Efficiently reconstructing compound objects by quantum imaging with higher-order correlation functions

Alexander Mikhalychev, Bänz Bessire, Ilya Karuseichyk, Anton Sakovich, Manuel Unternährer, Dmitry Lyakhov, Dominik L. Michels, André Stefanov, D. MogilevtsevCommunications 2019 · DOI

Abstract

Quantum imaging offers a path to enhancing reconstruction precision by exploiting the quantum correlations of imaging fields, particularly in low-intensity regimes. However, reconstructing images from higher-order correlations typically results in daunting nonlinear estimation problems that scale poorly with the number of parameters. We propose a "Sliding-Window Method" (SWM) that drastically reduces this complexity. By connecting the features of the Fisher Information Matrix (FIM) with the parametric locality of the problem, we reconstruct parameters stepwise through an efficient iterative inference scheme that scales linearly with the total number of pixels. This approach enables super-resolving reconstruction of complex grey-scale objects even at the few-photon level.

Introduction

Quantum imaging leverages the statistical correlations between photons—rather than just average intensity—to resolve objects with higher precision than classical optics allows. This is vital for applications like biological imaging, where low light levels are necessary to prevent sample damage.

Despite its potential, the field faces a "computational wall." Reconstructing an image from higher-order correlation functions (such as $G^{(2)}$ or $G^{(3)}$) is a nonlinear optimization challenge. As the number of pixels (parameters) increases, the computational effort required to solve these equations typically explodes, often negating the practical benefits of the quantum approach.

Schematic of the quantum imaging system showing the source, object, and image plane where correlation functions are measured.

The Sliding-Window Method (SWM)

To overcome this bottleneck, we introduced the Sliding-Window Method (SWM). This iterative inference scheme relies on the concept of parametric locality. By analyzing the Fisher Information Matrix (FIM), we discovered that in many near-field imaging geometries, a specific measurement is only strongly influenced by a small, local neighborhood of pixels. This results in a "banded" FIM, where information is spatially localized.

Instead of solving for every pixel simultaneously, the SWM focuses on a small "core" of pixels surrounded by a "border." The algorithm optimizes the core parameters and then "slides" the window across the entire object. This transformation turns a global nonlinear problem into a sequence of local optimizations, ensuring the computational cost scales linearly ($O(M)$) with the number of pixels $M$.

The SWM process (f) and reconstruction results (h) showing how the iterative windowing and refinement procedure converges on the true object profile.

Key Results and Insights

Experimental validation using pseudo-thermal light and entangled photons demonstrated several breakthroughs:

  • Super-Resolution: The SWM successfully reconstructed 1D and 2D objects with resolution significantly exceeding the Rayleigh limit. For example, a digit "5" was reconstructed using third-order correlation functions ($G^{(3)}$), proving the method's efficacy with higher-order quantum data.
  • Computational Efficiency: The linear scaling makes high-resolution quantum imaging computationally feasible for the first time, even for complex "compound" objects.
  • The "Goldilocks" Correlation: We identified an optimal correlation width for the light source. If photons are too perfectly correlated or too independent, resolution suffers; the best results occur when the correlation width matches the scale of the smallest desired detail.
  • Constraint Bias: For binary (black and white) objects, the physical constraint that pixel values must lie between 0 and 1 acts as a "bias" that helps the algorithm push even further beyond classical resolution limits.

Conclusion and Future Work

The Sliding-Window Method provides a robust framework for practical quantum imaging. By exploiting the local nature of information in light fields, we have bypassed the traditional computational barriers of nonlinear reconstruction. Looking forward, this informational approach could be adapted for quantum state tomography and quantum diagnostics, helping researchers understand high-dimensional quantum systems where the "curse of dimensionality" currently limits exploration.

Summary updated

Abstract

Abstract Quantum imaging has a potential of enhancing the precision of objects reconstruction by exploiting quantum correlations of the imaging field, in particular for imaging with low-intensity fields up to the level of a few photons. However, it generally leads to nonlinear estimation problems. The complexity of these problems rapidly increases with the number of parameters describing the object and the correlation order. Here we propose a way to drastically reduce the complexity for a wide class of problems. The key point of our approach is to connect the features of the Fisher information with the parametric locality of the problem, and to reconstruct the whole set of parameters stepwise by an efficient iterative inference scheme that is linear on the total number of parameters. This general inference procedure is experimentally applied to quantum near-field imaging with higher-order correlated light sources, resulting in super-resolving reconstruction of grey compound transmission objects.
BibTeX
@inproceedings{mikhalychev2019efficiently,
  author = {Alexander Mikhalychev and Bänz Bessire and Ilya Karuseichyk and Anton Sakovich and Manuel Unternährer and Dmitry Lyakhov and Dominik L. Michels and André Stefanov and D. Mogilevtsev},
  title = {{Efficiently reconstructing compound objects by quantum imaging with higher-order correlation functions}},
  booktitle = {Communications},
  year = {2019},
  doi = {10.1038/s42005-019-0234-5},
}
Figure