Automatic Pixel Selection Criteria for Image Registration

Image registration is the process of matching two or more images. It involves choosing one image as reference. The remaining images are modified to match the one chosen as reference. Two images match between them whenever every region of the original scene is mapped on the same set of pixels in both images. The registration problem consists in finding the geometric transformation to accomplish the required modification of the image. Over the years, a broad range of techniques have been developed for the various types of images and transformations. Existing automatic image registration algorithms use all the pixels of an image. This does not take into account the fact that the information content of different pixels may be rather different. Information content is measured in relation with its contribution to define the geometric transformation. This paper introduces a method that automatically identifies pixels with high content of information. Pixels with low content of information are discarded.