Search papers, labs, and topics across Lattice.
This paper introduces GenPolar, a Stokes-informed diffusion framework that estimates linear polarization from a single RGB image, addressing the challenges of instability in the Angle of Polarization (AoP) in weak polarization regions. By predicting channel-wise linear Stokes components from intensity observations and employing an observability-aware loss for AoP supervision, the method achieves high fidelity in degree of linear polarization (DoLP) and stability in AoP. Extensive experiments validate that GenPolar outperforms existing methods, leading to significant improvements in practical applications like material detection and de-reflection.
Achieving state-of-the-art stability in Angle of Polarization estimation from a single RGB image could revolutionize how we approach polarization-based applications without specialized hardware.
Polarization cues benefit applications such as material detection and de-reflection, yet acquiring them typically requires dedicated hardware. This motivates us to estimate the linear polarization from a single RGB image. However, the task is inherently ill-posed, with the Angle of Polarization (AoP) becoming particularly unstable in weak polarization regions, where the polarimetric signal is overwhelmed by noise, leading to erratic angle estimates. To address these limitations, we propose GenPolar, a Stokes-informed diffusion framework grounded in the Mueller formalism from an intensity observation. Specifically, GenPolar predicts channel-wise linear Stokes components (S1,S2) from intensity S0, from which degree of linear polarization (DoLP) and AoP are analytically derived; AoP is further supervised with an observability-aware loss. In addition, to enable efficient and high-fidelity inference, we adopt a two-stage training strategy. Firstly, a multi-step conditional diffusion model is trained with a physics-based loss. Subsequently, we distill it into a one-step generator, which further supports stable Low-Rank Adaptation (LoRA) of the VAE encoder to mitigate domain-specific autoencoding bias. Extensive experiments across rotating-polarizer, division-of-focal-plane, and hybrid datasets demonstrate that GenPolar achieves state-of-the-art performance in both DoLP fidelity and AoP stability. Crucially, these improvements translate to significant and consistent gains in downstream applications, including material detection and de-reflection.
