Image superimposition and the problem of selecting the set of corresponding point pairs

The results of experimental research of both homogeneous and heterogeneous image superimposition quality by means of multiple view geometry. One of the most important tasks in the problem of image superimposition which has not been solved till now is the task to make the right selection of a set of key (corresponding) point pairs on superimposed images. The research shows image superimposition by two methods. First, exhaustive search of all possible combinations into groups of 4 points results in a group (groups) of key points providing the best image superimposition. Second, suitable subset of key points is found using the algorithms of RANSAC group. The comparison of results has been made, recommendations to automate the process of key point selection have been given.