Object Representation with Local Features in Geodesic Distance Space

K. Slot and M. Gozdzik (Poland)


Object recognition and feature extraction


The following paper proposes a method for object representation that employs a local feature-based object description in a geodesic distance space. Local features are known to provide recognition robustness against affine object transformations, including rotation and scaling, as well as partial occlusion. Geodesic distance has been proved to provide an attractive metrics for representing object’s deformations. It has been shown that the proposed object representation – a set of features arranged in a geodesic space – allows for correct recognition in case of several types of nonlinear deformations.

Important Links:

Go Back