|
Activity Number:
|
150
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, July 30, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #310082 |
|
Title:
|
Using Geometrical Tools for Dimension Reduction of Images
|
|
Author(s):
|
Evgenia Rubinshtein*+ and Anuj Srivastata
|
|
Companies:
|
University of Central Arkansas and Florida State University
|
|
Address:
|
2840 Dave Ward Dr, Conway, AR, 72034,
|
|
Keywords:
|
Dimension reduction ; image analysis ; stochastic optimization ; kurtosis ; variance
|
|
Abstract:
|
We are interested in low-dimension linear representations of images, motivated the fact that image analysis often requires dimension reduction before statistical analysis, in order to apply sophisticated procedures. We present geometric tools for finding linear projections that optimize a given criterion for a given data set. We formulate this problem as multidimensional optimization on Stiefel manifold and use different criteria. The gradient vector is represented as the orthogonal projector onto the tangent space of the manifold. We use stochastic gradient methods to solve this problem in order to search for the global maximum. We demonstrate these results using several image datasets, including natural images and facial images.
|
