p2pHT: A Point to Point Hough Transform

Initial results on a fast approach for detecting straight line segments based on the combination of local search and Hough Transform (HT) is discussed in this paper. Though HT is a robust method for extracting patterns from noisy images, it does not regard the level of occlusions and the minimum allowable line segment length. We propose a Sliding Window (SW) concept which eliminates most of the spurious data in the Hough Space (HS) and takes into consideration the mentioned parameters. Moreover, HS is only estimated at points where peaks are expected, thus reducing the time complexity. Experimental results show clear improvement over the conventional HT.