Principal direction-based Hough transform for line detection

A robust and fast line detection method based on Hough transform (HT) is proposed in this paper. Edge pixels are extracted based on the summation and ratio of principal curvatures. Probabilistic sampling on the edge pixels is applied to reduce the count of voting. Then a one-to-one voting strategy is applied by taking advantages of the information of principal direction. The principal direction is also conducive for the successive accurate line segment extraction. The experiments demonstrate that the proposed method shows better locating accuracy and computation efficiency compared with several significant variations of HT.