Efficient computational geometry method for detecting dotted lines in noisy images

Abstract

In this paper we present an efficient O(n log n) time, linear space, algorithm for detecting a line, or line segment, represented by a set pf nL collinear points contained in a rectangular window with an additional set of nN independent, uniformly distributed random noise points; n = nL + nN. Empirical results show that the algorithm is very reliable for nL/nN ≥ 1/3.