US 12711752 B2 / Aug 18, 2026

Polarized image enhancement using deep neural networks

Methods and systems directed to processing of a polarized image are disclosed. A method may involve determining a polarization characterization for a polarized image. The polarization characterization is indicative of polarization data associated with a plurality of polarization directions of incident light in the polarized image. The method may also involve extracting, from the polarized image, a first collection of global features and a second collection of local features. The method may further involve performing, based on the polarization characterization, a global feature fusion to fuse global features in the first collection, and a local feature fusion to fuse local features in the second collection. The method may involve compositing the polarization characterization with the fused global features and the fused local features to generate a reconstructed image. The method may also involve providing the reconstructed image to an image processing resource to perform one or more image processing tasks.