Friday,
February 8, 3:15 p.m. PSA 206
Aaron Luttman
Div Sci Math, Bethany Lutheran College
Inverse Problems for Botanical and Astronomical Image Analysis
Abstract
Many image analysis problems are formulated as inverse problems, where the goal is to
minimize a particular energy functional over some set of allowable functions. Two such
problems are image segmentation and image deblurring. In botany it is useful to capture
image data of leaves as they fluoresce in the infra-red, and the goal is to segment the videos
or images into regions of fluorescence and non-fluorescence. The botanical problem will be
described as well a variational technique with numerical methods for video segmentation in
this context. Astronomical data measured from the ground is blurred as it passes through
the atmosphere, and this effect must be reversed in order to analyze the data. This
deblurring is formulated as an inverse problem, and we present theoretical analysis and
numerical results demonstrating that Poisson negative log-likelihood estimation can be used
to reconstruct such astronomical data when regularized using the total variation of the
reconstruction.
For further information please contact:
mittelmann@asu.edu
