Accuracy of the straight line Hough Transform: The non-voting approach

摘要

The Standard Hough Transform (HT), or the SHT, has established itself as a de-facto default technique of straight line HT evaluation. Its accuracy has been sporadically addressed in the past typically for justifying modifications within the SHT voting framework. Assessing SHT’s accuracy in previously less elaborated aspects, this paper comes to a conclusion that the voting nature of the SHT limits its ability to resolve a number of existing accuracy deficiencies. The paper considers a non-voting alternative called Hough–Green Transform (HGT). Quantitative accuracy assessment is conducted on both artificially simulated and real images using an approach based on image reconstruction from projections. The HGT is shown to surpassing the SHT significantly while being almost as accurate as the continuous Radon or Hough Transforms.