A Sub Peak Tracker Based Hough Transform for Accurate and Robust Linear Edge Extraction

Hough transform is a robust and efficient method for line segment extractions from complicate scenes. In most existing Hough transform methods, after peaks are detected in Hough space, one individual line is defined corresponding to each peak. However, in this way, the result is vulnerable to tiny orientation change of the linear edge in real image. The orientation change of linear edge leads the pixels of the edge to be contributed to several adjacent cells in Hough space, which makes it difficult to extract the edge by one peak. In this paper, a Sub Peak Tracker (SPT) is proposed to find sub peaks around the detected peaks by analyzing the butterfly pattern distribution in Hough space. Based on the SPT, an improved Hough transform algorithm is developed for robust and accurate linear edge extraction. The robustness and high accuracy of the algorithm in extracting linear edge are verified by experimental testing.