Related Experiment Video
Updated: Jun 11, 2026
07:14
Microfabrication of Implantable Optics Integrated in a Microstructured Imaging Window for Advanced In Vivo Imaging
Published on: April 11, 2025
Deep-learning reconstruction of aspheric microlens arrays from intensity maps
Applied Optics
|June 10, 2026
Summary
This study uses deep learning to reconstruct complex aspheric surfaces of microlens arrays from light intensity data. The novel approach enables accurate surface metrology for advanced optical elements.
Area of Science:
- Optics and Photonics
- Artificial Intelligence
- Materials Science
Background:
- Designing complex aspheric surfaces for precise light control is a significant challenge in optics.
- Current metrology techniques struggle with the intricate geometries of aspheric and freeform optical elements.
Purpose of the Study:
- To develop a deep-learning-based method for reconstructing dielectric aspheric microlens arrays.
- To address the limitations in achieving precise control over light paths in complex optical designs.
Main Methods:
- Formulated as an image-to-image translation task using deep convolutional neural networks.
- Trained the network to convert light intensity maps (from multi-angle, multi-wavelength illumination) into surface spatial maps.
- Employed physically accurate rendering software and a microlens array template for training and validation.
Main Results:
- Achieved highly accurate reconstructions of aspheric surfaces.
- Demonstrated stable reconstruction of complex surfaces from intensity-only measurements.
- Validated the efficacy of the deep learning approach using a microlens array template.
Conclusions:
- Deep convolutional networks can reliably reconstruct aspheric surfaces from intensity data.
- This data-driven approach offers a practical route for surface metrology of microlens arrays and refractive elements.
- Highlights the potential of deep learning to advance aspheric and freeform optics design and manufacturing.
