Robust vectorization using graph-based thinning and reliability-based line approximation

A vectorization method for line patterns which converts digital binary images into line-segment vectors is proposed. The proposed method suppresses shape distortion as well as pseudo-feature points, so it produces more compact and natural-shaped vector data than existing methods. The experimental results of applying the proposed method to geographical maps show that the vectorization method reduces data volume by 74%-87% and suppresses shape distortion more than is possible with a straightforward method