Title :
Fast spectral algorithms of invariants calculation
Author :
Labunets, Ekaterina ; Labunets, Valeri ; Egiazarian, Karen ; Astola, Jaakko
Author_Institution :
Signal Process. Lab., Tampere Univ. of Technol., Finland
Abstract :
The recognition of objects independent of their position, size and orientation is an important problem in pattern recognition. In this paper we propose a new fast algorithm of moment invariant computation, which needs almost no multiplications. We use modular arithmetic of the finite Galois field GF(Q) to map the geometrical moments calculation to a fast Fourier-Mellin-Galois transform, which reduces the computational complexity of moments from O(N4) to O(N2log2 N). We introduce orthogonal Fourier-Mellin-Galois moments based on a complete set of orthogonal characters of the multiplicative group of the GF(Q). These moments are modular remainders modulo Q of the classical geometrical moments
Keywords :
Galois fields; computational complexity; computational geometry; digital arithmetic; fast Fourier transforms; group theory; object recognition; spectral analysis; computational complexity; fast Fourier-Mellin-Galois transform; finite Galois field; geometrical moments; modular arithmetic; modular remainders; moment invariant computation; multiplicative group; object recognition; orthogonal Fourier-Mellin-Galois moments; pattern recognition; spectral algorithms; Arithmetic; Computational complexity; Fast Fourier transforms; Feature extraction; Laboratories; Layout; Object recognition; Pattern recognition; Polynomials; Signal processing algorithms;
Conference_Titel :
Image Analysis and Processing, 1999. Proceedings. International Conference on
Conference_Location :
Venice
Print_ISBN :
0-7695-0040-4
DOI :
10.1109/ICIAP.1999.797595