Single-Shot Lensless Imaging with Physics Guided Genetic Programming

TL;DR

This paper presents a novel single-shot lensless imaging method that uses a physics-guided genetic programming algorithm to reconstruct complex objects from a single intensity measurement, enhancing robustness and scalability.

Contribution

It introduces a genetically programmed iterative algorithm that couples wave propagation with adaptive meta-optimization for high-fidelity, out-of-distribution object reconstruction in lensless imaging.

Findings

Successfully reconstructs amplitude objects like USAF targets and silicon beads.

Achieves high-quality phase and amplitude reconstruction across various wavelengths and distances.

Demonstrates practical application in a portable optical digital bead assay.

Abstract

Lensless optical imaging eliminates the need for refractive optics, enabling compact and low-cost cameras with a large field-of-view, supporting point-of-care diagnostics and industrial monitoring. Practical deployments, however, remain constrained by ill-posed image reconstruction pipelines that require multiple measurements, careful calibration or object-specific training, thus limiting robustness and scalability. In this work, we introduce a single-shot lensless imaging framework that reconstructs complex objects from only a single recorded intensity pattern using a genetically programmed iterative algorithm. Our method couples a wave-propagation model with an adaptive meta-optimisation strategy to jointly estimate the object amplitude, object phase, and effective object-detector distance. Experiments demonstrate high-fidelity recovery of amplitude objects, including a USAF target and 2~$\mu$m silicon beads on a glass slide, as well as a phase-dominant biological sample consisting of U2OS cells on a glass slide. Across multiple object types, wavelengths, and propagation distances, the same learned policy maintains high reconstruction quality with minimal retuning, indicating strong out-of-distribution generalisation. As a practical demonstration, the framework is integrated with a $\beta$-amyloid-based optical digital bead assay under wide field-of-view acquisition. The resulting platform combines single-shot capture, compact hardware, and accurate reconstruction of complex fields, enabling rapid, portable assays in which throughput, alignment tolerance, and cost are critical.