Restoration of Distorted Laser-Spot Image Based on Generative Adversarial Networks
Shui-Qiu Diao, Jianjun Li, Rui Zhang, Feng Wen, Baojian Wu, Kun Qiu · 2024
We introduce the generative adversarial networks (GANs) into the application of the free-space laser communication, performing the laser-spot image restoration with the computervision method. After the training of our model is completed, the average Pearson correlation coefficient, average peak signal-to-noise ratio and mean-square-error of the generated restored laser spot image reach 0.964, 27.713dB and 0.003, respectively. According to the numerical results, the GAN method could effectively resist the interference of atmospheric turbulence, and accurately restore the amplitude distribution of the laser captured from the free-space channel.