Atmospheric Turbulence Image Recovery Algorithm Based on Lucky Imaging and Generative Adversarial Networks

In order to restore the degraded images of atmospheric turbulence in video sequences, an algorithm combining Lucky Imaging and Generative Adversarial Network(GAN) is proposed in order to restore the degraded images of atmospheric turbulence in video sequences. In this paper, the spatial lucky imaging algorithm is used to select the lucky regions in the limited video sequence images, and superimpose them after stitching-sorting, so as to eliminate the geometric distortion caused by atmospheric turbulence, and on this basis, the DeblurGAN-v2 model is introduced to further improve the image quality. In this paper, the real turbulence degradation image taken by the high-speed camera is taken as the research object, and the method proposed in this paper is used to carry out experiments, and compared with the image resampling, grayscale transformation and Butterworth high-pass filtering algorithms, and the results of different algorithms are evaluated by objective evaluation indicators. The experimental results show that the Brenner gradient function, Laplacian gradient function, SMD, Entropy, Energy, PIQE and Brisque indicators of this method are improved by 291%, 66%, 127%, 10%, 74% and 159% compared with other methods, respectively. From the subjective effect, the algorithm combining Lucky Imaging and GAN can significantly improve the visual quality of the image and effectively reduce the degree of blur and geometric distortion of the image.