A robust algorithm of principal curve detection

In this paper, we present a new method for the detection of principal curves in complicated feature images. Based on the criteria of the shortest path of curves and directional deviation of paths, principal curve detection is carried out in graph domain. DFS searching scheme is adopted in exploration of a graph network. The motivation of this research is to find road boundaries and house contours from printed map images. By extensive experiments, the algorithm has shown good efficiency and robustness with real map images. The technique described in this paper can also be used in other applications, such as in character recognition, to separate characters from other unwanted document components that overlap with the characters.