Title :
New multiscale planar shape invariant representation under a general affine transformations
Author :
Daoudi, Mohamed ; Matusiak, Stanislaw
Author_Institution :
Dept. of Comput. Sci., ENIC/INT, Villeneuve d´´Ascq, France
Abstract :
We introduce a set of invariant and local descriptors, which are independent under a general affine group transformation (GA(2): rotation, uniform scaling, translation, and stretching) of planar curves: a new invariant in affine scale-space (IASS). For their extraction, we propose a method based on multiple convolutions between the affine-parametrized curve and Gaussian kernel. The IASS representation of planar curves generalizes, in the affine case, a curvature scale space invariant description under similarity transformation (rotation, uniform scaling and translation)
Keywords :
computer vision; geometry; group theory; image segmentation; invariance; GA(2); Gaussian kernel; affine scale-space; affine-parametrized curve; general affine transformations; local descriptors; multiscale planar shape invariant representation; planar curves; rotation; similarity transformation; stretching; translation; uniform scaling; Cascading style sheets; Computer science; Computer vision; Feature extraction; Image recognition; Image reconstruction; Quantization; Robust stability; Shape; Topology;
Conference_Titel :
Pattern Recognition, 2000. Proceedings. 15th International Conference on
Conference_Location :
Barcelona
Print_ISBN :
0-7695-0750-6
DOI :
10.1109/ICPR.2000.903662