A new algorithm for line detection based on the Randomized Hough Transform
Chen Linpeng, Zhang Guoliang, Jin Guangming, Tian Qi · 2007
The Hough transform is an elegant way of extracting global features like line segments from binary edge images. However, long computation time and large memory requirements prevents it from being used for practical computer vision tasks. In this paper, we introduce a new randomized Hough transform to improve line detection accuracy and robustness, as well as computational efficiency. The method is based on the fact that choosing all feature points as seed points, and random picking a part of feature points for pairing with seed points, and corresponding accumulator cells are incremented in the space. The experimentation have proved that the new algorithm is more effective and robust than others.