Adaptive Regularization of Infrared Image Super-resolution Reconstruction

For conventional reconstruction algorithms, regularization parameter is randomly selected and image reconstruction cannot achieve the desired display effect. Thus this paper presents a simple and efficient adaptive regularization technique of infrared image super-resolution reconstruction algorithm that combines L1-norm with the total variation regularization. Regular terms select regularization parameters adaptively based on the difference between the estimated low-resolution images and the actual ones. The experiment results show that the contrast of infrared images reconstructed has increased to 1.4 times as the traditional algorithm and the image edge effectively has been enhanced with the signal-to-noise ratio improved dramatically.