An adaptive method has been developed for the automatic tracking of objects from a sequence of digital television-image signals. The method is based on a combination of the segmentation and matched-filtering methods. Invariance to changes of the size, shape, and statistical characteristics of the object is achieved by regenerating the data of the reference array. Matched filtering is accomplished using a grey-scale image in the region of higher spatial frequencies from a binary mask of the reference array. Dynamic signal accumulation provides high noise immunity, and this makes it possible to track objects on a complex background with a small SNR.
© 2007 Optical Society of AmericaPDF Article