Stochastic search approach to object location

Object location within an image involves establishing a location within an image for which a particular error function is minimised, and thus is essentially an optimization problem. Within the content of location via template matching, a random search approach, called a cluster search, based on an extension of mathematical methods due to Matyas has been studied to determine the most appropriate parameter /spl beta/. For well chosen /spl beta/ the cluster search has been shown to be on average significantly faster than the fastest deterministic search schemes, notably coarse fine approaches.<
>