Markerless method is useful for tracking real objects in augmented reality. As far as I know, markerless tracking can be done by tracking and matching. Is there another tracking method or any hybrid methods that use both?
Maybe the simplest approach is the POSIT algorithm (http://code.opencv.org/projects/opencv/wiki/Posit). I've seen it applied for augmented reality (http://www.youtube.com/watch?v=_Dln257k2Sc)