2D Neural Fields with Learned Discontinuities

Simple item page
dc.contributor.authorLiu, Chenxien_US
dc.contributor.authorWang, Siqien_US
dc.contributor.authorFisher, Matthewen_US
dc.contributor.authorAneja, Deepalien_US
dc.contributor.authorJacobson, Alecen_US
dc.contributor.editorBousseau, Adrienen_US
dc.contributor.editorDay, Angelaen_US
dc.date.accessioned2025-05-09T09:11:07Z
dc.date.available2025-05-09T09:11:07Z
dc.date.issued2025
dc.description.abstractEffective representation of 2D images is fundamental in digital image processing, where traditional methods like raster and vector graphics struggle with sharpness and textural complexity, respectively. Current neural fields offer high fidelity and resolution independence but require predefined meshes with known discontinuities, restricting their utility. We observe that by treating all mesh edges as potential discontinuities, we can represent the discontinuity magnitudes as continuous variables and optimize. We further introduce a novel discontinuous neural field model that jointly approximates the target image and recovers discontinuities. Through systematic evaluations, our neural field outperforms other methods that fit unknown discontinuities with discontinuous representations, exceeding Field of Junction and Boundary Attention by over 11dB in both denoising and super-resolution tasks and achieving 3.5× smaller Chamfer distances than Mumford-Shah-based methods. It also surpasses InstantNGP with improvements of more than 5dB (denoising) and 10dB (super-resolution). Additionally, our approach shows remarkable capability in approximating complex artistic and natural images and cleaning up diffusion-generated depth maps.en_US
dc.description.number2
dc.description.sectionheadersDrawn to Detail: Sketch-Based Modeling and Non-Photorealistic Rendering
dc.description.seriesinformationComputer Graphics Forum
dc.description.volume44
dc.identifier.doi10.1111/cgf.70023
dc.identifier.issn1467-8659
dc.identifier.pages15 pages
dc.identifier.urihttps://doi.org/10.1111/cgf.70023
dc.identifier.urihttps://diglib.eg.org/handle/10.1111/cgf70023
dc.publisherThe Eurographics Association and John Wiley & Sons Ltd.en_US
dc.rightsAttribution-NonCommercial 4.0 International
dc.rights.urihttps://creativecommons.org/licenses/by-nc/4.0/
dc.subjectCCS Concepts: Computing methodologies → Image representations; Reconstruction; Neural networks
dc.subjectComputing methodologies → Image representations
dc.subjectReconstruction
dc.subjectNeural networks
dc.title2D Neural Fields with Learned Discontinuitiesen_US
Files
Original bundle
Now showing 1 - 1 of 1
No Thumbnail Available
Name:
cgf70023.pdf
Size:
209.42 MB
Format:
Adobe Portable Document Format
Download
Collections
EG 2025 - Full Papers - CGF 44-Issue 2
44-Issue 2