Line Extraction Using Image Carving

Zeng, Z. and Yan, H.

A line extraction method based on image carving is presented in this paper. The gradient magnitude of a pixel is taken as the point-importance of the pixel. The length of the line where the pixel lies, is considered as the lineimportance of the pixel. Iteratively, those pixels, whose point-importance value is less than an updated pointimportance threshold and whose line-importance value is less than an updated line-importance threshold, are carved off. In other cases, the pixels are preserved. A broken line can be identified and repaired using the gradient image of the last iteration. This method can extract 'important' lines, remove 'unimportant' ones, and repair broken ones.

Cite as: Zeng, Z. and Yan, H. (2003). Line Extraction Using Image Carving. In Proc. Pan-Sydney Area Workshop on Visual Information Processing (VIP2002), Sydney, Australia. CRPIT, 22. Jin, J. S., Eades, P., Feng, D. D. and Yan, H., Eds. ACS. 117.

(from crpit.com) (local if available)