Masters Project, Boise State University, May 2005
By Darin Dutson
Many image processing applications require locating one image within another image. Applications which may require this include object tracking, object identification, and image registration. For purposes of astronomy, matching two images can help identify an instrument’s position, or enable automatic comparison of an image with a star catalog for identification of a previously uncataloged object.
A method is presented in this paper for matching an acquired telescope image with a star catalog. This amounts to automatically aligning the images, so that it may be determined if a telescope is positioned correctly with respect to a star catalog. This is accomplished by extracting centroids of the brightest stars from a star catalog in the region of interest. This is also done for the acquired image by first thresholding it, finding its connected components, and calculating the centroids of the connected components. The acquired image points are then translated relative to the star catalog points, and the directed, censored Hausdorff distance is calculated. The translation step that results in the smallest distance is considered to be the matching location.
Experiments indicate that the proposed method will execute in fractions of a second for a field of view over a range of several thousand pixels in both the x and y directions. Different methods of translation are suggested, depending on the range of translation, that are shown to decrease execution time. Additional improvements are possible with algorithm and code optimizations.