Tuneable edge enhancement filters for optical correlation

Abstract

Difference of Gaussian (DOG) filters, a form of wavelet filter, are an extremely useful preprocessing tool for the enhancement of image edge data. The DOG filter is highly tuneable by control of the standard deviations of its constituent Gaussian distributions. These can be used in the frequency plane of optical correlators and implemented in the form of a static non-updateable filter. This article reports on a simple optical technique using photorefractive materials whereby a filter, updateable in real time and very similar to the DOG filter, is implemented by tuning the intensity of the hologram reference wave to give enhanced modulation in a selectable frequency band. This is called a tuneable photorefractive (TPR) filter. The results of the DOG and TPR filters are compared with those of the phase-only filter and classical matched spatial filter with respect to the criteria of signal-to-noise ratio, Horner efficiency and discrimination capability.