Implicit representation optimization method for enhanced ghost imaging through scattering medium

In this paper, we proposed an optimization method for Ghost Imaging (GI) through scattering medium. Traditional GI systems encounter challenges in image quality when scattering medium is present between the light source and the object. To address this issue, we framed the GI image reconstruction as a maximum likelihood estimation problem and implicitly represent the scattering medium through nonlinear fitting of Zernike basis functions. We explored three methods for representing the scattering medium: neural representation, Zernike basis function representation, and matrix representation. Our results indicated that the neural representation is particularly effective at fitting nonlinear combinations of Zernike basis functions and demonstrated superior performance. Compared to other GI reconstruction techniques, our method successfully reconstructed the object’s image even in the presence of a scattering medium. We validated our method through numerical simulations and physical experiments, demonstrating its potential for image reconstruction in scattering environments under GI architecture.