HR-NeRF: advancing realism and accuracy in highlight scene representation

NeRF and its variants excel in novel view synthesis but struggle with scenes featuring specular highlights. To address this limitation, we introduce the Highlight Recovery Network (HRNet), a new architecture that enhances NeRF's ability to capture specular scenes. HRNet incorporates Swish activ...

Full description

Saved in:
Bibliographic Details
Main Authors: Shufan Dai, Shanqin Wang
Format: Article
Language:English
Published: Frontiers Media S.A. 2025-04-01
Series:Frontiers in Neurorobotics
Subjects:
Online Access:https://www.frontiersin.org/articles/10.3389/fnbot.2025.1558948/full
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850150938245857280
author Shufan Dai
Shanqin Wang
author_facet Shufan Dai
Shanqin Wang
author_sort Shufan Dai
collection DOAJ
description NeRF and its variants excel in novel view synthesis but struggle with scenes featuring specular highlights. To address this limitation, we introduce the Highlight Recovery Network (HRNet), a new architecture that enhances NeRF's ability to capture specular scenes. HRNet incorporates Swish activation functions, affine transformations, multilayer perceptrons (MLPs), and residual blocks, which collectively enable smooth non-linear transformations, adaptive feature scaling, and hierarchical feature extraction. The residual connections help mitigate the vanishing gradient problem, ensuring stable training. Despite the simplicity of HRNet's components, it achieves impressive results in recovering specular highlights. Additionally, a density voxel grid enhances model efficiency. Evaluations on four inward-facing benchmarks demonstrate that our approach outperforms NeRF and its variants, achieving a 3–5 dB PSNR improvement on each dataset while accurately capturing scene details. Furthermore, our method effectively preserves image details without requiring positional encoding, rendering a single scene in ~18 min on an NVIDIA RTX 3090 Ti GPU.
format Article
id doaj-art-dbf3824e991848a88cc83673feb7e8f1
institution OA Journals
issn 1662-5218
language English
publishDate 2025-04-01
publisher Frontiers Media S.A.
record_format Article
series Frontiers in Neurorobotics
spelling doaj-art-dbf3824e991848a88cc83673feb7e8f12025-08-20T02:26:24ZengFrontiers Media S.A.Frontiers in Neurorobotics1662-52182025-04-011910.3389/fnbot.2025.15589481558948HR-NeRF: advancing realism and accuracy in highlight scene representationShufan DaiShanqin WangNeRF and its variants excel in novel view synthesis but struggle with scenes featuring specular highlights. To address this limitation, we introduce the Highlight Recovery Network (HRNet), a new architecture that enhances NeRF's ability to capture specular scenes. HRNet incorporates Swish activation functions, affine transformations, multilayer perceptrons (MLPs), and residual blocks, which collectively enable smooth non-linear transformations, adaptive feature scaling, and hierarchical feature extraction. The residual connections help mitigate the vanishing gradient problem, ensuring stable training. Despite the simplicity of HRNet's components, it achieves impressive results in recovering specular highlights. Additionally, a density voxel grid enhances model efficiency. Evaluations on four inward-facing benchmarks demonstrate that our approach outperforms NeRF and its variants, achieving a 3–5 dB PSNR improvement on each dataset while accurately capturing scene details. Furthermore, our method effectively preserves image details without requiring positional encoding, rendering a single scene in ~18 min on an NVIDIA RTX 3090 Ti GPU.https://www.frontiersin.org/articles/10.3389/fnbot.2025.1558948/fullscene representationview synthesisimage-based renderingvolume rendering3D deep learningspectral bias
spellingShingle Shufan Dai
Shanqin Wang
HR-NeRF: advancing realism and accuracy in highlight scene representation
Frontiers in Neurorobotics
scene representation
view synthesis
image-based rendering
volume rendering
3D deep learning
spectral bias
title HR-NeRF: advancing realism and accuracy in highlight scene representation
title_full HR-NeRF: advancing realism and accuracy in highlight scene representation
title_fullStr HR-NeRF: advancing realism and accuracy in highlight scene representation
title_full_unstemmed HR-NeRF: advancing realism and accuracy in highlight scene representation
title_short HR-NeRF: advancing realism and accuracy in highlight scene representation
title_sort hr nerf advancing realism and accuracy in highlight scene representation
topic scene representation
view synthesis
image-based rendering
volume rendering
3D deep learning
spectral bias
url https://www.frontiersin.org/articles/10.3389/fnbot.2025.1558948/full
work_keys_str_mv AT shufandai hrnerfadvancingrealismandaccuracyinhighlightscenerepresentation
AT shanqinwang hrnerfadvancingrealismandaccuracyinhighlightscenerepresentation