Ghost edge imaging with untrained neural networks

Edge detection based on ghost imaging technology can directly capture the edge details of a target without acquiring the entire image of the object. In this paper, we propose a method of ghost edge imaging based on untrained neural network. The method initially generates a set of shifted random binary speckle patterns, then illuminates the object to obtain eight sets of detection values. These eight sets are recombined into two sets of detection values, which respectively contain horizontal and vertical edge information and are fed into a manually designed, pre-training-free neural network for processing to yield sharper edges. We implemented the proposed method through simulations and experiments, demonstrating its ability to successfully recover the edges of target objects at lower compression ratios than traditional methods. This method outperforms some widely used edge detection methods based on ghost imaging in terms of signal-to-noise ratio. The neural network used in this method does not require pre-training and exhibits good generalization capability.