Abstract
We have adapted a Progressive Probabilistic Hough Transform (PPHT) as presented by Galambos et al. in [1] to the detection of circles of multiple radii. The PPHT is found to be more computationally efficient than the Standard Hough Transform. We have found that, particularly for natural images, the PPHT is a poor technique to use for extraction of geometric primitives with more than two parameters, such as circles. This is due in part to the statistical/probabilistic nature of the transform, and in part to the distribution of votes that are cast in accumulator space for circles.
