Professor Goyal receives a Best Paper Award for 3D sensing method which uses probabilistic modeling to slice through the Gordian Knot of a data transfer bottleneck.
by A.J. Kleber
A knack for unconventional thinking around computational imaging challenges has earned Professor Vivek Goyal and his colleagues a prestigious Best Paper Award at the 2026 IEEE International Conference on Computational Photography (ICCP).
The paper, titled “High-Flux Count-Free Single-Photon 3D Cameras,” addresses two primary challenges with a promising 3D sensing method. Single-photon cameras (SPCs) have the capacity to detect the smallest units of light, as the name implies, making them highly sensitive and very fast. However, their sizable data transfer requirements can cause a severe bottleneck between detection and processing, and in high-flux scenarios (where a large amount of light is present), distortions occur as a result of the “dead time” intervals after each photon detection event.
Goyal’s paper presents innovations capable of solving the data bottleneck while avoiding additional dead time inaccuracies. The use of free-running capture, where the detector is activated as soon as possible after each delay, makes the most efficient possible use of the available photons; in addition, swapping a traditional “counting” method of data transmission for the use of estimated quantiles reduces the data rate. The key challenge with this approach is to interpret the resulting, unusual data format.
Compressive capture techniques can exacerbate the distortion issue, as they fail to retain sufficient historical data for post-capture correction using conventional methods. For this reason, the paper presents an alternative computational approach rooted in analysis of the convergence of quantile estimators when fed detection times that have been distorted by dead times. The method estimates scene distances and reflectances by finding the combinations that are most consistent with the data under this modeling. (In simpler terms, the researchers make complex calculations based on probability, using the estimations derived from the operation of the camera.)
Through hardware emulation and full-scene and single-pixel simulations, this approach is shown to reliably capture distance and reflectance across a wide range of illumination conditions, making high-resolution single-photon cameras which can operate under severe bandwidth constraints in real-world high-flux scenarios. The approach Goyal and his colleagues have developed could hold the key to making 3D sensing in portable devices more common, as they could be manufactured more cheaply and require far less power to operate.
However, the real triumph of a breakthrough like this one, according to Goyal, is that through attempting to address a very specific issue, a novel technique is developed; one with the potential for much more general use. “My coauthors and I are excited to continue to develop the concept of quantile sensing, first for other uses of time-correlated single-photon counting, and then more broadly.”
Professor Vivek Goyal’s research interests revolve around computational imaging, information representation and signal processing. A Fellow of the AAAS, IEEE and Optica, he is the recipient of a 2025 BU College of Engineering Dean’s Catalyst Award, a 2024 John Simon Guggenheim Memorial Foundation Fellowship, several best paper awards, and a 2023 Frontiers of Science Award in Computational Optics, among other accolades.