Community Research and Development Information Service - CORDIS

Abstract

Many applications of computer vision and image processing require the localisation of straight lines in images, or in edge detected images. In most cases the precision with which the lines can be located is of practical importance, and the question of whether or not sub-pixel precision can be achieved can be crucial. The paper presents a line-following algorithm which performs a least- squares straight line fitting during the following process, and which locates edges of sufficient length and contrast to fractional pixel accuracy. Another process, based on Gaussian fitting using a standard minimisation routine to refine the parameters of the found lines, is presented. This method improves the accuracy and can locate edges to 1/10 th pixel precision. An adaptation of the resulting algorithm to operate on zero crossings in the output of a Laplacian of a Gaussian filter, with further improvement in performance, is also described. Experimental results comparing the three processes and demonstrating the achievement of sub-pixel accuracy are presented.

Additional information

Authors: PECKHAM R J, JRC Ispra (IT)
Bibliographic Reference: Paper presented: British Machine Vision Conference, Leeds (GB), Sept. 21-24, 1992
Availability: Available from (1) as Paper EN 36777 ORA
Record Number: 199210617 / Last updated on: 1994-12-02
Category: PUBLICATION
Original language: en
Available languages: en
Follow us on: RSS Facebook Twitter YouTube Managed by the EU Publications Office Top