Superhuman Vision

Sonia Shum '27

Eyes play a major role in the lives of humans. They enable nuanced perception and interpretation of the surrounding world, and beyond physical sight, they are a pillar of social interaction, attracting researchers in neuroscience and psychology. As vehicles for life experiences and windows into human behaviour, eyes have been subject to tracking technologies to determine eye movement and gaze direction. A recent discovery by researchers at the University of Arizona Wyant College of Optical Sciences developed a new and improved human eye-tracking technique, involving deflectometry and advanced computation.

Currently, eye-tracking methods only capture “directional information of the eyeball from a few sparse surface points, about a dozen at most,” said Florian Willomitzer, principal investigator of the study (University of Arizona, 2025). One example is glint tracking, a method known as pupil-center corneal reflection (PCCR), in which a near-infrared light source illuminates the eye and casts a reflection on the cornea. By tracking this reflection relative to the position of the pupil’s center, an estimation of gaze direction can be made (Miseviciute, n.d.). While this method is accurate, it measures only a small surface spot on the cornea, relying on and producing only limited data points. The new deflectometry technique developed by Willomitzer’s lab improves current eye-tracking methods, increasing efficiency, data quantity, and accuracy.

Deflectometry is a 3D imaging technique for reflective surfaces, commonly used to scan high-performance optics, such as large telescope mirrors. Simply put, the reflection of a known grid of light, typically displayed on a screen, off a reflective 3D surface, is captured by a camera. The recorded image can be analysed to reveal information about the object’s surface based on how the surface distorts the known light pattern. The specular surface acts as “an optical device in the ray path between [a] camera and a screen” (Balzer et al., 2011). By applying knowledge of light trajectories, scientists can derive information about the object’s surface (Balzer et al., 2011). Since the millions of pixels on the screen each act as a light source, the camera can capture an image with this high number of pixels, each depicting the light reflecting off the specular surface, producing dense data which increases the accuracy of estimations derived from data analysis (University of Arizona, 2025).

Deflectometry has many different aspects in various applications. In the case of eye-tracking, human eyes are the target specular surface due to their reflective quality. Willomitzer’s group pointed a screen with a known light grid at the eye, recording reflections of the light off the eye with a stereo camera which captures the data in one instantaneous image (University of Arizona, 2025). In the research experiment, a subject observed a red cross move along a screen while a stereo camera tracked light reflection, capturing more than 40,000 surface points on the subject’s eye, which is 3,000 times the number of acquired data points by conventional approaches (University of Arizona, 2025). Willomitzer’s team analysed the data using advanced computational methods commonly used in vision research. First, they accurately reconstructed the dense 3D surface of the eye by analysing distortions in the reflected light patterns, focusing especially on the cornea and sclera. Next, Jiazhang Wang, the study’s first author, explained that computational reconstructions “use[d] this surface data together with known geometrical constraints about the eye’s optical axis to accurately predict the gaze direction” (University of Arizona, 2025). Further, the researchers used a newly developed surface optimization algorithm to avoid making strong assumptions about the shape or surface of the measured eye, since each human eye varies slightly. By combining deflectometry with these computational advances, the team achieved impressive tracking accuracies ranging from 0.46º to 0.97º for human eyes and around 0.1º error for the artificial eye models. Willomitzer reported how this “combination of precise measurement techniques and advanced computation allows machines to ‘see the unseen,’ giving them ‘superhuman vision’ beyond the limits of what humans can perceive,” (University of Arizona, 2025).

Beyond demonstrating technical excellence, the team’s innovation has significant practical implications. Their improved eye-tracking technology can be integrated with virtual and augmented reality systems, improving user experience and reducing system complexity. By embedding a fixed light pattern into a headset frame or visual content, deflectometry data of light reflection off the eye can be obtained and analysed for gaze direction. A further development would use infrared light patterns to avoid visual interference from visual light grids (University of Arizona, 2025). Moreover, eye-tracking techniques can be useful in industrial engineering, such as automotive driving assistance, along with improving the analysis of artwork and paintings as well as tablet-based 3D imaging methods to measure the shape of skin lesions (Burkhart, 2025). In addition to practical applications, dense information about the cornea and sclera’s surfaces can be obtained using this novel technology, assisting human eye research and opening the door for medical applications. Researchers could potentially develop more accurate reconstructions of the human eye’s surface and apply findings to advance understandings of eye disorders (University of Arizona, 2025).

Given how the deflectometry approach to eye-tracking produced promising results, Willomitzer’s group now plans to commercialise this technology at Tech Launch Arizona and looks to apply their innovation to real-world products in industry (Willomitzer, 2025). Looking ahead, the team seeks to incorporate artificial intelligence and explore other 3D imaging methods to improve their current approach. They also hope to start a new wave of next-gen eye-tracking technology and expand applications to neuroscience and psychology research. As Willoitzer emphasised, their work “highlights the versatility of optical technologies, which we hope will benefit the optics community as a whole” (Burkhart, 2025).


References

Balzer, J., Hofer, S., & Beyerer, J. (2011, June). Multiview specular stereo reconstruction of large mirror surfaces. Research Gate. Retrieved April 23, 2025, from
https://www.researchgate.net/figure/Basic-principle-of-deflectometry-The-specular-su rface-is-quasi-invisible-to-the_fig1_221363507
Burkhart, F. (2025, April 1). Arizona sharpens imaging for eye-tracking. Optics.org. Retrieved April 23, 2025, from https://optics.org/news/16/4/5
Miseviciute, I. (n.d.). How do eye trackers work? Tobii. Retrieved April 23, 2025, from https://www.tobii.com/resource-center/learn-articles/how-do-eye-trackers-work University of Arizona. (2025, April 1). 'Superhuman vision': Powerful 3D imaging technology paves way for next-generation eye-tracking. Phys.org. Retrieved April 23, 2025, from
https://phys.org/news/2025-03-superhuman-vision-powerful-3d-imaging.html Wang, J., Wang, T., Xu, B., Cossairt, O., & Willomitzer, F. (2025). Accurate eye tracking from dense 3D surface reconstructions using single-shot deflectometry. Nature Communications, 16(1). https://doi.org/10.1038/s41467-025-56801-1
Willomitzer, F. (2025, April). Method and System for Eye Tracking using Deflectometric Information. Tech Launch Arizona. Retrieved April 23, 2025, from
https://inventions.arizona.edu/tech/Method_and_System_for_Eye_Tracking_using_D eflectometric_Information