Currently the dirty region optimization doesn't optimize the case where a large object or image is drawn in the background with dynamic controls drawn on top. In this case, if anything on top of the large background object changes (e.g., a scrolling or animating control), the entire scene is repainted, thus defeating the dirty region optimization.

We should consider a form of visibility (occlusion) testing as part of the dirty region optimization to optimize the case where we can determine that no portion of the underlying image has been dirtied by a change to nodes on top of it.