Rice University logoGeorge R. Brown School of Engineering
 
Electrical and Computer Engineering
 
 

 

Rice, Northwestern Team Wins Honorable Mention at ICCP 2017


Congratulations to ECE faculty members Ashok Veeraraghavan, Richard Baraniuk and their labs on winning “Best Paper Honorable Mention” at the International Conference on Computational Photography (ICCP)!

Their paper “Coherent Inverse Scattering via Transmission Matrices: Efficient Phase Retrieval Algorithms and a Public Dataset,” was authored by ECE graduate students Chris Metzler, Sudarshan Nagesh, Rice ECE Research Scientist Manoj Sharma, ECE faculty Richard Baraniuk and Ashok Veeraraghavan, and their collaborator at Northwestern, Oliver Cossairt.

2017.05.Scattering.Setup

 

The research deals with the challenge of imaging through ‘scattering media’ such as fog, smog or imaging thick biological tissues. In all these cases, light scatters due to interactions with the material. Consequently, the captured images lose both spatial resolution and contrast.

“We have all experienced this phenomenon as drivers when the visibility reduces dramatically due to fog,” Veeraraghavan explained.

One of the most promising techniques to overcome this effect is to characterize the effect of scattering using a complex transfer function, referred to as the transmission matrix. Unfortunately, existing methods to estimate transmission matrices are computationally intractable, prohibiting any practical use. The group tackled this problem by developing a new phase retrieval algorithms and techniques for estimating transmission matrices that are far more efficient than existing methods. By applying this algorithm and calibrating the system in a new way, the group achieved a 100x reduction in computational times, moving these techniques closer to practical realism.

In addition to the progress their research has made, the team also released the first publicly available transmission matrix dataset.

“Imaging through scattering media is a challenging and significant problem,” Metzler explained. “It is our hope that our new and publicly available dataset will help further our understanding of the geometric and statistical redundancies exhibited by these transmission matrices, eventually allowing for fast and efficient (and perhaps one day real-time) estimation of transmission matrices.”

ICCP fosters a community of computational photography researchers, found over a span of fields and disciplines. The 2017 conference was held May 12-14 at Stanford University.

-Jennifer Hunter