Combination of Local and Global Line Extraction

✍ Scribed by Ville Kyrki; Heikki Kälviäinen

• Publisher: Elsevier Science
• Year: 2000
• Tongue: English
• Weight: 759 KB
• Volume: 6
• Category: Article
• ISSN: 1077-2014
• DOI: 10.1006/rtim.1999.0183

✦ Synopsis

I

n this paper we study how to combine local and global line extraction. The Hough transform is usually used to detect line segments in an image. However, the standard Hough transform (SHT) suers from time and storage complexity, and it is incapable to utilize local line extraction. Recently an approach, called the connective randomized Hough transform (CRHT), has been proposed to take advantage of local detection, such as the connectivity of neighboring edge points and the line ®tting of these detected points. However, this approach contains problems with images with many distorted lines. We suggest a new line detection approach, called the extended connective randomized Hough transform (ECRHT), to alleviate these problems. The use of gradient information of edge points is also discussed. Experiments with simulated and real-world data demonstrate the bene®ts of the ECRHT, as compared to the SHT and the CRHT.