Infrared Image Super Resolution Using GAN With Infrared Image Prior

Super-resolution reconstruction technology based on deep-learning is rarely used in the field of infrared image. This paper will apply the Generative Adversarial Network super-resolution approach to the infrared super-resolution task. The natural image gradient prior is introduced into the super-resolution algorithm, and the visible image of the corresponding scene and the field of view is innovatively used as the style map, and the corresponding shallow network perceptual loss and deep network perceptual loss are added to the super-resolution objective function. The reconstructed image is more abundant and more detailed in the subjective visual reconstruction of the image texture than the existing algorithm in the simulation experiment.