Search papers, labs, and topics across Lattice.
The paper introduces CFSR, a multi-modal prior-driven framework for shadow removal that integrates 3D geometric cues with large-scale foundation model semantics to enhance physical interpretability in image restoration. By employing a Geometric & Semantic Dual Explicit Guided Attention mechanism, CFSR enforces physical lighting constraints, effectively bridging the 2D-3D domain gap and improving localized texture recovery alongside global illumination consistency. Extensive experiments reveal that CFSR outperforms existing methods on several challenging benchmarks, establishing new state-of-the-art results in shadow removal tasks.
CFSR redefines shadow removal by integrating 3D geometry with semantic understanding, achieving unprecedented restoration quality in challenging scenarios.
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.
