By Colbi Edmonds
Boston University ECE professor Vivek Goyal was awarded a 2019 IEEE Signal Processing Society Best Paper Award for his work on “Photon-Efficient Computational 3-D and Reflectivity Imaging With Single-Photon Detectors” paper. The paper, which was co-authored with Goyal’s students Dongeek Shin and Ahmed Kirmani along with Jeffrey Shapiro of MIT, was published in the IEEE Transactions on Computational Imaging in 2015. The award is for papers of “exceptional merit” that appeared in any of the IEEE Signal Processing Society’s journals between 2013 and 2018. The award will be publicly presented at the 45th IEEE International Conference on Acoustics, Speech, and Signal Processing, a flagship conference of the Society. The conference and award presentation will be held remotely for the first time on May 4-9, 2020.
Goyal’s paper describes models, methods and a thorough evaluation of high-quality 3-D and reflectivity imaging with extreme photon efficiency of about 1 detected signal photon per pixel despite significant ambient light. This is in contrast to traditional methods that require hundreds of detected photons per pixel. For years, detectors with single-photon sensitivity have been used to measure histograms of detection times, which are subsequently processed as noisy discrete-time versions of the received waveform. Goyal’s work shows that photon-by-photon modeling combined with appropriate regularization and a method for approximately unmixing the signal and noise components in the detections can provide a 20-fold depth error reduction over state-of-the-art results.
The awarded paper is just one in a sequence of works from Goyal’s team that have upended the thinking of what is possible in photon-efficient imaging, but the innovation does not stop here.
In another recent research thrust, Goyal is questioning a bit of conventional wisdom. It is generally believed that if one wants to use the timings of detections of single photons — such as for 3D imaging — it is important to make the detections infrequent enough that detector dead times are unimportant. Goyal’s team showed that the estimation biases that occur because of detector dead times can be completely avoided (by using the right mathematics and algorithms), so the effects of dead times do not have to be avoided. If it isn’t important to avoid dead times, then the image collection can be a lot faster, in some cases as much as 20 times faster. To build upon this idea, Goyal recently won a Google Research Award for a project entitled “Information Processing Foundations for Single-Photon Detectors.”
To date, Professor Goyal has won several awards for his outstanding research, such as the 2017 IEEE Signal Processing Society Best Paper Award and six conference best paper awards. He is an IEEE and OSA Fellow.