Geometry of the Hough transforms with applications to synthetic data (Q2071520)

The Hough transform is a standard technique for feature extraction. It is used in a lot of applications and in particular, in image analysis and digital image processing. In addition, such a technique has initially been used to detect curves in images. In this paper and using geometrical arguments, the authors provide a bound for the number of Hough transforms to be considered for a successful optimization of the accumulator function in the recognition algorithm. As a particular case, it is shown the robustness of the presented results when applied to synthetic datasets strongly perturbed by noise. For the exact case, it is provided an algebraic approach that leads to a better bound of theoretical interest.