Résumé:
The objective of this work is to design and develop a shape descriptor for
describing and matching of 2D shapes, more robust and e cient than other
existing descriptors. The proposed approach introduces a new topological descriptor called the size function (or SFD) based on the representation of contour points distributed in annular curves according to sublevels calculated by
the radial measurement function, which allows to represent and characterize
the shape in a simple and more reduced way. The size function describes the
shape based on the characteristics of the real functions de ned on the objects
of the shape. These functions allow to quantify the geometric and topological
properties of the shapes, thus providing powerful descriptors for their characterization.
Thanks to the topological information contained in this new descriptor, it is
more e cient to match shapes that are not strictly similar. In this regard, a
matching algorithm is designed to compare a query shape with a collection of
target shapes. The similarity score of the matching is evaluated according to
the normalized distance interval [0..1].
We have validated the proposed approach using a standard 2D shape database.
The obtained test results are favorable and promising and show the success
and e ciency of this approach. These results also show that the new descriptor
model is competitive with other descriptors for description and matching tasks
and is more robust to noise, occlusions and invariant to geometric transformations. In the future, this descriptor may be a good candidate to successfully
apply it to various classi cation and recognition problems