|
Activity Number:
|
190
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 4, 2008 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #301223 |
|
Title:
|
Bayesian Scale Space Analysis of Image Differences
|
|
Author(s):
|
Leena A. Pasanen*+ and Lasse Holmström
|
|
Companies:
|
University of Oulu and University of Oulu
|
|
Address:
|
Dept. Of Mathematical Sciences, P.O. Box 3000, Oulu, International, FIN-90014, Finland
|
|
Keywords:
|
Image analysis ; Bayesian methods ; Scale space ; SiZer ; BSiZer ; Applications
|
|
Abstract:
|
We consider the detection of features in noisy images that appear in different spatial scales, or resolutions. In particular, our goal is to capture scale dependent differences in a pair of images of the same scene taken at two different instances of time. New approaches are proposed that use Bayesian statistical modeling and simulation based inference. The methods can be viewed as further development of SiZer technology, originally designed for nonparametric curve fitting. A strength of the Bayesian approach is straightforward inference and incorporation of domain specific prior information on the images in question. While we demonstrate the performance of the new methods mostly using artificial test images, the approach is believed to have applications e.g. in satellite based remote sensing. Hence, we also include a preliminary analysis of a pair of Landsat satellite images.
|
