A new approach for line recognition in large-size images using Hough transform

Applications of the Hough Transform (HT) have been limited to small-size images for a long time. For large-size images, peak detection and line verification become much more time-consuming. Many HT-based line detection methods are not able to detect line width. This paper proposes a new approach for detecting line segments using HT, with applicability to large size images, especially for those situations where line width is critical. Our approach applies a boundary recorder to eliminate redundant analyses, and employs an image-analysis-based line-verification method to overcome the difficulty of using a threshold to distinguish short lines from noise. It avoids overlapping lines by removing the pixels of detected line segments, a method which is more robust than only clearing the N×N neighborhood. This approach could be easily extended to improved HT methods that perform global accumulation. Experimental results show that this approach is very time efficient for large-size images.