Physics-Aware Style Transfer for Adaptive Holographic Reconstruction

2025-07-01

AI Summary

In inline holographic imaging, reconstructing the object’s complex amplitude from intensity-only diffraction patterns constitutes a severely ill-posed inverse problem. Existing deep learning approaches rely on high-fidelity ground-truth complex-amplitude labels, limiting their practical deployment. To address this, we propose a physics-informed style transfer framework that, for the first time, models object distance as an implicit style variable encoded in the diffraction pattern. By establishing a cycle-consistent image translation pathway mediated through a style domain, our method enables end-to-end, label-free inverse mapping learning using intensity data alone. Integrating the optical Fresnel diffraction model with a cycle-consistent generative adversarial network, the framework achieves adaptive and generalizable complex-amplitude reconstruction. Experiments demonstrate high accuracy and robustness in label-free, real-time imaging of dynamic red blood cells, significantly reducing reliance on annotated data and enhancing practicality and deployability in biomedical applications.

Technology Category

• Computer Vision: Medical and Biological Imaging
• Search and Optimization: Learning to Search
• Machine Learning: Deep Generative Models & Autoencoders

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Inline holographic imaging presents an ill-posed inverse problem of reconstructing objects' complex amplitude from recorded diffraction patterns. Although recent deep learning approaches have shown promise over classical phase retrieval algorithms, they often require high-quality ground truth datasets of complex amplitude maps to achieve a statistical inverse mapping operation between the two domains. Here, we present a physics-aware style transfer approach that interprets the object-to-sensor distance as an implicit style within diffraction patterns. Using the style domain as the intermediate domain to construct cyclic image translation, we show that the inverse mapping operation can be learned in an adaptive manner only with datasets composed of intensity measurements. We further demonstrate its biomedical applicability by reconstructing the morphology of dynamically flowing red blood cells, highlighting its potential for real-time, label-free imaging. As a framework that leverages physical cues inherently embedded in measurements, the presented method offers a practical learning strategy for imaging applications where ground truth is difficult or impossible to obtain.

Problem

• Reconstructing objects' complex amplitude from diffraction patterns
• Learning inverse mapping without high-quality ground truth datasets
• Adaptive holographic reconstruction for biomedical imaging

Innovation

• Physics-aware style transfer for holography
• Cyclic image translation without ground truth
• Adaptive learning from intensity measurements