𝔖 Bobbio Scriptorium
✦   LIBER   ✦

Variable generalized hough transform based on error analysis of curve gradient

✍ Scribed by Masashi Morimoto; Yasuhito Suenaga


Publisher
John Wiley and Sons
Year
2000
Tongue
English
Weight
372 KB
Volume
31
Category
Article
ISSN
0882-1666

No coin nor oath required. For personal study only.

✦ Synopsis


The generalized Hough transform (GHT) is known as an extension of the Hough transform to extract arbitrary shapes from images. With the GHT, errors in the gradient of object boundary curves result in deterioration of performance and decreased accuracy of object extraction. In this paper, the influence of gradient errors on histograms in GHT parameter space is analyzed, and the results of the analysis are used to develop a variable generalized Hough transform (V-GHT), being an extension of the variable Hough transform (VHT) that was reported earlier by the authors. With the proposed method, the PDF or standard deviations are found for calculated parameters to design a variable filter; this variable filter is then applied to parameter space, aiming at better performance and accuracy. Evaluation experiments with circles prove that the proposed V-GHT offers a better hit rate, error rate, and extraction accuracy. The proposed method was also validated experimentally in terms of moving object tracking.