Unified probability explanation for ghost imaging with thermal light

TL;DR

This paper introduces a probability model that explains ghost imaging with thermal light, showing that pixel value distributions follow a Gaussian pattern, unifying understanding across different correlation methods.

Contribution

It provides a unified probabilistic explanation for ghost imaging mechanisms, applicable to various correlation functions and demonstrating the Gaussian distribution of pixel values.

Findings

Pixel values follow Gaussian distribution regardless of reference pattern form.

Variance of pixel values explains reconstruction noise.

Linear relationship between reconstructed means and original gray values.

Abstract

Ghost imaging (GI) is an intriguing imaging technology which achieves the object images through intensity correlation between reference patterns and bucket signal. Here, we propose a probability model to explain the imaging mechanism of this modality, by assuming that the reference patterns fulfill an arbitrary identical distribution and that the objects are of gray-scale. We have proven that the probability of the reconstructed pixel values in the pixel region of the same original gray value obeys a Gaussian distribution, no matter which functional form of the reference patterns is used in correlation calculation. Both simulation and experiments have demonstrated that the probability of recovered pixel values are highly consistent with their Gaussian theoretical distribution, while their variance explains the appearance of reconstruction noise. In addition, we have also extend this theory to other classic correlation functions, e.g., normalized GI and differential GI. The results have shown that there is a linear relationship between reconstruction means in specified pixel regions and original gray values, which might provide a unified explanation for GI with thermal light.