Estimating the phase congruency of localised frequencies

Abstract

Phase congruency is a new method for detecting features in images. One of its significant strengths is its invariance to lighting variation within an image, as well as being able to detect a wide range of interesting features. We present a method for estimating the phase congruency of localised frequencies that cannot be measured separately by Gabor filters. We show that by measuring the ratio of the standard deviation to the mean energy between different phase shifted Gabor filters we are able to estimate whether the localised frequencies are phase congruent. We then show example results from applying this estimation procedure to a set of images. We also show improvements when compared to another phase congruency detector. We conclude that the concept of estimating the phase congruency of localised features is possible, but more work is needed to mature the technique to a robust feature detector.