Stable Feature Classification in the Wavelet Domain
Abstract
This research project studied the mathematics of adapted wavelet transforms, which underlie some highly successfull algorithms for feature detection and classification. The investigators surveyed five feature detection methods, evaluating them empirically on reasonably large data sets to find their weakest points. In addition, new methods have been found for characterizing and constructing wavelets. Newly discovered equations characterizing orthonormal wavelet bases can be applied, along with more classical constructions, to create new families of wavelets and similar functions. These provide connectivity and allow new dilation structures and translations by new lattices. Research supported by this grant in its final year has resulted in ten referred publications that appeared in 1999 or were accepted for publication.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 02, 2000
- Accession Number
- ADA379900
Entities
People
- Guido L. Weiss
- M. V. Wickerhauser
Organizations
- University of Washington