Reducing the Optimisation Effort in Batch Mosaicing

Abstract

Image mosaicing methods have become very popular thanks to the rapid progress in obtaining optical data using robotic platforms. Global alignment which is known as finding best image registration parameters taking into account overlapping image pairs is one of the essential steps of image mosaicing process in order to obtain seamless mosaics. Global alignment usually employs non-linear optimisation methods to minimise the error metric defined over correspondences detected between overlapping image pairs. In this paper, we propose a method that uses graph theory principles to reduce the total number of overlapping image pairs used in the global alignment process without disturbing the final mosaic quality. This reduction allows for obtaining image mosaics with reduced computational cost and time. The method is validated through two experiments which involve challenging underwater datasets.