What is it about?
When light travels through turbulent air, biological tissue, or other obscurants, its wavefront becomes distorted, like looking through a frosted window. This causes cameras to capture blurry images. Fixing this in real time has until now always required either a bright reference point in the scene, known as a guidestar, or a dedicated wavefront sensor. This paper presents the first system to remove both requirements while still correcting aberrations optically and in real time. A triangle-shaped asymmetric aperture is placed in the optical path and paired with two machine learning algorithms. These algorithms estimate wavefront distortion directly from ordinary natural scene images and immediately drive a spatial light modulator to optically correct the distortion. All of this is achieved without any guidestar, wavefront sensor, or controlled illumination.
Featured Image
Photo by Logan Voss on Unsplash
Why is it important?
Every existing real-time adaptive optics (AO) system depends on either a guidestar or a dedicated wavefront sensor to measure distortion, and without one, real-time correction was considered out of reach. We are the first to break this barrier. Our closed-loop system corrects optical aberrations in real time using only natural scene images captured through an asymmetric aperture, with no guidestar and no wavefront sensor. Compared to state-of-the-art guidestar-free methods, our approach uses an order of magnitude fewer measurements and three orders of magnitude less computation. This makes real-time adaptive optics practical for settings where guidestars are unavailable, including ground-based telescopes imaging through atmospheric turbulence, biomedical imaging through tissue, and robust computer vision in adverse weather.
Perspectives
The long-standing assumption in adaptive optics (AO) has been that real-time wavefront correction requires either a guidestar or a dedicated sensor, and the optics community has worked around this constraint for decades. What excites me most about this work is that we show the constraint itself can be lifted. By treating wavefront sensing as a phase retrieval problem over natural scene measurements and enabling it through a simple asymmetric aperture, we achieve something that was previously thought to require specialized hardware. Our system performs real-time, closed-loop optical correction with no guidestar and no wavefront sensor. I hope this will enable new research directions in (1) seeing through atmospheric turbulence with ground-based telescopes to rival space telescopes at a fraction of the cost, (2) seeing through tissue to observe biology at the cellular scale, (3) seeing through fog and smoke for robust computer vision in adverse weather conditions, and (4) seeing through fiber bundles to enable minimally invasive endoscopy.
Weiyun Jiang
Rice University
Read the Original
This page is a summary of: Guidestar-Free Adaptive Optics with Asymmetric Apertures, ACM Transactions on Graphics, April 2026, ACM (Association for Computing Machinery),
DOI: 10.1145/3809501.
You can read the full text:
Resources
Contributors
The following have contributed to this page
