A new method for extraction of approximately vertical lines from natural images
Jian Cun Zhao · 哈尔滨工业大学学报：英文版 · 2002
For natural images with complex background and noise, this new approach appears is more effective than other techniques for three key reasons: 1)edge elements treated interdependently to overcome the weakness of standard Hough transform (HT); 2)gradient orientation information taken into account in the process of HT; 3) an effective method used to merge the HT result, i.e. many clustered short lines into single straight line to represent the edge of object to be recognized. This method consists of three steps are: 1) edge detection; 2)modified Hough transform to extract clustered short lines; 3) merging clustered short lines into a single line after few modifications. The method presented in this paper could also be used for extraction of non vertical straight lines.