|
We proposed a simple but extremely robust visual tracking framework that can adapt to the appearance changes of the object efficiently. Via our proposed visual tracking framework, the "well-known" drifting issue plauged with most of visual tracking algorithms can be solved nicely. In addition, via the proposed framework, an accurate confidence measurement can be provided to assess each tracked object so that the tracking failures can be identified automatically. The above two significant contributions can be demonstrated by the successful application in the multi-view face tracking under significant head movements, facial expression changes and partial occlusions.
|
