This tool is used to perform image-to-map rectification, i.e. transform the geometric properties of an image to match that of a projected map, or image-to- image registration. Image-to-map rectification is commonly used when an image must be overlayed with other geographic data, e.g a road network, and image-to- image registration is usually only carried out for change detection applications where two images must be exactly co-registered on a pixel-to-pixel basis. The two operations follow the same approach. A collection of ground control points (GCPs) must be digitized on the input image. These GCPs are input to the tool in the form of a shapefile of a Point ShapeType. This file can be created through on-screen digitizing into a blank shapefile created using the Create New Shapefile tool. Each control point in the image GCP file must have a corresponding GCP digitized in the reference coordinate system (i.e. either in the map coordinates or the coordinate system of a second reference image for rectification and registration respectively) and contained in the map GCP shapefile.
Once the user specifies the names of the input files and output image file, they will be presented with the rectification dialog, which displays information about the transformation error associated with each GCP and the overall root-mean- square-error (RMSE). The user can optionally deactivate individual, poorly performing GCPs to reduce the RMSE. The transformation is based on a two-variable (X, Y) polynomial transformation. The order of the polynomial can range between 1 (an affine transformation) and 5. Satellite imagery can typically be rectified using a low-order transform (e.g. 1-2), while sub-orbital imagery usually requires a higher-order transform (e.g. 2-4). Although higher-order polynomials generally reduce the overall error, it is often the case that the improved fit is localized to the areas immediately adjacent to the GCPs and undesirable warping can occur in regions not covered by GCPs. As such, it is recommended that the lowest order polynomial that allows for a reasonably good fit to the data be applied. Affine transformations can account for geometric distortions resulting from translation, scaling, rotation, and skew. The success of a rectification operation is highly dependent on the number, placement, and quality of GCPs. Furthermore, the higher the order of polynomial transformation the greater the number of GCPs that are required. The mathematically defined minimum number of GCPs for a 1, 2, 3, 4, and 5 order polynomial transformation are 3, 6, 10, 15, and 21 respectively. However, in reality many more than the mathematical minimum number of GCPs are required to rectify imagery adequately.
This tool requires a higher level of user interaction than most other tools and is not intended for a scripting environment.