Image matching problem for certain cases of perspective photography
Yu. B. Blokhinov, D. A. Gribov, Alexey Chernyavskiy · Journal of Computer and Systems Sciences International · 2008
The modern state of matching problem for digital images is addressed. Image matching presumes the process of overlapping of two or more images shot at different time instants from different perspective views or, maybe, from different sensors. In addition the most precise geometric correspondence should be reached. The main stages of the matching problem are considered in detail. The advantages and disadvantages of the solutions proposed in this field by different researchers are analyzed. The solutions developed earlier that are the most successful from the authors’ opinion are employed, and original solutions of certain problems are also proposed. For reliable control and rejection of false points, the authors have developed original algorithms of metric filtering and iterative stochastic filtering SwarmSAC. The objective of this paper is to develop and present a methodology of identification and matching of fragments of digital images received for a wide range of shooting conditions.