Have a look at this work: http://www.jprr.org/index.php/jprr/article/viewFile/355/148
It covers scale, rotation and translation ... I dont know if you can get this running in real time, but it is a very nice paper and definitely a good starting point.
An algorithm was proposed in 2005 by Hong and Schowengerdt based on many control points (around 50) found automatically, by correspondence matching of edge gradient contours in both SAR and optical images.
The algorithm was tested for rural and urban areas and proved robustness. However, it works well only on reasonably flat terrain.
I've attached the paper describing the method and findings.