An apparatus and method for supporting augmented reality or other computer vision applications are presented. Embodiments enable communication between natural feature and auto-focus engines to increase an engine's accuracy or decrease a processing time of the engine. An auto-focus engine may communicate a location of an auto-focus window to a natural feature detection module and/or a change in location of a previous auto-focus window to a next auto-focus window. The natural feature detection module uses the communicated information to limit an initial search area and/or to set a next tracking search window. A natural feature tracking module may communicate a changes from a previous location of a natural feature to a next location of the natural feature to an auto-focus engine. The auto-focus engine uses the change to set a next auto-focus window.