Hi,
i don’t think our solution will work here. What we are doing is a more simple approach. I would call it “progressive enhancement using multi sampling”. We only enable it if the user is not intracting with the scene to improve the quality of the “static” image. As soon as the user starts the interaction again we disable it and fallback to hardware anti aliasing. If we would enabled it while the user is moving an object for example we would see very strong ghosting (like motion blur).
Our solution works like this:
- between render frames we modify the camera projection matrix to create a subpixel camera jitter
- using post processing we store the current render frame and combine it later with the next camera frame
- we combine up to 32 frames