A Method of Region Considered Segmentation for Automatic Image Registration

Automatic image registration is still an actual challenge in several fields. Even though many image registration algorithms have been developed over the years, it is far from the broad use of applications. This paper deals with problem faced in the area of remote sensing consisting of image registration. Generally remote sensing images may have different gray –level characteristics and multiple features, so simple techniques cannot be applied directly. In this work, new automatic image registration through region considered segmentation is proposed. This approach performs region considered segmentation for the pair of images to be registered, followed by features characterization of the extracted objects with control point selection and objects matching. It leads to the registration of satellite images (multisensor and multiband) with differences in rotation and translation. The region considered segmentation for automatic image registration leads to a more sub pixel accuracy. Index Terms – Region considered segmentation; image matching; remote sensing; control point selection. 1