[This](https://www.youtube.com/watch?v=qt7HNKae9e0) video shows how well it works and what you need for it to work, but he never explains how it works. How does it work?
In: 2
30 to 60 fps is easy, you just insert a “made up” frame between each frame. The difficult bit is how you make up the new frame.
One simple strategy is to simply pixel for pixel take the average value of the pixels in two sequential frames and use that to make a new frame. However since most moving stuff moves further than a couple of pixels per frame, this would result in an odd blurry mess.
Lets call some frames 0, 1 and 2.
0 is the current frame
2 is the next real frame
1 is the frame we want to generate.
A reasonable algorithm works something like this:
1. Pick a small square of pixels from image 0
2. Search image 2 to find the most similar square of pixels nearby.
3. Work out the vector this bunch of pixels has moved and save that info
4. Repeat 1-3 for next small square of pixels until all areas of the image have been mapped.
5. Warp frame 0 based on the vector map so that everything moves half the distance between frames 0 and 2.
This is very crude and would come up with a bunch of artifacts and odd behaviour, however it’s the crux of how all these algorithms work. Some algorithms may look more than 1 frame forward/back to better track features.
Latest Answers