Imaging and Computing Seminar
Vivek Goyal, Electrical Engineering and Computer Science, MIT
Compressive Depth Acquisition Cameras: Principles and Demonstrations
Measuring time elapsed from transmitting a pulse to receiving a reflected response is a standard method for distance estimation. Light detection and ranging (LIDAR) systems and time-of-flight (TOF) cameras use this principle along with scanning by the illumination source or 2D sensor arrays to acquire depth maps. This talk introduces depth map acquisition with high spatial and range resolution using a single, omnidirectional, time-resolved photodetector and no scanning components. In contrast to compressive photography, the information of interest -- scene depths -- is nonlinearly mixed in the measured data. The reconstruction uses parametric signal modeling to recover a set of depths present in the scene. Then, a convex optimization formulation that exploits sparsity of the Laplacian of the depth map of a typical scene is used to determine correspondences between spatial positions and depths.
We have demonstrated depth map reconstruction for both near and medium-range scenes even in low light conditions. An initial prototype uses patterned illumination created by a 64-by-64-pixel spatial light modulator. With this prototype, we constructed depth maps of scenes comprising two to four planar shapes using only 205 spatially-patterned, pulsed illuminations of the scene. A second apparatus achieves spatially-patterned reception with a digital micromirror device and a single photon-counting detector. With the same mathematical framework, we constructed depth maps with and without the presence of a partially-transmissive occluder.
Joint work with Ahmed Kirmani, Andrea Colaco, Greg Howland, John Howell, and Franco Wong.