Robust Image Registration via Cepstral Analysis

Abstract

Cepstrum based image registration has been largely overlooked over the past two decades. Initial investigation of the method was limited to alignment of images exhibiting only translational differences. Later attempts to extend the method to handle rotations proved disappointing. This paper presents a cepstrum-based registration method that provides performance comparable to the state of the art over a wide range of scale factors and rotation angles, with reduced computational load.