Graphs representing quadtree structures using eight edges

Quadtree is an important data structure for image processing. In this paper, a graph based data structure is proposed for representing two-dimensional binary images. This data structure uses eight edges for each node to represent neighborhood relations in quadtrees. New neighbor finding and connected component labeling algorithms using the structure are also introduced. Empirical tests are given for the algorithms in comparison with usual neighbor finding and connected component labeling algorithm based on quadtrees [2] - [3]. In our experiments using natural images, the proposed neighbor finding algorithm is more than twice as fast as algorithm using quadtrees and the proposed connected component labeling algorithm is 40% faster than quadtree based algorithm.