Originally Posted by Raistlin_HT
If enough 'future' frames are analyzed when you are doing the interpolation of previous frames, the I would disagree - you shouldn't really have many problems with the 'guesses'.
Ok, let's imagine we're watching some fireworks on TV. And let's imagine there's a ball flying through the image at the same time - in front of the fireworks. Now in frame 1 the ball is in the left part of the image. If the ball is flying really fast, it might be in the midst of the image already in frame 2 and on the right side of the image in frame 3. At the same time the fireworks looks different in each frame. Now how can a video processor "guess" a frame between frame 1 and 2? The VP would have to find out that there's ball flying from left to right. So the VP would have to place this ball in an intermediate position on the screen for the interpolated frame. But there's some part of the fireworks which was covered by the ball in frame 1 and another part of the fireworks covered by the ball in frame 3. Now the VP has to guess how the fireworks could look like in the intermediate frame. But there's lots of information missing about the fireworks, because big parts of it were covered by the ball.
Do you see what I mean? We humans with our intelligence would probably be able to create a good intermediate frame, but we'd have to use some guessing ourselves. A computer program (which is basically what a VP runs through) has a much much harder time doing such things, cause it has no real intelligence. The VP would have to understand every little bit that is happening on screen to be able to create good intermediate images. We humans know "oh, this is fireworks, this usually looks like this" and "oh, this is a building, when the fireworks flies there, the shadow of the building will likely change this way", but a VP doesn't have this kind of information.
Basically, there will be situations where the VP will not have enough information to create interpolated images that look good to our eyes/brain. There will be situations where some parts of the image are almost always covered, but shine through a bit, nevertheless. We humans will see some minor parts of covered things and know instantly what they are, which form they have and how they usually move. A VP doesn't recognize "well known objects" like this. It only sees pixels and geographical forms. A VP doesn't know how specific things usually move/behave.