Output details

Automatic Bootstrapping and Tracking of Object Contours

<23> A new contour-based object-tracking framework was introduced in this work. It is fully automatic with a novel motion-based bootstrapping, in contrast to manual initialisation which is a routine practice in contour based tracking. The framework allows simultaneous deformable object tracking and online shape learning. It shows significant improvements in handling occlusions and noise interference, which are fundamental issues in Computer Vision. This new framework provides a new paradigm for contour-based object tracking and a new approach to track object not only their position but also their shape evolution. It is published in a top Image Processing journal.