Rotationally invariant correlation filtering
- 1 September 1985
- journal article
- Published by Optica Publishing Group in Journal of the Optical Society of America A
- Vol. 2 (9) , 1411-1418
- https://doi.org/10.1364/josaa.2.001411
Abstract
A method is presented for analyzing and designing optical correlation filters that have tailored rotational invariance properties. The concept of a correlation of an image with a rotation of itself is introduced. A unified theory of rotation-invariant filtering is then formulated. The unified approach describes matched filters (with no rotation invariance) and circular-harmonic filters (with full rotation invariance) as special cases. The continuum of intermediate cases is described in terms of a cyclic convolution operation over angle. The angular filtering approach allows an exact choice for the continuous trade-off between loss of the correlation energy (or specificity regarding the image) and the amount of rotational invariance desired.Keywords
This publication has 12 references indexed in Scilit:
- Rotation-Invariant Pattern RecognitionOptical Engineering, 1984
- Unified synthetic discriminant function computational formulationApplied Optics, 1984
- Rotation-invariant digital pattern recognition using circular harmonic expansionApplied Optics, 1982
- Optical pattern recognition using circular harmonic expansionApplied Optics, 1982
- Linear combinations of fillers for character recognition: a unified treatmentApplied Optics, 1980
- Image analysis via the general theory of moments*Journal of the Optical Society of America, 1980
- Multivariant technique for multiclass pattern recognitionApplied Optics, 1980
- Optical character recognition based on nonredundant correlation measurementsApplied Optics, 1979
- A Threshold Logic Network for Shape InvarianceIEEE Transactions on Electronic Computers, 1967
- Signal detection by complex spatial filteringIEEE Transactions on Information Theory, 1964