thank you for the explanation and examples. For my viewing method the
small images embedded in the text are not the best viewing. Could you
possibly add these images on your StereoPix site if you haven't already
I sketched the illustrations and deleted their original after I sent the message, sorry.
However, I rebuilt an image based on the LRL of the previous mail so that it can be displayed by a 3D viewing software (see attachment).
There are some inaccuracies (zero is not exactly on the screen for example) due to a quick manipulation, but the point is still illustrated.
Here's a more extreme example of mine that may help:
Not sure it is a good example in the sense that it mixes several techniques. It has a well executed 3D window, though.
Unfortunately, it remains a window violation in the bottom left corner, but I should admit that it is almost not noticeable.
Floating windows are a very good way to permit moving objects into negative space (in front of the screen or other image plane), of which there is much more room than positive space (behind the screen or other image plane). Since convergence is normal and divergence is not, you can easily fuse objects that are many times more in front of the screen than behind it.
If you set the window at 2 meters (in real space) in front of the camera, you have only two meters of potential negative space (and a part of it is too close to be fused by most people) and far far more distance where the depth is perceived behind (assuming appropriate settings). I guess that you mean something else.
Since stereoscopic displacement is lateral only, window violations occur only at the sides, not the top or bottom edges.
But the view defined by a window is a pyramidal volume, so it makes sense to me to name top and bottom objects outside of this volume as being in window violation as well.