The tracking problem is distinguished from the recognition problem in that its search processes are more local rather than global. The problem of locating a face is a recognized problem. In order to locate a human face, the system needs to capture an image using a camera and a frame grabber, to process the image, to search the image for important features, and then to use these features to determine the location of the face. In order to track a human face, the system not only needs to locate a face, but also needs to find the same face in a sequence of images. This requires the system’s ability to estimate the motion while locating the face. Furthermore, the system needs to control the camera, e.g., panning, tilting, and zooming
to track faces outside a certain range