Non-Line-of-Sight Imaging

The critical challenge of non-line-of-sight (NLOS) imaging is that diffuse reflections scatter light in all directions. The most effective existing methods use time-resolved acquisition to measure the transient response of a scene to a pulsed light source, but require large scanning large virtual apertures with many measurement points in order to perform reconstructions.

We propose a method for seeing around corners that uses visible vertical edges to recover directional information, combined with transient measurements. Our method enables the imaging of large-scale scenes with a large field-of-view, despite a small scanning aperture and few measurement locations.

Work with Charles Saunders, Julián Tachella, John Murray-Bruce, Yoann Altmann, Jean-Yves Tourneret, Stephen McLaughlin, Robin Dawson, Franco Wong & Vivek Goyal

Dead-time effects in Single-photon Lidar

Single-photon detectors used in time-correlated single photon counting (e.g., for lidar, fluorescence lifetime imaging) have a dead time after each photon detection, which blocks registration of subsequent photons arriving within that dead time and causes a distortion of the detection time distribution. Our work enables fast data acquisition with dead time-limited detectors by deriving estimators that depend on accurately modeling detection time sequences as Markov chains.

Work with Yanting Ma, Robin Dawson & Vivek Goyal

Dithered depth measurement

In single photon detection systems (especially SPAD arrays) that have time bins longer than typical laser pulse durations, the resulting measurement errors are dominated by quantization. Our work proposes an optical time-of-flight system that uses subtractive dither to improve image depth resolution and investigates how modeling the measurement noise with a generalized Gaussian distribution can further improve estimation error.

Work with Robin Dawson & Vivek Goyal

Robust single-photon lidar

Recent photon-efficient computational imaging methods are remarkably effective with as little as 1 detected photon per pixel, but they are not demonstrated at signal-to-background ratio (SBR) below 1.0 because their imaging accuracies degrade significantly in the presence of high background noise. We introduce a new approach to depth and reflectivity estimation that emphasizes the unmixing of contributions from signal and noise sources and has defined the state-of-the-art performance for SBR as low as 1/25.

Work with Vivek Goyal