Image Restoration Using Machine Learning

High quality images are essential in scientific discovery, forensics, and medical diagnosis. Images can be degraded by blur caused by lens or atmospheric effects, by motion of the subject, or by defective recording devices.

The talk will focus on using ideas from machine learning and scientific computing to guide us to improved estimates of the true image and of the uncertainty in our estimate. The methods include exploiting training data, using Bayesian estimation, and designing optimal filters.

Some of this work is joint with Julianne Chung, Matthias Chung, Glenn Easley, and Bert Rust.

Speaker Bio: Dianne Prost O'Leary is a NIST faculty member in the Applied and Computational Mathematics Division of ITL, a professor of computer science at the University of Maryland, and also holds an appointment in the university's Institute for Advanced Computer Studies (UMIACS) and in the Applied Mathematics and Scientific Computing Program. She earned a B.S. from Purdue University and a Ph.D from Stanford University. Her research is in computational linear algebra and optimization, with applications including solution of ill-posed problems, image deblurring, information retrieval, and quantum computing. She has authored over 100 research publications on numerical analysis and computational science, two textbooks, and 30 publications on education and mentoring. She is a member of AWM and a Fellow of SIAM and ACM. She was awarded a Doctor of Mathematics degree, honoris causa, from the University of Waterloo in 2005 and was chosen to be the 2008 AWM-SIAM Sonia Kovalevsky Lecturer. Further information about her work can be found at http://www.cs.umd.edu/users/oleary.

Presentation Slides: PDF

Note: Visitors from outside NIST must contact Cathy Graham; (301) 975-3668; at least 24 hours in advance.