Here at Panda, we are constantly impressed with the requests that our customers have for us, and how they want to push our technology to new areas. We’ve been experimenting with more techniques over the past year, and we’ve officially pushed one of our most exciting ones to production.
Introducing frame rate conversion by motion compensation. This has been live in production for some time now, and being used by select customers. We wanted to hold off until we saw consistent success before we officially announced it 🙂 We’ll try to explain the very basics to let you build an intuition of how it works – however, if you have any questions regarding this, and how to leverage it for your business needs, give us a shout at email@example.com.
Motion compensation is a technique that was originally used for video compression, and now it’s used in virtually every video codec. Its inventors noticed that adjacent frames usually don’t differ too much (except for scene changes), and then used that fact to develop a better encoding scheme than compressing each frame separately. In short, motion-compensation-powered compression tries to detect movement that happens between frames and then use that information for more efficient encoding. Imagine two frames:
Now, a motion compensating algorithm would detect the fact that it’s the same panda in both frames, just in different locations:
We’re still thinking about compression, so why would we want to store the same panda twice? Yep, that’s what motion-compensation-powered compression does – it stores the moving panda just once (usually, it would store the whole frame #1), but it adds information about movement. Then the decompressor uses this information to construct remaining information (frame #2 based on frame #1).
That’s the general idea, but in practice it’s not as smooth and easy as in the example. The objects are rarely the same, and usually some distortions and non-linear transformations creep in. Scanning for movements is very expensive computationally, so we have to limit the search space (and optimize the hell out of the code, even resorting to hand-written assembly).
Okay, but compression is not the topic of this post. Frame rate conversion is, and motion compensation can be used for this task too, often with really impressive results.
For illustration, let’s go back to the moving panda example. Let’s assume we display 2 frames per second (not impressive), but we would like to display 3 frames per second (so impressive!), and the video shouldn’t play any faster when we’re done converting.
One option is to cheat a little bit and just duplicate a frame here and there, getting 3 FPS as a result. In theory we could accomplish our goal that way, but the quality would suck. Here’s how it would work:
Yes, the output has 3 frames and the input had 2, but the effect isn’t visually appealing. We need a bit of magic to create a frame that humans would see as naturally fitting between the two initial frames – panda has to be in the middle. That is a task motion compensation could deal with – detect the motion, but instead of using it for compression, create a new frame based on the gathered information. Here’s how it should work:
These are the basics of the basics of the theory. Now an example, taken straight from a Panda encoder. Let’s begin with an example of how frame duplication (the bad guy) would look like (for better illustration, after converting FPS we slowed down the video, and got slow motion as a result):
See that jitter on the right? Yuck. Now, what happens if we use motion compensation (the good guy) instead:
It looks a lot better to me, the movement is smooth and there are almost no video artifacts visible (maybe just a slight noise). But, of course, other types of footage are able to fool the algorithm more easily. Motion compensation assumes simple, linear movement, so other kinds of image transformations often produce heavier artifacts (they might be acceptable, though – it all depends on the use case). Occlusions, refractions (water bubbles!) and very quick movement (which means that too much happens between frames) are the most common examples. Anyway, it’s not as terrible as it sounds, and still better than frame duplication. For illustration, let’s use a video full of occlusions and water:
Okay, now, let’s slow it down four times with both frame duplication and motion compensation, displayed side-by-side. Motion compensation now produces clear artifacts (see those fake electric discharges?), but still looks better than frame duplication:
And that’s it. The artifacts are visible, but the unilateral verdict of a short survey in our office is: the effect is a lot more pleasant for motion compensation than frame duplication. The feature is not publicly available yet, but we’re enabling it for our customers on demand. Please remember that it’s hard to guess how your videos would look like when treated with our FPS converter, but if you’d like to give it a chance and experiment a bit, just drop us an email at firstname.lastname@example.org