Ghost imaging is a technique for indirectly reconstructing images by utilizing the second-order or higher-order correlation properties of the light field, which exhibits a robust ability to resist interference. On the premise of ensuring the quality of the image, effectively broadening the imaging range can improve the practicality of the technology. In this paper, a dual-path computational ghost imaging method based on convolutional neural networks is proposed. By using the dual-path detection structure, a wider range of target image information can be obtained, and the imaging range can be expanded. In this paper, for the first time, we try to use the two-channel probe as the input of the convolutional neural network and successfully reconstruct the target image. In addition, the network model incorporates a self-attention mechanism, which can dynamically adjust the network focus and further improve the reconstruction efficiency. Simulation results show that the method is effective. The method in this paper can effectively broaden the imaging range and provide a new idea for the practical application of ghost imaging technology.

Computational ghost imaging (CGI) is an imaging technique that uses the second-order correlation properties of optical fields. It uses random illumination patterns to illuminate the object, collects the signal via a single-pixel bucket detector, and later reconstructs an image of the object. By treating the illumination patterns as keys and the bucket measurements as ciphertexts, CGI can be regarded as an encryption process. This paper proposes a multi-party collaborative decryption scheme based on CGI, which aims to reduce key transmission data whilst enabling cooperative decryption among multiple users and verifying the legitimacy of the ciphertext source via digital watermarking. This method uses a chaotic system to generate a measurement matrix, and only the chaotic parameter needs transmission. The receiver can reconstruct the measurement matrix, thereby reducing the amount of key transmission. Further, verifiable secret sharing (VSS) is employed to divide the chaotic parameter into multiple shares, which are then distributed to different recipients. Decryption is only feasible when a threshold number of recipients collaborate to reconstruct the chaotic parameter. To further improve security, hashed identity information is embedded into the ciphertext as a watermark, enabling confirmation of the ciphertext’s origin. Experimental results show the feasibility of the proposed scheme and its ability to resist tampering, offering a new approach to secure multi-user communication.