Application of a reciprocal confluence tree unit to similar-picture retrieval

Abstract

S. K. Chang et al. [1] defined three types of pattern matching to retrieve “similar” symbolic pictures from a symbolic picture database. Symbolic pictures are composed of icons. Each pair of icons have 9 possible spatial relationships, depending on whether their x an y coordinates are <, =, or >. In this paper we propose the use of a reciprocal confluence tree unit [2] to compute and represent each symbolic picture as a large integer. An algorithm is then given for determining whether or not a pattern picture matches a subject picture. This algorithm runs in O(n² log(m) + m²) time, where n is the number of icons in the pattern and m is the number of icons in the subject.