Latest film restoration tricks

johnmeyer wrote on 1/27/2012, 2:31 PM

The following short before/after clip shows some new tricks I've developed for film restoration. These same tricks apply to video.

This 8mm film was taken in 1954, and the camera (not the projector) malfunctioned and didn't bring the film to rest before opening the shutter. As a result, not only were the frames jumping up and down, but many of those which jumped were blurred. I went through by hand (thousands of frames) and marked each jump frame. I then developed software which deleted the jump frame and then used motion estimation software (similar to Twixtor which is used for synthesizing slow motion) to create a brand new synthesized frame which I then inserted in place of the damaged frame. So, on the right-side "after" version of the clip, about 50% of those frames are computer-generated.

I have used this same technique -- although thankfully on a much smaller number of frames -- to remove frames where a photographer's flash ruined the frame, or where there was a dropout or, on older analog video, where there was a glitch or dropout.

Back to post

Comments

4110 wrote on 1/29/2012, 12:00 PM

My parents bought their 8mm camera in 1948. Some of their film jumped and some didn't. That was true 60 years ago when we looked at it with their original projector. It is still true now with the digitized film. Based on that history I am pretty sure it was a camera issue and not a projector issue. I have been working on the digitized film with Vegas. Mercalli 2 corrects some, but not all of the jitter problem. In fact, there are some sections that Mercalli seems to make worse.

Your efforts are impressive but also intimidating and I expect it will take me some time to get up the learning curve. I have my fingers crossed.

Thanks again,

David

johnmeyer wrote on 1/29/2012, 2:45 PM

My parents bought their 8mm camera in 1948. Some of their film jumped and some didn't. That was true 60 years ago when we looked at it with their original projector. It is still true now with the digitized film. Thanks for posting that. It gives me more confidence that I am not going nuts. I say this because I actually re-mounted the bad film and looked at it back in the projector, one frame at a time, to confirm that the problem was really on the film. It was. However, I still didn't want to believe my own eyes because I didn't see how the camera could have malfunctioned in that way. Apparently, from what you say, this was not a unique or isolated problem

I have been working on the digitized film with Vegas. Mercalli 2 corrects some, but not all of the jitter problem. Yes, I was able to use that on about 2/3 of the footage. The key is this: if the frames move location, but are not blurred, then Mercalli will work. However, you have to get it set up correctly in order for it to correct the rather violent vertical motion. Deshaker can sometimes do a better job. You just have to try both.

Your efforts are impressive but also intimidating and I expect it will take me some time to get up the learning curve. I have my fingers crossed.The first part -- using the Vegas script -- should be easy. Just save the code, using Notepad, and use the extension ".js". Assign the script to a keyboard shortcut and then you can do a "first-order" correction by duplicating frames. If the objects in the film itself are not moving much, you can sometimes get reasonable results just by duplicating.

The next part, using AVISynth, is also quite simple once you install AVISynth and Debugmode frameserver. You may have to be running a 32-bit O/S to use these two things. (Actually, you can render to an intermediate rather than use the frameserver). Once you have AVISynth installed, you just point to the frameserver signpost file, or to the intermediate file, and then open the script in VirtualDub and render out. The only really difficult thing is finding all the jumps and replacing them with a duplicate. After the first 1,000, it gets pretty tedious.

larry-peter wrote on 1/29/2012, 3:16 PM

I'm with musicvid. This is astounding. You should find a way to brand and sell this approach.

johnmeyer wrote on 1/29/2012, 3:45 PM

You should find a way to brand and sell this approach.Thanks for that thought. Unfortunately, a lot of what I do fits into the category of "technology without a market." I have over 100 AVISynth scripts that I've developed, some of which do equally remarkable things to film and video. However, there are very few commercial entities who need this sort of thing (most film and video done by companies are using film or video being shot current day, not resurrecting old stuff).

A good example of this is the film to video technique I figured out a few years ago. There are actually two companies who figured out how to do this a few years before I did. They both tried marketing to companies who own the rights to old Kinescope versions of television shows. Using this technology, you can make it look like the TV show was recorded on videotape instead of film. My technique does the same thing. The problem is -- and this is even more true of doing the same thing for amateur film -- is that no one wants to pay what it actually costs, because the number of people viewing this old stuff is a fraction of the number of people who are willing to watch new stuff. With amateur stuff, where the viewing audience is usually just the member of one family, the cost is completely out of the question (unless you are lucky enough to get the job restoring the 1960 World Series Kinescop footage found in the attic of Bing Crosby's house a few years ago).