PhaSR: Generalized Image Shadow Removal with Physically Aligned Priors

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Lee, Chia-Ming, Lin, Yu-Fan, Hsiao, Yu-Jou, Jiang, Jin-Hui, Liu, Yu-Lun, Hsu, Chih-Chung
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911549456646144
author Lee, Chia-Ming
Lin, Yu-Fan
Hsiao, Yu-Jou
Jiang, Jin-Hui
Liu, Yu-Lun
Hsu, Chih-Chung
author_facet Lee, Chia-Ming
Lin, Yu-Fan
Hsiao, Yu-Jou
Jiang, Jin-Hui
Liu, Yu-Lun
Hsu, Chih-Chung
contents Shadow removal under diverse lighting conditions requires disentangling illumination from intrinsic reflectance, a challenge compounded when physical priors are not properly aligned. We propose PhaSR (Physically Aligned Shadow Removal), addressing this through dual-level prior alignment to enable robust performance from single-light shadows to multi-source ambient lighting. First, Physically Aligned Normalization (PAN) performs closed-form illumination correction via Gray-world normalization, log-domain Retinex decomposition, and dynamic range recombination, suppressing chromatic bias. Second, Geometric-Semantic Rectification Attention (GSRA) extends differential attention to cross-modal alignment, harmonizing depth-derived geometry with DINO-v2 semantic embeddings to resolve modal conflicts under varying illumination. Experiments show competitive performance in shadow removal with lower complexity and generalization to ambient lighting where traditional methods fail under multi-source illumination. Our source code is available at https://github.com/ming053l/PhaSR.
format Preprint
id arxiv_https___arxiv_org_abs_2601_17470
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle PhaSR: Generalized Image Shadow Removal with Physically Aligned Priors
Lee, Chia-Ming
Lin, Yu-Fan
Hsiao, Yu-Jou
Jiang, Jin-Hui
Liu, Yu-Lun
Hsu, Chih-Chung
Computer Vision and Pattern Recognition
Shadow removal under diverse lighting conditions requires disentangling illumination from intrinsic reflectance, a challenge compounded when physical priors are not properly aligned. We propose PhaSR (Physically Aligned Shadow Removal), addressing this through dual-level prior alignment to enable robust performance from single-light shadows to multi-source ambient lighting. First, Physically Aligned Normalization (PAN) performs closed-form illumination correction via Gray-world normalization, log-domain Retinex decomposition, and dynamic range recombination, suppressing chromatic bias. Second, Geometric-Semantic Rectification Attention (GSRA) extends differential attention to cross-modal alignment, harmonizing depth-derived geometry with DINO-v2 semantic embeddings to resolve modal conflicts under varying illumination. Experiments show competitive performance in shadow removal with lower complexity and generalization to ambient lighting where traditional methods fail under multi-source illumination. Our source code is available at https://github.com/ming053l/PhaSR.
title PhaSR: Generalized Image Shadow Removal with Physically Aligned Priors
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2601.17470