Tracking moving objects using adaptive resolution

An algorithm is presented to visually track moving objects in a natural environment. Although various methods already exist to solve this task they usually fail to detect the precise motion of the target when the background gets more complex. They also require high computing power due to time consuming calculations of correlation-type motion detectors. The presented approach uses adaptive resolution to reduce the size of a window for comparison of two image regions. To determine a suited size the information given by a simple motion detector is integrated. This improves the performance of the system in terms of both speed and robustness