Skip navigation
Please use this identifier to cite or link to this item:
Title: Enhancing Perceptual Loss with Adversarial Feature Matching for Super-Resolution
Authors: Tej A.R.
Sukanta Halder S.
Shandeelya A.P.
Pankajakshan, Vinod
Published in: Proceedings of the International Joint Conference on Neural Networks
Proceedings of International Joint Conference on Neural Networks, IJCNN 2020
Abstract: Single image super-resolution (SISR) is an ill-posed problem with an indeterminate number of valid solutions. Solving this problem with neural networks would require access to extensive experience, either presented as a large training set over natural images or a condensed representation from another pre-trained network. Perceptual loss functions, which belong to the latter category, have achieved breakthrough success in SISR and several other computer vision tasks. While perceptual loss plays a central role in the generation of photo-realistic images, it also produces undesired pattern artifacts in the super-resolved outputs. In this paper, we show that the root cause of these pattern artifacts can be traced back to a mismatch between the pre-training objective of perceptual loss and the super-resolution objective. To address this issue, we propose to augment the existing perceptual loss formulation with a novel content loss function that uses the latent features of a discriminator network to filter the unwanted artifacts across several levels of adversarial similarity. Further, our modification has a stabilizing effect on non-convex optimization in adversarial training. The proposed approach offers notable gains in perceptual quality based on an extensive human evaluation study and a competent reconstruction fidelity when tested on objective evaluation metrics.
Citation: Proceedings of the International Joint Conference on Neural Networks, 2020. IEEE Computational Intelligence Society (CIS)
Issue Date: 2020
Publisher: Institute of Electrical and Electronics Engineers Inc.
Keywords: Generative Adversarial Networks
Perceptual Loss Functions
Single Image Super-Resolution
Convex optimization
Function evaluation
Optical resolving power
Quality control
Condensed representations
Human evaluation
Ill posed problem
Nonconvex optimization
Objective evaluation
Perceptual quality
Photorealistic images
Stabilizing effect
ISBN: 9.78173E+12
Author Scopus IDs: 57207764881
Author Affiliations: Tej, A.R., Indian Institute of Technology, Roorkee, India
Sukanta Halder, S., Carnegie Mellon University, United States
Shandeelya, A.P., International Institute of Information Technology, Bhubaneswar, India
Pankajakshan, V., Indian Institute of Techno
Appears in Collections:Conference Publications [ECE]

Files in This Item:
There are no files associated with this item.
Show full item record

Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.