Thermal optical aberrations correction via postprocessing in a raw domain

Optical imaging systems are susceptible to severe performance degradation when exposed to extreme temperature fluctuations. While extensive athermalization can mitigate such environmental effects, the complex design process, material limitations, and prohibitive cost greatly restrict its practical application. Herein, we propose what we believe to be a novel postprocessing scheme to compensate for the thermal-induced degradation. We simulate degraded raw image pairs based on optical point spread functions (PSFs) across a range of temperatures. To perform effective restoration, a PSF-guided neural network is specially designed. Experiments conducted on a customized digital-single-lens-reflex (DSLR) camera lens demonstrate that our solution successfully removes the thermal optical aberrations in both synthetic and experimentally captured scenes. Notably, our proposed approach outperforms several state-of-the-art (SOTA) deblurring methods, showcasing great promise for broad applicability in thermal-challenged imaging systems.