Abstract
Recently, many space-borne imaging sensors have been available. As satellites have various height and various spectral bands, we can use many types of data, for example, to analyze the change of the earth's environment. These image data have their own geometrical property, and therefore, they have relative disparity among them. The disparity makes it difficult for us to apply multitemporal and/or multisensor methods for efficient analysis of the environment. In change detection processing, for example, the disparity between multitemporal images causes apparent change. The disparity between stereo pair images also gives us topological information such as terrain height. It has been indispensable to register remotely sensed images with high accuracy. There have been a lot of methods reported for the image registration.
In the most of the former methods, we needed to choose a coordinate transfer function and to select manually many pairs of control point. Accuracy of the registration wholly depends on the coordinate transfer function and the accuracy with which control point pairs were selected. Manual selection of the point pairs is very time consuming and it often causes decrease in accuracy due to mis-selection. Efficient and accurate algorithm has been acquired.
In this paper, we propose an automated registration method ARTSIM (Automated Registration meThod for Satellite IMages). ARTSIM automatically generates as many pairs of control point as possible on the images to be registered. A spatial correlation technique is used to remove the mis-selection of control point pair. In order to avoid the error due to using inadequate coordinate transfer function, we adopt a piecewise registration where images are divided into many triangles so that their vertices are the control points generated, then piecewise Affine transformation is applied to each triangle pair. For achieving high level automation, ARTSIM compensates relative rotation and shift between the images from rough information about their initial disparity.
Here, we describe the principle and the procedures of ARTSIM. The validity of the method is confirmed by numerical simulation. Application results of ARTSIM to actual remotely sensed images are also shown.
