This paper presents a study of anisotropy in digital images, i.e. the detection of main directions and the quantification of their occurrence rate. Human vision is usually very powerful for such a feature-based analysis, because it simultaneously performs a multi-level analysis, from the local inspection of details to the more global analysis of spatial distribution of patterns. We show that our method is able to perform the anisotropy feature analysis using this global approach, unlike classic methods of directions analysis. Moreover, our method directly processes grey level images, uses the inner part of the patterns instead of their contours and is able to inspect all the directions of a picture. These specifications eliminate most of the limitations of usual methods.
[1]
M. Coster,et al.
Précis d'analyse d'images
,
1989
.
[2]
Wen-Hsiang Tsai,et al.
Scale- and orientation-invariant generalized hough transform3-a new approach
,
1991,
Pattern Recognit..
[3]
H. Maître.
2 - Un panorama de la transformation de Hough
,
1985
.
[4]
Murat Kunt,et al.
Traitement numérique des signaux
,
1996
.
[5]
Claude Bonnet,et al.
La perception visuelle des formes et du mouvement
,
1988
.
[6]
Jean Serra,et al.
Image Analysis and Mathematical Morphology
,
1983
.
[7]
R Ghiglione,et al.
Traité de psychologie cognitive - 1 : Perception, action, langage
,
1989
.
[8]
M. Denis.
Image et cognition
,
1989
.