An improved RHT algorithm to detect line segments

Abstract

Line segment detection is widely used in image process and recognition field. Conventional Hough Transform (CHT) is one of the popular techniques to detect lines accurately. Randomized Hough Transform (RHT) decreases the huge computation of CHT, but the random samples produce large invalid samples, so it increases the time and space complexities. Another disadvantage of RHT is that some important line properties, such as line length, the start and end positions of the detected line, can not be gotten directly. This paper proposes an improved RHT algorithm to decrease the invalid samples by finding the relation between the improved chain code direction and line points. The neighbour point distances are computed to segment lines so as to retrieve the line length and positions. Experiment showed this improved algorithm can detect and locate the line segments faster and more precise than RHT in the Cartesian coordinate system, especially for the noisy-image and complicated image with lots of connected regions.