Join us online!
The potential of hyperspectral imaging in many application domains is severely hindered by the challenges in acquiring them. A typical capture has to deal with two challenges: poor signal-to-noise ratio due to low photon counts at a spatio-spectral voxel, and the large signal dimensionality that increases acquisition and reconstruction times. In this talk, I will show that both of these challenges can be circumvented by adopting a holistic approach to sensing and processing, where-in we measure the most relevant components of the hyperspectral image. At the heart of these computational imaging designs are two key ideas: first, a specialization to the particular scene being sensed using an adaptive measurement strategy, and second, performing key computations in the optical domain prior to sensing. This dramatically reduces the number of measurements required, while maintaining robustness to noise as well as reducing the computational costs associated with reconstructing high-resolution hyperspectral images.
Aswin C. Sankaranarayanan is an Associate Professor in the ECE department at CMU. His research interests are broadly in computational imaging, signal processing and computer vision. Aswin did his doctoral research at the University of Maryland where his dissertation won the distinguished dissertation award from the ECE department in 2009. He is the recipient of the CVPR 2019 best paper award, the CIT Dean's Early Career Fellowship, the NSF CAREER award, the Spira Teaching award, the Eta Kappa Nu (CMU Chapter) Excellence in Teaching award, and the Herschel Rich Invention award from Rice University.