Why is it sensible to use wavelets in matched filtering?
Heiko Schwarzer, Stephan Teiwes, Frank Wyrowski · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 1996
The application of wavelets in optical pattern recognition has been first proposed by Sheng et al. The concept of a wavelet matched filter has been motivated by its improved discrimination properties compared to the classical matched filter and its improved signal-to-noise ratio compared to the phase-only filter. However, similar features can also be achieved using the approach of fractional power filters. Thus, the question for striking advantages resulting from the composition of wavelet filters and matched filters is still open. In this paper we give a positive answer to this question by discussing the problem of detecting defects on textured surface materials. This problem is of high practical relevance in industrial manufacturing.