Improvement of methodology for homography superposition of images

The article presents an improved technology for projective image alignment. The original technique includes two stages: preliminary comparison of singular points and selection of points that satisfy geometric constraints. The improvement is achieved by introduction of another stage, with selectionof those points that correspond to unique image fragments among the singular points. Points corresponding to similar fragments are removed from the image, since they determine a large number of potentially incorrect hypotheses, which verification increases total processing time. Such improvement allows to reduce alignment time, which is one of the general characteristics of the developed technology. The improved technology was tested on images taken by an unmanned aerial vehicle. The experiment showed a significant decrease of the image processing time.