Skip to Main content Skip to Navigation
Conference papers

Transformées de Hough denses : un cadre efficace et générique pour la reconnaissance de formes paramétrées

Abstract : More than fifty years old , the Hough transform (HT) is one of the first algorithms of computer vision. It is an elegant and generic framework to detect parametrised forms in pictures , that applies to both analytical (lines , circles ,. . .) and non-analytical shapes (objects). HTs have generated a large amount of works , regarding the types of shape , parametrisation , quantisation or acceleration of algorithms. Nevertheless , almost all proposed algorithms are sparsely applied to contours or salient points images. However , simply estimating the spatial derivatives is sufficient to densely apply a HT , directly from the gray level. In addition , for analytical forms such as lines or circles , this estimate provides a one-to-one correspondence from the image space to the parameter space , much faster than conventional projections (one-to-many or many-to-one votes). In this document we make a brief recall of the HT and its conventional approaches , before presenting the dense Hough transform (DHT) in the multi-scale derivatives framework. One-to-one DHTs for the fast detection of lines and circles are presented , then the generalised DHT and its application to object detection .
Complete list of metadata
Contributor : Antoine Manzanera Connect in order to contact the contributor
Submitted on : Wednesday, December 16, 2015 - 5:33:24 PM
Last modification on : Friday, December 3, 2021 - 11:34:10 AM


  • HAL Id : hal-01245133, version 1



Antoine Manzanera, Thanh Phuong Nguyen. Transformées de Hough denses : un cadre efficace et générique pour la reconnaissance de formes paramétrées. Traitement et Analyse de l'Information : Méthodes et Applications TAIMA'15, May 2015, Hammamet, Tunisie. ⟨hal-01245133⟩



Les métriques sont temporairement indisponibles