Precision single-particle localization using radial variance transform

We introduce an image transform designed to highlight features with high degree of radial symmetry for identification and subpixel localization of particles in microscopy images. The transform is based on analyzing pixel value variations in radial and angular directions. We compare the subpixel localization performance of this algorithm to other common methods based on radial or mirror symmetry (such as fast radial symmetry transform, orientation alignment transform, XCorr, and quadrant interpolation), using both synthetic and experimentally obtained data. We find that in all cases it achieves the same or lower localization error, frequently reaching the theoretical limit.