SCRIPPS INSTITUTION OF OCEANOGRAPHY SPECIAL SEMINAR -
DATE: June 6th, Thursday, 12:45 p.m.
LOCATION: Eckart 227
SPEAKER: Stanley Chan
TITLE: Rethinking Atmospheric Turbulence
Atmospheric turbulence is one of the most devastating distortions in long-range imaging systems for naval missions, surveillance, navigation, weather monitoring, and astronomical exploration. Image processing methods for mitigating atmospheric turbulence has been studied for at least two decades, but their performance remains unsatisfactory: Many algorithms are agnostic to the actual turbulence model; Some are overly complicated; Most can only handle a narrow set of imaging conditions.
The purpose of this talk is to rethink the physical forward model of atmospheric turbulence, and to propose new algorithms to recover the images. Specifically, I will discuss (1) The surprising similarity between a turbulence model and the large deviation theory used in statistical learning; (2) How the new model based on statistical learning can unify the mainstream image processing algorithms; (3) How to design a simple and effective algorithm for image recovery. Towards the end of the talk I will discuss developing synergy between ECE and SIO to tackle the open challenge of sensing in harsh environment. I will also share my experience in creating an award-winning machine learning course at Purdue College of Engineering, enrolling 300+ MS/PhD students across all engineering departments.
Faculty Host: Bill Hodgkiss (email@example.com)