A polarization-based dehazing method leveraging multi-angle DoP distribution and dual constraints

• First quantitative analysis of polarization’s advantages in resolving atmospheric scattering parameters. • Introduces MDoP polarization feature to better separate foreground from background while recovering details. • Builds transmittance model using MDoP-atmospheric light distribution similarity. • Proposes multi-feature regularization to improve dehazing robustness and visual quality. Physical model-based dehazing methods have a strong theoretical foundation, with precise transmission gradient estimation critical for image quality. However, existing methods often emphasize overall transmission accuracy while neglecting local details and textures. Polarization-based approaches provide extra physical information, improving foreground-background separation and edge sensitivity, but polarization data are often underused for enhancing clarity. This article proposes a novel polarization-based dehazing method: First, a multi-angle degree of polarization (MDoP) is computed from images at different polarization angles, better representing object-background differences. Second, a dual transmission-gradient estimation model is developed, leveraging MDoP and atmospheric light distribution along with scene depth for improved transmission estimation. Third, multiple indicator constraints enhance local transmission accuracy while ensuring global consistency.Experiments demonstrate that the method effectively captures scene variations, producing detailed, clear, haze-free images without extra hardware or complex procedures. It shows excellent stability and adaptability across diverse environments.