Lightweight and rapid adaptive super-resolution interpolation reconstruction algorithm for super-pixel segmentation of super-wide field-of-view long-wave infrared images

Super-wide field-of-view (SWFOV) long-wave infrared (LWIR) images possess notable advantages, including strong anti-interference capabilities and an expansive imaging field of view, making them suitable for all-weather and multi-disciplinary applications. However, significant distortion and suboptimal texture information in SWFOV LWIR images impede advancements in SWFOV LWIR imaging detection. This study presents an innovative approach utilizing superpixel segmentation for SWFOV LWIR images, wherein each superpixel is classified based on its gradient to facilitate adaptive super-resolution interpolation and reconstruction. Notably, the “sinc” function is employed to construct the interpolation weighting function, particularly for high-gradient regions. The results demonstrate that the reconstructed SWFOV LWIR images exhibit enhanced object edge information, while the dentate edge ringing effect is notably mitigated. Evaluation of the reconstructed images following a 2× reconstruction of dataset 1 reveals that the energy of gradient function of the proposed algorithm exhibits a 21.93% improvement compared to the sub-optimal algorithm. In addition, the Vollath parameter, Roberts parameter, and Laplace parameter increase by 7.16%, 14.84%, and 15.62%, respectively. Furthermore, our algorithm in this paper has certain advantages, such as low-intensity noise and a calculation time of around 0.5 s, making it have certain application value for the SWFOV LWIR imaging system that requires real-time high-definition imaging.