Wyniki wyszukiwania - BazTech

Ograniczanie wyników

1 Image Processing & Communications

1 2001

Powiadomienia systemowe

Sesja wygasła!

Znaleziono wyników: 1

Liczba wyników na stronie

Wyniki wyszukiwania

Sortuj według:

Ogranicz wyniki do:

Quazi-optimal Zernike feature selection

Labunets V., Maidan A., Astola J., Egiazarian K.

Image Processing & Communications

2001

Vol. 7, no 1-2

31-40

Moment invariants have found wide applications in image recognition since they were proposed. The recognition problem is often connected with image reconstruction technique to determine a desired set of invariants for use their in a recognition system. For image reconstruction low order moments are important because they contain information about general shape of the image. But these moments are not efficient for recognition because general shapes of different objects can be very similar and do not allow to distinguish one object from another. The main difficulty in the application of the moment invariants is the absence of theoretical methods for estimating their efficiency in recognition tasks. In this paper, our goal is to analyse the significance of Zernike moments of different orders from the viewpoint of pattern recognition theory. We propose simple intuitive method of optimal filtering in invariant domain, namely, to select image features in order to minimize errors (misclassification rates) excluding noise sensitive and unstable features.