Efficient Image Matching Algorithms Based on Procedures of Searching for 2D Templates
A model of algorithms based on estimates calculation with 2D support sets (2D-AEC) defines recognition algorithms which could be directly applied to images and their fragments. When one uses these algorithms in practice, a problem arises of decreasing their computational complexity. Some constraints on 2D-AEC algorithms allow one to reduce this problem to the problem of efficient search a binary matrix for a certain template. Efficient two-step procedures of searching for a rectangle were suggested earlier. In this paper conditions are determined under which these procedures have the least complexity among all two-step procedures for rectangle search. This result determines image recognition problems for which the performance of 2D-AEC algorithms based on suggested search procedures cannot be improved by applying other two-step search procedures.
- 1.Gurevich, I.B.: Determining a Class of Algorithms of Estimates Calculation Based on Two-Dimensional Information for Image Recognition. In: Methods and Means of Processing Graphic Information, Intercollegiate Collection of Papers, Yu.G. Vasin, (ed.), Gorky: Gorky State Univ. (1986) 47–66 (in russian)Google Scholar
- 2.Gurevich, I.B., Nefyodov, A.V.: Algorithms for Estimates Calculations Designed for the Case of 2D Support Sets. Part 1: Rectangular Support Sets. Pattern Recognition and Image Analysis, 4, Vol. 11 (2001) 662–689Google Scholar
- 3.Zhuravlev, Yu.I., Nikiforov, V.V.: Recognition Algorithms Based on Estimates Calculation. Kibernetika, 3 (1971) 1–11 (in russian)Google Scholar