Efficient image matching algorithms based on procedures of searching for 2D templates

Abstract

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. © Springer-Verlag 2003.