Adaptive nonlinear dynamic range compression for enhanced pixelated polarimetric fusion imaging

Long-wave infrared division-of-focal-plane (DoFP) polarimetric imaging can simultaneously acquire target intensity and polarization information, but spatial undersampling, channel nonuniformity, and weak polarization responses often cause resolution degradation, insufficient target saliency, and limited multidimensional information representation. This paper proposes a pixelated polarization fusion enhancement method based on adaptive nonlinear dynamic range compression (NDRC), including pixel-level data correction, sliding-window Stokes reconstruction, weak-polarization enhancement, and NDRC-HSV fusion. Pixelwise quadratic nonlinear radiometric calibration reduces the fitting error of representative pixels to approximately 1/12.7 of that obtained with a linear model, thereby improving the response consistency of the LWIR DoFP detector. Sliding-window Stokes reconstruction recovers polarization details at a resolution close to that of the original focal plane. The proposed NDRC enhances weak polarization responses through adaptive logarithmic expansion, noise-constrained sigmoid mapping, progressive fusion, and lightweight post-processing. In the metal-sphere scene, the target–background mean difference and target-to-background ratio are increased to approximately 58.4 and 1.29 times those of the original DoLP result, respectively. Compared with Direct HSV, the proposed fusion method increases ̅ ̅ ̅ ̅ ̅ ̅ 𝑆 R O I and 𝛥 ⁢ 𝑆 R O I − B G to approximately 25.7 and 32.0 times in the dark region of the aircraft model, and to approximately 31.6 and 35.0 times in the eyeglass region of the portrait, respectively. Across eight representative scenes, the average gradient of the fusion results improves by 9.8%–38.3% over 𝑆 0 , demonstrating its applicability to different targets, materials, and viewing conditions. The results confirm that the proposed method improves weak-target detectability, edge clarity, and multidimensional information visualization in LWIR polarimetric images.