Infrared image super-resolution algorithm based on global channel attention

Abstract

Infrared imaging technology is widely used in security, chemical industry, military and other fields due to its strong penetration and excellent anti-interference ability. However, limited by the wavelength of the infrared band and the manufacturing process of detectors, the resolution of infrared images is usually low, so improving their spatial resolution is of great significance. Aiming at the problems of insufficient accuracy of traditional interpolation methods and the fact that most existing deep learning methods are designed for visible light and difficult to adapt to complex degradation, this paper proposes an infrared image super-resolution network based on edge-conditioned global channel attention. This method combines multi-scale feature modeling and edge-guided channel recalibration to enhance the expression of structural information, suppress noise interference, and achieve high-quality reconstruction. Experimental results show that the proposed method outperforms existing methods in both objective metrics and subjective visual effects.