Low-light polarized image enhancement via multi-scale feature fusion network

Polarization imaging provides additional physical information beyond conventional RGB images, making it valuable in various vision tasks. However, the quality of polarized images is often degraded due to inconsistent light intensities across different angles and the challenges posed by low-light environments. In this paper, we design a multi-scale feature fusion-based network for enhancing low-light polarized images (MPLENet), which progressively enhances low-light polarized image features through three specialized sub-networks: illumination restoration, perceptual enhancement, and polarization information refinement. Specifically, we introduce a lightweight multi-scale residual block and an intensity-polarization attention fusion mechanism. Experimental results demonstrate that our method achieves state-of-the-art performance on both the PLIE and LLCP datasets.