Most dramatic restoration vid?

[Mephesto]

What is the most dramatic video restoration you veterans have seen done on here?

While the open-source community has its drawbacks, I've always loved witnessing the power of the internet collective that brought us things like Wikipedia, Avisynth, codec listening tests and so on.

I wanna know from the more active veterans on here what the greatest accomplishment in video restoration they've seen.

Show me the most seriously messed up video that you've seen completely transformed.

[sanlyn]

Hm. "Seriously messed up", but "completely transformed"? Haven't seen one of those yet.

[Mephesto]

I said "most". I'm sure the most messed up vid to be restored didn't look like utter garbage.

[manono]

If you haven't seen them before, check out the VideoFred before and after videos at Vimeo, the ones such as this one:

http://vimeo.com/49963017

[Brad]

He lost me at "denoised".

[sanlyn]

I said "most". I'm sure the most messed up vid to be restored didn't look like utter garbage.

Ok, I was reasing a bit but I get yer point. Manono has a point with VideoFred's projects (most of it was done on Doom9, but that shouldn't disqualify should it?). Seriously, though, "most messed up" covers a lot of territory. I worked on a godawful opera VHS-DVD transfer for 16 months in this forum, was in terrible shape, but I wouldn't say I completely revamped it. I could have done more (and I still drag out a few scenes to play with new stuff I learned).

This would bear some looking, to find what would be the "most" messy and the "most complete" revamp. I'd start with work jagabo, poisondeathray, 2BDecided, manono, others...and I'd have to dig back in the forum a few years to find some stalwarts that haven't been around for a while.

So OK, it might take some of us a while to dig up worthy candidates. There are so many, don'tcha know.

[robjv1]

Part of the issue in identifying the right clips is that some of the most dramatic errors are easy fixes (visual distortions fixed with a TBC, color correction) and others require meticulous work to fix what amounts to very little noticeable improvement. The first method may take only the right equipment and a little time, while the latter can take many hours of work.

[Mephesto]

I'm asking for everyone to post what is in their opinion the most extreme case that has been greatly improved/restored.

Those VideoFred vids were pretty good. Why is the master copy blurrier if its a higher resolution? Has deblurring been done?
Got any other examples?

Part of the issue in identifying the right clips is that some of the most dramatic errors are easy fixes (visual distortions fixed with a TBC, color correction)

Most visual distortions are not a simple case of changing the hue. More commonly, you'll be stuck with oversaturations of one color that bleeds in certain parts without an obvious way to fix it.

[manono]

Here's one I made to show someone. Nothing really special or difficult about it though:

https://www.youtube.com/watch?v=yRjF3me6X6U

[Mephesto]

Very good, needs some RemoveDirtMC though. What did you use to stabilize?

[manono]

Stab is all.

[2Bdecided]

2BDecided

No, I never need to work with really bad sources of my own. Worst I have to tackle is noise in low light camcorder videos (and I'm never very happy with the results).

I do need to upscale very small video clips though...
http://forum.doom9.org/showthread.php?p=808013#post808013
...and the results can be dramatic.

Cheers,
David.

[robjv1]

Part of the issue in identifying the right clips is that some of the most dramatic errors are easy fixes (visual distortions fixed with a TBC, color correction)

Most visual distortions are not a simple case of changing the hue. More commonly, you'll be stuck with oversaturations of one color that bleeds in certain parts without an obvious way to fix it.

I agree. I'm just saying 'dramatic' is not always an indication of a 'complex' restoration.

[Mephesto]

Stab is all.

Bullshit, Stab couldn't fix the violent vertical pans that happened later in the video. The wobbling in general is more smooth compared to the usual jerky panning of only a couple pixels. I don't see this being the work of Stab at all.

I udo need to upscale very small video clips though...
http://forum.doom9.org/showthread.ph...013#post808013

Pretty cool, looks a lot better than Lanczosresize.

I agree. I'm just saying 'dramatic' is not always an indication of a 'complex' restoration.

What the ****'s up with nerd communities and its love for dancing with semantics? Would me using the word 'extreme' instead satisfy you?

[robjv1]

Geez, take a pill. I have no issue with your wording, just trying to get to the bottom of what you're looking for -- as was sanlyn. Manono posted a link to a YT video and I have a few other examples along those lines but wasn't sure if that fit the bill.

[sanlyn]

Okay, we're getting closer. The key phrases in the original post: "What is the most dramatic video restoration you veterans have seen done on here?", "the greatest accomplishment in video restoration they've seen:, and "the most seriously messed up video that you've seen completely transformed."

Those are tall orders in themselves, but what I get from this is:

1. Work done here in this forum, not some other forum.

2. Dramatic = serious imaging and/or audio problems, mistakes, screw-ups, bad captures, mangled tape, etc. "Seriously messed up" wouldn't be 2 bad frames and a frame hop in a 20-second video. It would have to be worse than that.

3. I take it that "completely transformed" doesn't necessarily mean "absolutely perfectly transformed". I guess that leaves out minor improvements like removing a slightly blue color cast.

So if those requirements seem more or less self-defining, let's proceed.

Doesn't make it all that easy. Seen some godawful video around here. The trick is picking the worst input and the biggest improvement on output.

[manono]

What did you use to stabilize?

Stab is all.

Bullshit, Stab couldn't fix the violent vertical pans that happened later in the video. The wobbling in general is more smooth compared to the usual jerky panning of only a couple pixels. I don't see this being the work of Stab at all.

Cute. You asked and I told you. Being called a liar doesn't really encourage anyone to want to help you out or even to be in any thread you're in. That heavy jerking after the live action began I wouldn't really call stabilize, at least not in the sense that most of that piece has a nearly continuous jitter which Stab handles pretty well. It fixes the jitter caused by the sprocket holes in the film being enlarged over time. Obviously, Stab wasn't designed to handle the stronger stuff. I used frame interpolators for the bigger jumps.

[sanlyn]

However, 300 frame hops in a 20-second video would likely qualify as "damage". More offerings on the way.

[Mephesto]

rob,
I don't think my post was that cryptic.

manono,
Frame interpolator explains it. I guess you technically are a liar when you said Stab is all that was used.

sanlyn,
Everything you said is mostly true. I'm writing a paper on amateur Internet communities and their accomplishments in several fields, so it doesn't have to be limited to videohelp. Doom9 counts too. Any non-professional community that used tools available on the web counts.

I realize the problem when it comes to finding a qualified extreme of things, though. You end up with results that deviate from the orthodox definition of the object.

E.g. You ask what the longest movie ever made is because you wanna give it a world record medal. Sorting by length in the film registry gives you a 40-hour-long piece that consists of a poet reading mountains of poetry on camera. You say that doesn't count as a real film because it has no plot or acting.

Then you find a 6-hour-long Chinese movie about China's history which has actual acting with all the qualities of a "regular" movie but punctuated with long lectures and reviews by historians. You say that doesn't count, at least not the boring lecture parts so you consider its real length 3 hours.

Then with further digging you're stuck at the 3-4 hour range because you can't decide which of these 100 titles count and which don't. Finally, you give up and just filter by popularity or number of oscars/awards and then sort by length and end up with Titanic which is 3 and a half hours.

So when I asked for the most extreme make-over you've seen a video get. Do yourself a favor and don't think too hard, just post what you think counts.

[manono]

Frame interpolator explains it. I guess you technically are a liar when you said Stab is all that was used.

Look, I know you're not stupid so you do this on purpose, right? Unless you really believe your own nonsense. Which would make me wrong and you stupid. You asked what I used to stabilize it and I answered stab. As in 'stab is all that was used to stabilize it'. Of course I used other things. One look at it and any fool can tell a bunch of stuff was used. So cut the crap. Interpolating isn't stabilizing. It's creating entirely new frames

[Mephesto]

[QUOTE=manono;2229684]

Look, I know you're not stupid

Funny, I've lost count how much you shat on my intelligence in the past.

so you do this on purpose, right? Unless you really believe your own nonsense. Which would make me wrong and you stupid. You asked what I used to stabilize it and I answered stab. As in 'stab is all that was used to stabilize it'. Of course I used other things. One look at it and any fool can tell a bunch of stuff was used. So cut the crap. Interpolating isn't stabilizing. It's creating entirely new frames

It was late last night when I watched it but now I notice the frame interpolator on the violent shake parts with the walking guy's morphed face.

Interpolating isn't stabilizing.

Frame interpolation as you've proved can be used as a stabilizer when de-panning won't cut it.

Either way, I was commending your restoration. so why so quarrelsome?

[Cornucopia]

Well, I doubt it is the MOST dramatic restoration vid, but I had a project that I worked on (in a past life, with prev. production co.) where my demo/archival footage got ruined and I no longer had access to the originals. And I mean RUINED!. You can read about it here (it was the first thread I posted a question to): https://forum.videohelp.com/threads/27866-Any-way-to-save-this-BAD-encode
I would say this ranks up there.

Anyway, at the time, I never got an answer to it, so I spent quite a bit of time on my own and was able to re-create a reasonable restore. It's far from perfect, and I know there is still much to be done to it, but to be honest time has past and I don't care about that project with such fervency as I originally had. I've resigned to letting it stay in the semi-restored version I find at least acceptable.

I daresay, there are folks on this site now that weren't around then who could probably whip it completely into shape. If so, I gladly submit it for your inspection or you could even try your hand at it...
What happened is: I edited a short movie for a client intent on submitting to a film festival. His footage was a combination of VHS and 16mm xfers bumped over to Betacam, so it was never that great to begin with. Editing was done on the old Media100 (Mac) system, with archival/demo sample encoding to have been done through Cleaner5. Something went wrong and I never got to view the material before it was burned to disc and I moved away.

Here is a 10sec clip of the extant original (mpeg2 output from Cleaner), along with an XVID-reencoded sample from the restored version. It has a very strange problem that I haven't ever encountered before or since, having to do with a disconnect between the pixel organization and the stated resolution, which made everything "fold over" multiple times.

Enjoy!

Scott

[Mephesto]

We have a winner! That's the wildest fix I've ever seen.

I've come across that problem before when cropping videos a few pixels and saving to x264vfw lossless AVI. The pixels shift to a diagonal manner in a weird pattern and turns to all kinds of colors, though with a proper decoder it doesn't happen. I've learned to either not use x264vfw or resize to mod8 before saving.

I'm inclined to say the problem is easy to remedy if you know the original resolution to reverse the simple mathematical pixel ordering error but anybody can TALK. I could do nothing close to that kind of fix with the video.

Don't trash your unsolved projects, though. I've kept a really scuzzy DVD for 5 years before I could reasonably clean it up. My takes in 2013 look way better than my first attempts in 2008 when I thought I had the most unmatched tools around. Back then I thought a simple temporal smoother was the cutting edge method in grain removal and that anything else wouldn't work or would ruin the quality too much. If I knew then what I know now... I would have deleted the original footage thinking it couldn't get better.

[2Bdecided]

That is amazing Cornucopia - how do you do it?!

I'm inclined to say the problem is easy to remedy if you know the original resolution to reverse the simple mathematical pixel ordering error but anybody can TALK.

The fact it's been through lossy MPEG-2 while in this chopped-up form means there will be weird artefacts after it's been put back together.

Cheers,
David.