Tracking an object in a video stream

摘要

A method of tracking an object such as a face 12 in a video stream is disclosed comprising running an object detector at a plurality of first locations 1 on a first frame, defining a coarse grid of base unit 10. This is repeated for second and subsequent frames by running an object detector at a plurality of second locations 2 and subsequent locations having a grid that is slightly offset each time so that eventually all the points on the fine grid are covered but in several passes. When an object such as a face 12 is located on one frame such as at position 18, the positional and/or scale information is propagated to the next frame such as at position 20 to assist in the tracking of that object in the next frame.