CFSR: Geometry-Conditioned Shadow Removal via Physical Disentanglement

TL;DR

CFSR introduces a physics-constrained, geometry-aware shadow removal framework that integrates 3D cues and foundation model semantics for improved restoration quality.

Contribution

It presents a novel multi-modal prior-driven approach combining geometric and semantic cues with explicit attention mechanisms for shadow removal.

Findings

Achieves state-of-the-art results on multiple benchmarks.

Effectively suppresses shadow-induced noise using a custom HVI color space.

Balances high-frequency detail recovery with global illumination consistency.

Abstract

Traditional shadow removal networks often treat image restoration as an unconstrained mapping, lacking the physical interpretability required to balance localized texture recovery with global illumination consistency. To address this, we propose CFSR, a multi-modal prior-driven framework that reframes shadow removal as a physics-constrained restoration process. By seamlessly integrating 3D geometric cues with large-scale foundation model semantics, CFSR effectively bridges the 2D-3D domain gap. Specifically, we first map observations into a custom HVI color space to suppress shadow-induced noise and robustly fuse RGB data with estimated depth priors. At its core, our Geometric & Semantic Dual Explicit Guided Attention mechanism utilizes DINO features and 3D surface normals to directly modulate the attention affinity matrix, structurally enforcing physical lighting constraints. To recover severely degraded regions, we inject holistic priors via a frozen CLIP encoder. Finally, our Frequency Collaborative Reconstruction Module (FCRM) achieves an optimal synthesis by decoupling the decoding process. Conditioned on geometric priors, FCRM seamlessly harmonizes the reconstruction of sharp high-frequency occlusion boundaries with the restoration of low-frequency global illumination. Extensive experiments demonstrate that CFSR achieves state-of-the-art performance across multiple challenging benchmarks.