|
Activity Number:
|
467
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 5, 2009 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
ENAR
|
| Abstract - #304668 |
|
Title:
|
Blind Source Detection in Images: A Multiple Testing Strategy
|
|
Author(s):
|
David A. Friedenberg*+
|
|
Companies:
|
Carnegie Mellon University
|
|
Address:
|
5562 hobart st. #409, Pittsburgh, PA, 15217,
|
|
Keywords:
|
multiple testing ; image analysis ; astrostatistics ; source detection
|
|
Abstract:
|
New technologies like high-resolution telescopes and remote sensing satellites produce massive images, in which a diverse collection of objects are embedded in a noisy background. Identifying the important components and estimating their properties is crucial in any analysis of these data. One statistical approach to this problem is conducting a hypothesis test at each pixel and using a multiple testing correction. However, a pixel is a unit that depends on the instrument and does not have scientific meaning. I have been investigating multiple testing procedures that control the proportion of false objects as opposed to false pixels. I demonstrate our techniques on data from the Chandra X-ray observatory satellite and show that we can simultaneously make a probabilistic guarantee about the rate of false sources and detect objects with power comparable to standard techniques.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2009 program |
