Graph-based vectorization method for line patterns

Abstract

A vectorization method for line patterns is proposed which converts digital binary images into line segment vectors. The vector data is more compact and more natural than that obtained by conventional methods using thinning operations. The proposed method consisted of four steps. First, thinning of an input binary image is performed. Then a medical line image obtained by the thinning operation is transformed into a graph, in which pixels on the medial line correspond to nodes and neighboring nodes are connected by edges. Next, extra edges unnecessary for preserving the topology of the medial line image are deleted. The deletion can be implemented as an iterative parallel operation. Finally, the graph is simplified by line approximation. Every step except the line approximation is suitable for parallel processing. The experimental results of applying the proposed method to geographical maps show that the method reduces data volume by about 30%, as compared with conventional methods. >