著者

鈴木 寿

出版者

一般社団法人日本応用数理学会

雑誌

応用数理 (ISSN:09172270)

巻号頁・発行日

vol.9, no.3, pp.207-219, 1999-09-16

被引用文献数

2

---

This article compactly describes progress of researches on the Hough transform that is famous as one of straight-line detection methods in fields of image processing. Many versions revised from the original Hough transform have been developed, and the generalized Hough transform can detect not only straight lines but also figures of arbitrary shapes. An important problem in grayscale image processing is to calculate the ratio of magnification, the angle of rotation, and the quantity of parallel translation between a template image and its varied, noisy image. Known solutions for this problem can be classified to characteristic point extracting methods, normalized correlatiom methods, moment methods, frequency space methods, energy minimization methods, θ-p Hough transform methods, generalized Hough transform methods, and others. Generalized Hough transform methods are robust even if the quality of images is low; nevertheless, practical applications were difficult since their computational complexity is not small. Various methods based on the Hough transform are realizable and practical today since computers have been highly advanced.