Sorry but I can't see the point of using images in this situation or technique (rotating layers for 3D).
The shapes and effects are simple enough to be done in AS with vector layers and then there would be absolutely no pixelation ever.
I could see using an image here or there in the background elements that are "static" but NOT for that water layer (or the trees in the first sample). If the camera zooms in you will ALWAYS get some "fuzzing" or blurring of the image unless the resolution is HUGE.
Each ENTIRE image has to be as big as the closest point to the camera it will be seen. This would require gigantic images. For the water example this would mean "infinity"

. This is the nature of raster. When you zoom in it is just like resizing a small image to a larger size, the details blur and soften, tiny pixels become huge blocks. It is like an "old style" first person 3D game. The closer you get to an object or wall the worse it looks.
It could be done but at what cost in render time? The massive resolution would probably be wasted 90% of the time just for those few moments when the camera is close to it.
Just my opinion.
-vern