Project_11
Blind Deconvolution
Errors, Errors Everywhere


Project 11 considers the problem of dealing with uncertainty in a mathematical model as well as in the data; this case involves spectroscopic data.

In the spectroscopic case, we are taking measurements of a spectrum, but on top of the errors in measurement that we expect, we know that there is also some kind of blurring going on. When we make a model of the process by which the exact physical system is measured, we want to account for the measurement errors, for which we can make a reasonable model, but also for the blurring, about which we initially know nothing. The fact that the blurring function is unknown is why this process is sometimes called blind deconvolution.

The computational tools to be used to try to handle this problem include both the standard least squares method, and a related method called total least squares.

Reference:

  1. Dianne O'Leary,
    Blind Deconvolution: Errors, Errors Everywhere,
    Computing in Science and Engineering,
    Volume 7, Number 1, January/February 2005.
  2. Dianne O'Leary,
    Scientific Computing with Case Studies,
    SIAM, 2008,
    ISBN13: 978-0-898716-66-5,
    LC: QA401.O44.


You can go up one level to the Computational Science Projects page.


Last revised on 10 February 2009.