Summary: Algorithmic Evaluation of Line Detection Problem
School of Information Science,
(Japan Advanced Institute of Science and Technology),
Asahidai, Tatsunokuchi, Ishikawa,
TEL: 0761-51-1295 (direct)
The Hough transform is a well-established scheme for detecting digital line components
in a binary edge image. A key to its success in practice is the notion of voting on an
accumulator array in the parameter plane. This paper discusses computational limitation
of such voting-based schema under the constraint that all possible line components in a
given image must be reported. Various improvements are presented based on algorithmic
techniques and data structures. New schema with less computation time and working space
based on totally di erent ideas are also proposed with some experimental results.
Index Terms | computer vision, computational geometry, duality transform, Hough
transform, topological walk.