Probing Light Transport for 3D Shape

Matthew O'Toole: There is a rising demand for high-performance 3D sensors in response to the rapid development of autonomous cars, 3D printers, & virtual/augmented reality systems.

February 19, 2018
3:30 pm - 4:30 pm
Location
Kemeny Hall 007
Sponsored by
Computer Science Department
Audience
Public
More information
Sandra Hall

There is a rising demand for high-performance 3D sensors in response to the rapid development of autonomous cars, 3D printers, and virtual/augmented reality systems.  These sensors often make use of controllable light sources to send light signals into an environment, and cameras to measure the signal reflected back in response.  This approach can, however, fail in critical scenarios where objects with complex material properties are present, when imaging objects under overwhelmingly bright sunlight, or when the object of interest is hidden behind an occluder.

In this talk, I will address these key challenges by presenting a new family of computational cameras that explicitly control what light paths contribute to an image.  First, I will identify the crucial link between stereo geometry and light transport used to attenuate the contribution of multiply-scattered light paths which make 3D imaging hard.  Second, I will show how this link can be further exploited to optimize the energy-efficiency of these camera systems, which enables 3D imaging under strong ambient lighting.  Finally, I will explain how sampling a specific set of light paths leads to the derivation of an efficient, closed-form solution for reconstructing images of objects hidden from view.

Bio: Matthew O’Toole is a postdoctoral scholar with the Department of Electrical Engineering at Stanford University.  His research focus is on computational imaging, a highly multi-disciplinary topic that makes use of novel combinations of computation, electronics, and optics to overcome the limitations of conventional imaging systems.  He completed his Ph.D. at the University of Toronto in 2016, and his thesis received the ACM SIGGRAPH Outstanding Dissertation Honorable Mention award in 2017.  His research accolades also include two runner-up best paper awards (CVPR 2014, ICCV 2007) and two best demo awards (CVPR 2015, ICCP 2015).  He co-organized two workshops on Computational Cameras and Displays at CVPR 2016 and 2017, and a course by the same name at SIGGRAPH 2014.  He is supported by a Banting Postdoctoral Fellowship from the Government of Canada.

Location
Kemeny Hall 007
Sponsored by
Computer Science Department
Audience
Public
More information
Sandra Hall