Automatic image registration of multi-temporal KOMPSAT-2 images in agricultural areas

The objective of this study is to extract welldistributed matching points in agricultural areas having many similar patterns of farming intensity and homogeneous paddies. Mutual information (MI) which is one of the representative intensity-based similarity measures was used to estimate the initial translation difference between multi-temporal highresolution images. Regular grids were then constructed over the entire images based on the initial translation and the center points of each grid were evaluated as the candidate points for precise matching. Based on these points, corresponding points were determined as the final matching points using the MI method. Multi-temporal high-resolution satellite images from Korea Multi-purpose Satellite-2 (KOMPSAT-2), which was launched in 2006 and has 1 m spatial resolution, were used for the experiments. Using the proposed method, we were able to extract evenly distributed matching points over the entire image and we obtained acceptable registration accuracies for all sites.