses, such as L <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</inf> , which drive the prediction towards a blurry average. This leads to fundamentally conflicting objectives when combined with adversarial losses, which degrades the final quality. We address this issue by revisiting the L <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</inf> loss and show that it corresponds to a one-layer conditional flow. Inspired by this relation, we explore general flows as a fidelity-based alternative to the L <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</inf> objective. We demonstrate that the flexibility of deeper flows leads to better visual quality and consistency when combined with adversarial losses. We conduct extensive user studies for three datasets and scale factors, where our approach is shown to outperform state-of-the-art methods for photo-realistic super-resolution. Code and trained models: git.io/AdFlow
Year
DOI
Venue
2022
10.1109/WACV51458.2022.00095
2022 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)
Keywords
DocType
ISSN
Computational Photography, Image and Video Synthesis
Conference
2472-6737
ISBN
Citations 
PageRank 
978-1-6654-0916-2
1
0.35
References 
Authors
0
5
Name
Order
Citations
PageRank
Andreas Lugmayr171.10
Danelljan Martin2134449.35
Fisher Yu3128050.27
Luc Van Gool410.69
Radu Timofte51880118.45