Restoring vs. distorting, audio and video?

[lordsmurf]

I spent a few minutes on this for you, seeing what might be possible.
See the attached Matrox MPEG-2 (10-bit I-frame only in AVI wrapper).

Here's what I did:

Step 1: Deinterlace with Avisynth, output to AVS file for further VirtualDub processing. Attached is my installed Avisynth plugins folder from "C:\Program Files\AviSynth 2.5\plugins" (which includes the needed plugins). You'll want to install Avisynth on your own, then "steal" the added contents from my folder into your own. And then use this script:

ffvideosource("d:\sample2.mpg")
yadifmod(order=1, mode=0, edeint=NNEDI2(field=1))

or

avisource("d:\sample2.avi")
ConvertToYUY2(interlaced=true)
yadifmod(order=1, mode=0, edeint=NNEDI2(field=1))

The first script is for MPEG-2 input, the second for HuffYUV input.

Note: VirtualDub deinterlacing is inferior to what Avisynth can do, hence this step. Everything else is fine in Vdub.

Step 2: First pass with VirtualDub, to corral some of that harsh color, and provide most of the noise reduction (NR). Median filter is what knocks out most/all of the dropouts ("comets" as some have called it). Output to lossless HuffYUV. FYI, YES those filters are in a specific order -- it is NOT random.

Step 3: Second pass with VirtualDub, to refine some of the color quality in ColorMill and add another NR pass.

Attached are my VCF settings files for both Vdub passes.

This is just a quick proof-of-concept type correction. Don't mistake this as the best I can do. It would take a good bit of tweaking with color and NR filters, to get that oh-so-perfect balance of removal vs loss, and good vs bad coloration.

My choice here is to reduce, but not entirely remove, the color. Remember that a whole channel is missing (1 of the 3 pieces, the YCb with no Cr), so you're looking to reduce that obvious problem. It cannot be corrected, but it can be minimized for your viewing pleasure. Nobody wants to watch a severely off-color video. Think of it as a color-tinted silent movie (but with sound, in your cases) similar to the famous Metropolis flick, which used gold and other additives to give pigments to otherwise B&W scenes.

Because I know you err on the side of "leave it alone", my approach to NR was very light-handed.

[jmac698]

I don't see any attachments...
Ok, that took a while.
Nice job! I admittedly suck at color grading, no practice.

[lordsmurf]

Sometimes posts take 3-4 steps to fully get up. This was one.
You saw it after my first save, before it was completely uploaded and finished for viewing.

Thanks.

[magillagorilla]

Great idea, B&W, it hadn't occured to me. Thanks!

OK please bare with me.
Here is another sample of some of the old VHS. This is in better shape, it's meduim quality in my set. Most of my clips look better some look like this and very few are really bad like the previous samples.

[jmac698]

No B&W! I fixed it.

[magillagorilla]

sample4
EDIT: sorry I posted more junk before I realized you guys were still working on the last samples. Let me look at your results on sample2 LS. Thanks!

OK, my restoration project involves 3 different types of media VHS. VHS-C and MiniDV.

sampel5
This is a sample which is typical of my VHS-C recordings. The fat guy is me 7 years ago. I could use a bit of cleaning as could most of my VHS-C. Should I just denoise a bit.

Here is a MiniDV sample. I had the camera set to crop for 16:9 for some reason. The interlace lines seem harsh for some reason.

Should I even mess with my miniDV footage? Most of it is really clean. I guess I'm asking if there is anything to gain from filtering?

LS, Thanks for the example. I'm going to mess with those settings tonight.

Just curious, what is the advantage of deinterlacing? I'd prefer to keep my temporal resolution rather than smudging it, though I have not tried Avisynth's deinterlace. I know that VLC and my TV will handel interlace.

[admin]

Quote:

Originally Posted by magillagorilla

Just curious, what is the advantage of deinterlacing?

It is required in order to run several of those NR filters -- especially the median filter.

Quote:

I'd prefer to keep my temporal resolution rather than smudging it, though I have not tried Avisynth's deinterlace.

You're not really smudging anything. It's not like VHS tapes had a lot of temporal or in-frame resolution anyway. Not that I'd want to reduce it further, ala VCD or drop-field deinterlacers, but this one is non-destructive.

The NNEDI2+Yadifmod combination works very well, mostly working to rebuild combed areas as progressive, but without simply throwing away data like most other methods will do.

A number of us have been working towards a reliable method to fully or near-fully remove magnetic tape drop-outs, and this is probably the first example I've seen where one of the proposed/theorized methods has succeeded in near-complete removal of this awful analog noise. Some of these filters and filter combos did not exist even 5 years ago. I have some client project files here that are the basis for "worst case" tests on dropouts, and need to try this script/filter combo on it.

So you're getting VERY current advice.

[Steve(MS)]

Well I must admit so far my adventures in noise reduction using vdub has been less than spectacular.
When correcting color, etc it has been somewhat better provided the original wasn't too bad.
No doubt most of my lack of success has been due to my ignorance of video computer software.
Further since I am on dialup, I haven't been able to look at the
images on this thread but I would like to add my novice opinion
to the dialog from the beginning,
First get the best possible VCR that you can, like some
of the JVCs recommended on this site.
Sometimes it helps to have more than one model so if one tape doesn't play on one, it may do better on another.
Next and this is important, get a hardware proc amp and do
your color,brightness, contrast adjustments before they go
to be recorded.
And I decided to bypass the computer capture and got some
really great JVC recorders.
The only video I run through the computer are the recordings
that need some extra little bit of work.
Once again for me the hardware approach for color, brightness, contrast along with several of those great JVC VCRs work best.
Unless you know what you are doing with the computer approach, you can spend a ton of time, which is fine for those so inclined to do so.
BTW, Admin I hope one of these days you can put up a guide
for basic NR and so forth using Vdub filters,maybe a set of your favorite ones, for us bunch that are computer and video computer restoration challenged.

[magillagorilla]

Steve,
What do you mean by JVC recorder? If you are talking about DVD recorders, they won't suit my needs. I feel that transferring the video to DVD just creates more work for me in the future for archiving. Plus, as individual computer files, I can create an unlimited combination of productions. DVD is rigid and I don't trust DVDr media. Hard drives are cheap and can be easily duplicated.

I agree a video processor would be ideal but it's not really in the budget. My playback decks are a Panasonic AG-1980P and AG1960. The AG-1980P does a great job.

Yes the software learning curve is high but if you can't spend money, you have to spend time.

[Steve(MS)]

No problem with how you want to approach it.
However I have seen firsthand where a tape will look not so good on one
model vcr and on another will look better, sometimes much better.
And it seems to me that proc amps do a better job with adjusting
color errors, etc.
I am using the k.i.s.s. method and thought I would mention it as an
alternative.

[magillagorilla]

I agree with your method. I just don't have the budget for multiple VCRs and processing equipment. Unfortunately in this case, k.i.s.s. equals expensive.

LS,
I finally got around to really playing with the example files you posted. You are correct, the AVISynth deinterlace looks really good. Is it recomended to deinterlace all of my VHS video. I mean, it looks pretty good to me (especially on a progressive display), any reason I shouldn't?

Your VHS YCB noCr Test Pass1.vcf settings for VDUB produce a file that crashes my media players (VLC, WMP, and VDUB). It's really weird. When I run it on a longer clip the file it outputs is un-playable. VDUB steps through the clip and applies the filters without errors but as soon as I try to open the resulting AVI, in anything, it crashes. The file gets about two frames in where 2/3 of the top of the frame is bright colored rainbow snow (digital noise) no picture at all. The bottom 1/3 looks like my video. Then all players dump memory errors. (sorry dont have those right now).

I thought my machine was freaking out or I had bad RAM or something so I ran another pass (on the same clip) using a different filter set and it plays fine. Reboot..... same problem.

I didn't have time to troubleshoot the filter set. I started to remove filters 1 at a time to see of there was an offending filter but I ran out of time to get through the all. (it was 2am and way time for bed).

My first suspest was the resize filter. Could it be resizing the video so that the AVI wrapper says the video is 720x480 but the actual size of the data is different, causing players to crash? Or maybe some kind of "out of range" setting for the AVI format? I dunno, any thoughts?

[kpmedia]

Quote:

Is it recomended to deinterlace all of my VHS video

No. Only deinterlace when required. Although the method being used here is pretty good, it's not perfect. None of the available methods are. Someday, far off, maybe we'll see one happen. It does get better every year. The only way to use a better deinterlacer later is to keep interlaced footage as-is for now.

Unless, of course, it needs to be done in order to allow for restoration, such as the median filter usage that was done here in these samples. This is an exception to the "never deinterlace" rules.

Quote:

Your VHS YCB noCr Test Pass1.vcf settings for VDUB produce a file that crashes my media players (VLC, WMP, and VDUB). It's really weird. When I run it on a longer clip the file it outputs is un-playable. VDUB steps through the clip and applies the filters without errors but as soon as I try to open the resulting AVI, in anything, it crashes. The file gets about two frames in where 2/3 of the top of the frame is bright colored rainbow snow (digital noise) no picture at all. The bottom 1/3 looks like my video. Then all players dump memory errors. (sorry dont have those right now)

Players generally don't like to play lossless video, like HuffYUV. This format was intended for editors, not players. If I try to play the "MT" (multi-threaded) variant of HuffYUV, it will crash most of my players, too. PowerDVD (or maybe it's WinDVD?) tend to work pretty well with HuffYUV codec -- go figure. A DVD player software works better than a general-use media player for that format, in my experience.

Quote:

Could it be resizing the video so that the AVI wrapper says the video is 720x480 but the actual size of the data is different, causing players to crash?

The AVI wrapper doesn't contain resolution data.

[magillagorilla]

Quote:

Players generally don't like to play lossless video, like HuffYUV

The codec is not the problem. The original HUFFYUV file plays fine. When I process the same clip using a different filter set the final HUFFYUV file also plays fine. The filter set LS posted produces an AVI file which crashes even VDUB. ....weird

Anyhow, I will have to deinterlace a lot of my footage because it's in bad shape. I am keeping the original HUFFYUV caps archived though to future proof for restoration techniques.