Registration of region of interest for object tracking applications in wide area motion imagery

Abstract

Image registration or stabilization is a task that has been focused on in many fields in image processing. Many methods are available for registering two images, but when dealing with wide area motion imagery, registration of the full scene can be taxing computationally. The proposed algorithm is an application utilizing two existing registration methods to stabilize a specific region of interest in an aerial image database. The registration tools implemented during this application are a phase correlation method and the efficient second-order minimization method. The combination of both registration functions act as a coarse-to-fine image registration algorithm. The goal of this application is to output a set of registered smaller sized images from the larger dataset for the use in detection and tracking of objects in wide area imagery. Experiments performed on several image streams show that the proposed technique is effective and gives high registration accuracy.