
Researchers from NJU and PKU have developed a lensless imaging method using deep learning that reconstructs high-fidelity dynamic images from single shots.
Key Details
- 1New 'McLDI-INR' framework integrates physical models with implicit neural representation for lensless imaging.
- 2Overcomes traditional need for multiple acquisitions and strong priors in dynamic (moving) imaging scenarios.
- 3Uses a binary mask and dual-domain loss (spatial and frequency) to improve image quality and reduce motion blur.
- 4Demonstrated improved reconstruction of moving objects (including biological samples) in both simulation and real-world experiments.
- 5Published in Intelligent Opto-Electronics, June 2026; supported by multiple Chinese science foundations.
Why It Matters
This approach could enable more compact and practical imaging systems, particularly benefiting portable microscopy, dynamic biological observation, and potentially point-of-care medical diagnostics. High-resolution single-shot reconstruction of moving targets is a longstanding challenge, making this advance significant for next-generation imaging AI and device design.

Source
EurekAlert
Related News

•EurekAlert
Deep Learning and MRI Identify Cognitive Impairment in CKD Patients
Researchers used multimodal MRI and a novel deep learning model to identify cognitive impairment in chronic kidney disease patients with strong validation.

•EurekAlert
Advances in 3D-Printed Devices for Structured-Light Imaging
Researchers review femtosecond laser 3D microfabrication for integrated structured-light optical devices.

•EurekAlert
HKU Unveils X-ray Telescope and AI Innovations at InnoCarnival 2026
Hong Kong University showcases an X-ray telescope and AI-powered health technologies at InnoCarnival 2026.