Wavelet-based identification and classification of local symmetries in microscopy images

We present a method for the identification and classification of local symmetries in biological images. We aim at obtaining a precise estimate of symmetric junctions in a scale and rotation invariant way. The proposed method is template-free, which allows the test of any combination of arbitrary symmetry orders in an effective way. Our measure of local symmetry is derived from a circular harmonic wavelet analysis. The basis functions exhibit different symmetry orders. We use this measure to formulate a classifier to label the different junctions into one of several symmetry classes. We present experimental results, and validate our method using both on synthetic images and biological micrographs.