Feature Selection on Affine Moment Invariants in Relation to Known Dependencies

0476980 - ÚTIA 2018 RIV CH eng C - Konferenční příspěvek (zahraniční konf.)
Zita, Aleš - Flusser, Jan - Suk, Tomáš - Kotera, Jan
Feature Selection on Affine Moment Invariants in Relation to Known Dependencies.
Computer Analysis of Images and Patterns : 17th International Conference, CAIP 2017. Cham: Springer, 2017 - (Felsberg, M.; Heyden, A.; Krüger, N.), s. 285-295. Lecture Notes in Computer Science, LNCS 10425. ISBN 978-3-319-64698-5.
[International Conference on Computer Analysis of Images and Patterns (CAIP 2017) /17./. Ystad (SE), 22.08.2017-24.08.2017]
Grant CEP: GA ČR GA15-16928S
Grant ostatní: GA UK(CZ) 1094216
Institucionální podpora: RVO:67985556
Klíčová slova: Affine invariants * Image moments * Feature selection * Machine learning * Pattern recognition
Obor OECD: Automation and control systems
http://library.utia.cas.cz/separaty/2017/ZOI/zita-0476980.pdf

Moment invariants are one of the techniques of feature extraction frequently used for pattern recognition algorithms. A moment is a projection of function into polynomial basis and an invariant is a function returning the same value for an input with and without particular class of degradation. Several techniques of moment invariant creation exist often generating over-complete set of invariants. Dependencies in these sets are commonly in a form of complicated polynomials, further-
more they can contain dependencies of higher orders. These theoretical dependencies are valid in the continuous domain but it is well known that in discrete cases are often invalidated by discretization. Therefore, it would be feasible to begin classi cation with such an over-complete
set and adaptively nd the pseudo-independent set of invariants by the means of feature selection techniques. This study focuses on testing of the infuence of theoretical invariant dependencies in discrete pattern recognition applications.
Trvalý link: http://hdl.handle.net/11104/0273427