Frame Interpolation and The Videophile - Page 5

Two points to make:

1. The problem with motion interpolation is that it is having to create new frames that weren't on the original film.

Imagine we have a football moving left to right on screen.
frame a:
O
and frame b:
O
but then we have to insert a new frame in between.
Then the ball should be in between:
O
Even this simple task is very tricky.
Firstly the software has to recognise the object that has changed position - the ball.
Secondly it has to cut out the ball and move it into an interpolated position - in the middle.
But it also has to not show the ball at the left-hand or right-hand position. It has to instead show whatever was behind the ball - presumably grass or something.
What a CG digital compositor would probably do is grab from frame b the bit of grass that was behind the ball in frame a, and grab from frame a the bit of grass that was hidden in frame b.

But this simple task would take an effects artist a few minutes per frame at least.

Your TV is having to do this in 1/24 of a second.

Which is impossible.

So it cheats. It uses warping and other motion compensaton/interpolation techniques to create inbetween frames that sort of fool the eye.

But what we end up with is every other frame being a frame made up by the computer as a poor approximation of what that frame would look like had it ever been filmed - with all the errors, artifacts, and inaccuracies of a process done at incredibly high speed.

And not surprisingly it looks weird.

For half the film you are watching images that no camera ever filmed, that no editor ever saw, and no director ever intended for you to see.

But my second point is:

When we go the cinema to see a film, or even when we see a movie on television, we subconsciously are seeing characteristics and artifacts of the picture.

If you switch channels, I reckon most people could tell within a second whether what's playing is a movie or a TV program.

Nowadays, as more movies are shot digitally, filmmakers are trying hard to work out what it is that makes a movie looks like a movie. Then they're trying to alter the digital picture to replicate that look.

There are a number of things.

1. Depth of field. Simply put this means that in films, very often the foreground will be in sharp focus, whilst the background will be blurry/out of focus. TV camera lenses have traditionally been unable to replicate what we call that 'shallow depth of field'.

2. Grain versus noise. Both video and film picture are the combination of little dots. But film is made up from a random pattern of dots, whereas video is made up of absolutely square pixels. And it's interesting that while stills photographers sometimes blow up grain, loving its natural texture, almost nobody ever loves video pixel noise. Some argue that it's because the eye/brain prefers natural texture to artificial.
But certainly, when trying to make video look like film, one thing you can do is reduce the video noise, and actually degrade the picture a bit by adding a bit of sampled film grain. There are programs which can simulate the grain given by different types of film.

3. Gamma. It's complicated but basically film and video react to light in different ways. Film still sees more detail in the brightest and darkest parts of the picture and sees colour in a different way. Another thing that filmmakers do to make digital images look like film is to alter the gamma, either in camera or in post-production.

4. Frame-rate. To bring this back to the original subject, traditionally the biggest give away that one is watching a video-originated image and not film is the frame rate. In the US, video cameras shoot 60 frames a second. They cheat a bit to do this. Imagine the picture broken down into horizontal lines. Then imagine that these lines are numbered 12345, etc. On the first frame, video cameras record lines 2,4,6,8, etc. Then on the second, they record 1,3,5,7,9 etc.
This is known as interlacing.

Nowadays, we can set digital cameras to not do this. We change them to film in what's known as a progressive' mode, which shoots full 24 frames a second.

In the old days, filmmakers used to go to great lengths to get rid of this interlacing.

We would literally throw away all the information for lines 1,3,5,7,9 etc., and duplicate the information for lines 2,4,6,8.
Why?
It was clear that the higher frame-rate was a give-away that the image had been shot on video not on film. So even if we degraded the picture somewhat by throwing out half the data, it still sort of looked better'.

Why would it be so important to filmmakers (most of us anyway) that we shoot something that looks like film, even if it isn't?

I think the answer is because audiences react differently to video and film. They don't know they're doing it, and it's working almost entirely on a subconscious level, but when someone thinks they're watching a movie, their mindset is that this is something more special, bigger budget, more worthy of their attention.

So imagine the frustration of filmmakers when new TVs undo all of the work we have put in to making something look like film and make it look indeed like a 'soap opera'. Modern TVs can get rid of all the lovely film grain' we've programmed in - they brighten and alter the picture undoing all our work to match a film gamma, and worst of all, they increase the frame rate, adding digital artifacts at the same time.

I don't know of any filmmaker who would want you to watch his film looking like that.

So, as a previous poster said - if you're watching sports or video games, or indeed soap operas, fine.
But if you're watching a movie, and you want to watch it the way the director made it, switch as much of that video enhancing crap off as you can!

Paul Spurrier

Originally Posted by Mikenificent1

Maybe with your brain, but not mine, I see judder all the time on film and hate it.