摘要 |
The invention relates to detecting and tracking objects in a sequence of images. In particular, a method, software and system for tracking a non-rigid object in a plurality of images. Initially, a first set of parameters for a parameterised shape model are generated (410) based on a first image. Then a second set of parameters are generated (415) for the parameterised shape model by fitting the parameterised shape model to the object in a second image of the plurality of images. This fitting is according to (i) the one or more constraints defined independently of the object or (ii) the first set of parameters. Also, the first and second sets of parameters define a tracking of the non-rigid object between the first and second images. |