Contourlet image preprocessing for enhanced control point selection in airborne image registration

Abstract

In applications such as airborne imagery, target tracking, remote sensing, and medical imaging; it is helpful to have an image set where all of the images lie on one fixed coordinate system. However, frequently a set of images cannot be captured from a fixed perspective using the same sensor or different sensors at the same time. Image registration presents a solution by mapping points from one image to corresponding points in another image; however existing registration methods are computationally expensive and not completely accurate. Hence, continual investigation of image registration methods is needed such as those using feature-based or intensity-based approaches, transformation models, spatial and frequency domain methods, and single or multi-modality data. In this paper, we investigate these processes by focusing on the identification of control points, which play a vital role in the process of registering images. By using the multi-resolution contourlet transform for image preprocessing, control points are better identified, which provides us a more reliable image registration for applications such as image fusion.