In brief

The algorithm performed well on both simulated patterns and real-world data, producing high quality reconstructed images.

© A*STAR Research

Imaging with computational lenses

17 Jun 2022

A modified image processing algorithm, created by A*STAR scientists, paves the way toward super-resolution imaging.

A picture paints a thousand words, and this is true of all forms of imaging ranging from the X-rays that are used to capture signs of pneumonia in human lungs, to visible light used in lab microscopes, and even electron beams used to image viruses. However, some computer imaging experts say that today’s imaging technologies have yet to reach their full potential.

At A*STAR’s Singapore Institute of Manufacturing Technology (SIMTech), Joel Yeo and a team of researchers are working to take digital X-ray imaging to new heights with the help of computational lenses, specialised algorithms that process and boost the resolutions of laser images. Better imaging is a boon to many industries from medicine to manufacturing, enhancing the speed and reliability of detection. Known as phase retrieval, this computer-aided imaging processing step has, however, proven challenging to optimise.

“Current applications of phase retrieval in research are still unable to recover the highest resolution possible,” said Yeo, adding that high resolution image quality relies on the use of costly, energy-hungry supercomputers.

Yeo’s team hypothesised that the key to achieving clearer, sharper laser images with minimal computational resources lay in a set of equations called the Fresnel propagator function. By incorporating these equations into existing phase retrieval algorithms, the team were able to reconstruct captured images with pixel sizes a few times smaller than previously reported.

The team was also able to zoom into the captured images to reveal fine details without the need for expensive optical components such as lenses and mirrors. Furthermore, with the new algorithm, the resolution of captured images was no longer limited to the pixel size of the detector—they were instead able to obtain pixel sizes of around one micrometre, even with a camera with large 13-micrometre pixel pitches.

“We tested our modified phase retrieval algorithm using simulated diffraction patterns, as well as on real-world data,” Yeo explained. “We were pleasantly surprised by the high quality of the reconstructed image, a clear indication that the enhanced algorithm could accurately model the physics of image formation.”

By reconstructing images with unprecedented clarity, this imaging innovation holds promising benefits for medical diagnostics, research and manufacturing. According to Yeo, the approach can also be used to improve applications that use laser imaging.

Next, Yeo and colleagues plan to apply their phase retrieval framework to enhance images captured via electron microscopy. “Being able to observe finer structures in proteins and molecules would greatly help researchers reveal the unknown and undiscovered properties of these nano-scaled objects,” said Yeo.

The A*STAR researchers contributing to this research are from the Singapore Institute of Manufacturing Technology (SIMTech).

Want to stay up to date with breakthroughs from A*STAR? Follow us on Twitter and LinkedIn!


Yeo, J., Seck, H.L., and Zhang, Y. Super-resolution, multi-plane phase retrieval via amplitude flow variants. Optics and Lasers in Engineering 146, 106715 (2021) │ article

About the Researcher

Joel Yeo is currently pursuing a PhD degree in the field of computational imaging and optics at the National University of Singapore (NUS). In 2018, he joined Nanyang Technological University and found new exotic propagation modes of light travelling through inhomogeneous media. He then joined the Singapore Institute of Manufacturing Technology’s (SIMTech) Advanced Imaging and Machine-Vision group in 2019 as a research engineer, where he derived a new model for efficient and selective generation of high-powered UV lasers and also developed phase retrieval algorithms for lensless computational imaging. In 2021, he was awarded the A*STAR Graduate Scholarship to pursue further studies.

This article was made for A*STAR Research by Wildtype Media Group