Probabilistic Hough transform for line detection utilizing surround suppression
Si-Yu Guo, Yaguang Kong, Qiu Hua Tang, Fan Zhang · 2008
A new probabilistic Hough transform algorithm for line detection was proposed. Instead of treating edge pixels in a binary edge image equally, a weight is bestowed to each edge pixel according to the surround suppression strength at the pixel, which can be used in either sampling stage or voting stage or both of the probabilistic Hough transform. This weight is used to put emphasis on those edge points located on clear boundaries between different objects, leading to higher probability of sampling from perceptually reasonable real lines in the edge image, as well as suppressed false peaks in Hough space formed by large amount of noise edges. Experiments on a real-world image base show that the new method gives higher line detection rate and accuracy, at the expense of moderate execution time acceptable for a broad range of applications, where the novel algorithm is preferable than other Hough transform methods tested.