| | |
Summary: Adjustable Invariant Features by Partial Haar-Integration
Bernard Haasdonk, Alaa Halawani and Hans Burkhardt
Computer Science Department
Albert-Ludwigs-University Freiburg
79110 Freiburg, Germany
haasdonk,halawani,burkhardt¡ @informatik.uni-freiburg.de
Abstract
A very common type of a-priori knowledge in pattern
analysis problems is invariance of the input data with re-
spect to transformation groups, e.g. geometric transforma-
tions of image data like shifting, scaling etc. For enabling
most general analysis techniques, this knowledge should be
incorporated in the feature-extraction stage. In the present
work a method for this, called Haar-integration, is gener-
alized to make it applicable to more general transforma-
tion sets, namely subsets of transformation groups. The re-
sulting features are no longer precisely invariant, but their
variability can be adjusted and quantified. Experimental re-
sults demonstrate the increased separability by these fea-
tures and considerably improved recognition performance
|
