Degraded image restoration of vortex beam array based on deep learning

Abstract

The phase aberration and intensity fluctuation of the vortex beam array caused by atmospheric turbulence decrease the decoding accuracy of the optical communication system. This paper proposes an end-to-end turbulence-degraded image restoration method based on deep learning to solve the problem. The K-means clustering algorithm is employed to obtain the coordinate information of each beam in the array, and the distorted vortex beam array is segmented. The neural network constructed is used to restore the degraded image of a single vortex beam obtained by segmentation. Then the restored intensity image of the vortex beam array is obtained by combining the existing coordinate information. The simulation results show that the intensity correlation coefficients of the 3 × 3 rectangular distribution Laguerre–Gaussian beam arrays are increased to more than 0.99 after restoring from 1000 meters of transmission in both varied and unknown turbulence intensities, alongside differing CCD signal-to-noise ratios. This method does not require wavefront reconstruction, which further improves the restoration speed and saves computational resources, and has good generalization ability and robustness in quickly restoring the distorted light intensity of vortex beams. The results provide a theoretical basis for studying atmospheric turbulence influence mitigation techniques for vortex optical communication.