SWIFT Image Registration with ORB Features

The technique of aligning and overlaying two or more images with an overlapping area to create a single integrated image is known as image registration. It is one of the essential parts of many applications like change detection, mosaicing and fusion of images. A general image registration process includes feature detection, feature matching, transformation model estimation, image re-sampling and result evaluation. Although feature detection determines the overall time complexity of the image registration process, the overall robustness and accuracy are determined by the quality of the transformation model. In literature, it is generally seen that feature detection algorithm has a trade-off between time complexity and robustness. This paper aims to enhance the robustness of the image registration process using a lower time complexity algorithm like the ORB feature detection algorithm.