Abstract




In conventional imaging systems, the results are poor unless there is a physical mechanism for producing a sharp image with high signal-to-noise ratio.  In this talk, I will present two settings where computational methods enable imaging from very weak signals:  range imaging and non-line-of-sight (NLOS) imaging.


 


Lidar systems use single-photon detectors to enable long-range reflectivity and depth imaging.  By exploiting an inhomogeneous Poisson process observation model and the typical structure of natural scenes, first-photon imaging demonstrates the possibility of accurate lidar with only 1 detected photon per pixel, where half of the detections are due to (uninformative) ambient light.  I will explain the simple ideas behind first-photon imaging and lightly touch upon related subsequent works that mitigate the limitations of detector arrays, withstand 25-times more ambient light, allow for unknown ambient light levels, and capture multiple depths per pixel.


 


NLOS imaging has been an active research area for almost a decade, and remarkable results have been achieved with pulsed lasers and single-photon detectors.  Our work shows that NLOS imaging is possible using only an ordinary digital camera.  When light reaches a matte wall, it is scattered in all directions.  Thus, to use a matte wall as if it were a mirror requires some mechanism for regaining the one-to-one spatial correspondences lost from the scattering.  Our method is based on the separation of light paths created by occlusions and results in relatively simple computational algorithms.