Computer Science > Computer Vision and Pattern Recognition
[Submitted on 29 Nov 2023 (v1), last revised 14 Mar 2024 (this version, v2)]
Title:Neural Fields with Thermal Activations for Arbitrary-Scale Super-Resolution
View PDF HTML (experimental)Abstract:Recent approaches for arbitrary-scale single image super-resolution (ASSR) have used local neural fields to represent continuous signals that can be sampled at arbitrary rates. However, the point-wise query of the neural field does not naturally match the point spread function (PSF) of a given pixel, which may cause aliasing in the super-resolved image. We present a novel way to design neural fields such that points can be queried with an adaptive Gaussian PSF, so as to guarantee correct anti-aliasing at any desired output resolution. We achieve this with a novel activation function derived from Fourier theory. Querying points with a Gaussian PSF, compliant with sampling theory, does not incur any additional computational cost in our framework, unlike filtering in the image domain. With its theoretically guaranteed anti-aliasing, our method sets a new state of the art for ASSR, while being more parameter-efficient than previous methods. Notably, even a minimal version of our model still outperforms previous methods in most cases, while adding 2-4 orders of magnitude fewer parameters. Code and pretrained models are available at this https URL.
Submission history
From: Alexander Becker [view email][v1] Wed, 29 Nov 2023 14:01:28 UTC (6,745 KB)
[v2] Thu, 14 Mar 2024 19:17:34 UTC (9,028 KB)
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
Connected Papers (What is Connected Papers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.