A new optimization criterion for extracting several straight lines in a binary image

This paper introduces a new algorithm for extracting several straight lines in a binary image. This algorithm is based on the optimization of a new criterion called “Line-Kproduct” (LKP) which is derived from the K-product algorithm we introduced in a previous work for unsupervised mixture estimation. A relaxation algorithm is proposed to find the minimum of LKP. Contrary to the classical Hough transform (HT) based methods our approach does not require the tricky configuration of extra parameters. Simulations finally illustrate the superiority of our approach compared to the HT-based algorithm.