Single Pixel Optical Encryption in Atmospheric Turbulence

Atmospheric turbulence causes random distortions in optical signals transmitted through the air. This paper proposes a method to improve image recovery using a single-pixel detector and a deep learning network designed to reduce turbulence effects in optical communication. Hadamard patterns are used to encrypt light from a target into ciphertext, which is measured by the single-pixel detector. Differential calculations then retrieve a degraded image after decryption. To enhance image quality, we incorporate a channel attention mechanism in the DeepRFTECA network for better feature fusion and use a ResFFT-Conv block to effectively integrate residual information. This approach ensures secure transmission and high-quality image recovery, demonstrated with remote sensing data and optical experiments.