Reconstruction of frames

[ianphillips]

This is probably a stupid question, but it's one that has bothered me for a while. All film frames can be translated into pixels and pixels are dots of color, so why doesn't anyone recreate missing frames? Sure it would take a lot of time, skill, and hard work, but a determined fan could do it. If I printed out a telesnap of any episode and gave it to you and told you to recreate it perfectly on gimp or photoshop or whatever then you would probably be able to do it. What is the difference between making that and creating a frame from scratch? So my question is, why haven't any fans ever recreated a single frame of a missing episode ever?

[RWels]

Can you give an example?

[R. Williams]

Total non-starter. PAL DVD resolution is 720x576, making 414720 individual pixels to get right. Grayscale images permit 256 different shades, so this means there are a total of 256⁴¹⁴⁷²⁰ possible different frames. This is an incomprehensibly large number, and the Sun will expand and destroy the earth before humanity could even make a fraction of a dent in it.

Of course this assumes that you just generate the frames randomly, but it gives you a sense of the scale of the problem. Don't forget that there are about 37500 frames which make up a 25 minute episode.

However, it is possible to interpolate between consecutive telesnaps from the same shot (credit to The Chris of Fenric):

[Simon Collis]

All books are composed of words, so why has anyone not ever recreated War & Peace?

Let's assume that we're looking at a 405-line picture. There's roughly 202.5 lines per field to make 405 lines. Realistically, a percentage of that is hidden - I believe, although I can't find a source online, that it equates to "376i" in modern terms. So let's assume that we're going to recreate a single frame from a telesnap in photo-realistic greyscale.

Now, the picture is 4x3 in size - meaning for every 3 pixels down, we have 3 across. For the sake of keeping it simple, let's assume that's 500 pixels across. So that's 500 x 376, which equals 188,000 pixels we have to shade by hand. Multiply that by 64 and we get to 12 million possible combinations.

So to get photorealism, that's going to take a very long time - look at the videos on YouTube of the people who draw photorealistic portraits with pen on paper and that takes them hours. So I'm guessing that making a 188,000 pixel photorealistic recreation of a telesnap would take maybe 18 hours?

Whereas scanning it would take seconds, and I know it's exactly as accurate as my source material.

I haven't really got a point here, I find. I just enjoyed doing the maths (which no doubt someone will point out to me I've got completely wrong, as usual, but hey)...

[scotttelfer]

So what I think you are getting at is trying to create an ultra realistic animation that appears to be live action it's that good.

Answer is, one person making one frame a day will take 100 years to get through it. Really it will take a lot longer per frame, and even then you have problems of how accurate you got it. Couple of frames is easy to fill in, not so much with a couple of minutes or even a whole episode.

[Sue Butcher]

Oh my, that's such a classic Pat Troughton change of expression! All we need are a few thousand off-air snaps of a missing episode, and we've got it in the bag.

[George D]

If someone simply time exposed 1- 8mm frame every second, it could possibly accomplish what you suggest.
Just not enough data to make it work

[Sue Butcher]

That would have been a very economical way of recording TV programmes. You'd be able to fit two hours of vision on one 8mm reel. Or with slower sampling, a whole day's broadcast. I wonder if anyone ever tried this just as an experiment? I wish I had!

[RWels]

Sept 10, 2016 12:56:48 GMT R. Williams said:

However, it is possible to interpolate between consecutive telesnaps from the same shot (credit to The Chris of Fenric):

Suppose I needed something like this to bridge a 1 second gap in a fan restoration project. Just once. Would I be able to get some help, or be able to do that myself?

[Ace St.John]

Sorri


Doesn't look reel

[Greg H]

Oct 28, 2016 20:43:43 GMT RWels said:

Suppose I needed something like this to bridge a 1 second gap in a fan restoration project. Just once. Would I be able to get some help, or be able to do that myself?

Heya. Shouldn't be too challenging if you are reasonably skilled with a computer. I spent a bit of time messing around with this technique years ago and got some fun results with it.
Search the net for something like video morphing software. I would imagine there is freeware that will do the job and I know there is some expensive video software that will certainly do it. There is also free trial versions of some of the more heavy weight software. Hope that's of some help.

[brianfretwell]

All Vidfired episodes have frame/field reconstruction and I believe some single frames damaged by bad splicing/projector lamp burns were replaced this way. There was a system for internet compression by leaving out several frames then reconstructing them, but I suspect that as this was before broadband it wasn't needed after that became widespread and development was dropped.

[jayglover]

they also did a recon of frames in the war machines dvd. It is revealed on the special extra "Wotan Assembly"

[Sue Butcher]

That was around a second of missing frames as I recall, but it wasn't at all noticeable.