Method for improving tracking using dynamic background compensation with centroid compensation

A method for tracking an object across a number of image frames, the method comprising:

• identifying a region containing the object in a first image frame to be stored as an exemplar view of the object;
• creating an initial appearance model based on the region in the first image frame and a background region in the first image frame;
• and then for subsequent video frames in a video sequence
• determining at least one of a location and size of a predicted region for tracking the object in the subsequent image frame using at least one of the initial appearance model and an updated appearance model;
• correcting at least one of the determined location and size of the predicted region in the subsequent image frame in accordance with at least one of the location and size of the region in the first image frame corresponding to the exemplar view of the object;
• and
• creating an updated appearance model.

A non-transitory computer readable storage medium having a program recorded thereon, the program being executable by a processor to track an object across a number of image frames, the program comprising:

• code for identifying a region containing the object in a first image frame to be stored as an exemplar view of the object;
• code for creating an appearance model based on the region in the first image frame and a background region in a second image frame;
• code for determining at least one of a location and size of a predicted region for tracking the object in the second image frame using the appearance model;
• and
• code for correcting at least one of the determined location and size of the predicted region in the second image frame in accordance with at least one of the location and size of the region in the first image frame corresponding to the exemplar view of the object.

Apparatus for tracking an object across a number of image frames, the apparatus comprising a processor coupled to a memory in which are stored the image frames and a program, the program comprising code means for performing the steps of:

• identifying a region containing the object in a first image frame to be stored as an exemplar view of the object;
• creating an appearance model based on the region in the first image frame and a background region in a second image frame;
• determining at least one of a location and size of a predicted region for tracking the object in the second image frame using the appearance model;
• and
• correcting at least one of the determined location and size of the predicted region in the second image frame in accordance with at least one of the location and size of the region in the first image frame corresponding to the exemplar view of the object.