Vol.40 No.12

Journal of Xi'an Jiaotong University

Jan.2006

retue.gif (1614 ×Ö½Ú)

zwb.gif (1647 ×Ö½Ú)

¡¡

Fast Parallel Algorithm for Detecting Lines from Image
Yu Fan£¬Wu Huisi£¬Qin Zheng£¬Yang Bo
(School of Electronics and Information Engineering£¬Xi'an¡¡Jiaotong¡¡University£¬Xi'an 710049£¬China)

Abstract£ºA quick lines detection algorithm under the general super parallel system environment with scalable cluster structure is presented£® It divides the original image into Nª«¡Á Nª« parts with same area by means of the partition method of optimal domain£¬ and maps each sub image onto the local coordinates of the nodes of the parallel system for detecting lines£® Two thresholds are preintroduced to the two different scanning periods of Hough transform to simplify calculation£® It is proved that the detection with 1/(2N£­1) of the original threshold on the divided sub image can satisfy the demands that are no target loss in the original image£® Experiment result indicates that the algorithm effectively reduces the computation complexity and raises the accelerating ratio up to 25£®320£¬ while being highly robust and noise avoidable£®
Keywords£ºparallel algorithm£» image processing£» line detection