Is it possible to push VHS video beyond its basic clarity/resolution?

[brassplyer]

Video at hand is commercial VHS video captured to DV format. There's nothing really "wrong" with the video other than the limitations of the SD technology it was filmed with and the further hit it takes going to VHS consumer format. The tape was bought new, has had few plays, the player is healthy and I suppose the video looks as good as it's meant to look.

To convert to DVD, I've been playing around with various filters in Virtualdub to see if I can enhance the image. After deinterlacing, I find that I can't really improve it past its original quality. Sharpening to any extent that you can really see adds artifacts. Skin tones start to look artificial, people in the background develop a "milk moustache".

Is there a trick for sow's ear/silk purse treatment short of jillion-dollar ILM-level enhancement of the images with CGI?

[budwzr]

Can you supply a clip?

[brassplyer]

Can you supply a clip?

You mean an untreated clip?

[budwzr]

Yes, from the original capture.10-20 secs is fine. Send the worst example.

[brassplyer]

Yes, from the original capture.10-20 secs is fine. Send the worst example.

Here's two with very different content. 10 secs each - no audio. Captured using a Sony Dig8 cam as the passthrough box. Neither are "worst" - as I say the video is basically solid for what it is, just wondering if it can be truly improved.

What I'm looking to see is if you can improve aspects of the video - sharpness, detail, noise reduction - without losing any of the original quality and not introducing artifacts.

If you only want to play with one, I'd say do Original_DV_video_1 since it seems to be the most resistant to improvement. The kids' faces, details of the clothes etc. are small in proportion to the overall video size which seems to make it hard to improve.

http://www.fileden.com/files/2008/3/11/1809238//Original_DV_video_1.avi


http://www.fileden.com/files/2008/3/11/1809238//Original_DV_video_2.avi

[budwzr]

Brightness/contrast/unsharp mask. Untouched original on the right.



I would go ahead and resize it 4X first. That's what you see here. The whole screen capture including checker pattern is 720X1280. But you don't want to stretch it, you're "re-casting" it, or "super-sampling" it.

After you finish, You can probably crop it to look like an HD aspect, like 720X360. If you render it out at 720p FIRST, then re-render it to SD, it will pop too.

So the end result in SD, would be:

[davexnet]

Looks to me like typical VHS, the limited color bandwidth is a give away.

[vhelp]

After deinterlacing, I find that I can't really improve it past its original quality. Sharpening to any extent that you can really see adds artifacts. Skin tones start to look artificial, people in the background develop a "milk moustache".

.
.

Is there a trick for sow's ear/silk purse treatment short of jillion-dollar ILM-level enhancement of the images with CGI?

yes, raise the bitrate.

when it comes to vhs, you want to keep the perceived noise as detail and incoporate higher bitrate to retain the vhs asthetics. imho, mpeg-2 is not suffecient for this medium. a better compression format would be h264, with a good bitrate. that means larger filesize but worth it if you want to see as close to transparent to the source as possible.

for film content, you ivtc, and for interlace video, you have the option to deinterlace or leave alone.

for interlace, i prefer to bob-deinterlace.

if you're posting (SD) to youtube, it would be better to upsample the video at 720p, this way you have a better chance at obtaining decent quality 360p and 480p since yt-converts anyway, plus giving an HD option of a 720p video. so, you want to upload a very high bitrate source for best yt-convert.

i prefer to leave your video clip alone. altering the color levels tend to contribute less to the overall re-encoded quality.

edit: slight improved deinterlaced.

[brassplyer]

Brightness/contrast/unsharp mask. Untouched original on the right.

I appreciate you taking the time - to be honest you're running into the exact same things I've been encountering. My experiments involved essentially the same kinds of filters you've mentioned. Note the 2nd kid from the left, middle row - see the "milk moustache" & hot spots on his vest?

I'm not versed on all the tech but my sense is there's only so many pixels to work with in those small areas and the tools available aren't fine enough or the nature of the video just doesn't allow them to affect the areas without adding noticeable artifacts.

[lordsmurf]

In terms of getting sharper video, this has to be done in hardware, with a detailer.
By the time you get to software, it's too late to adjust the analog frequencies.

The posted clip has a bunch of color issues, too. That's a separate topic.

[brassplyer]

In terms of getting sharper video, this has to be done in hardware, with a detailer.
By the time you get to software, it's too late to adjust the analog frequencies.

The posted clip has a bunch of color issues, too. That's a separate topic.

Do you feel that with typical consumer VHS there's a lot better image that can be gotten out of it than with a decent consumer VCR?

[robjv1]

You can only do so much.

The ideal is to capture using the best possible equipment, enhance detail if possible, clean up the color and color properties, fix the common VHS errors and clean up the noise commingled with the detail. That last one can go a long way in impoving the aesthetic quality of the footage and really improve the clarity. Sharpening is the same way -- it's all aesthetics.

There are so many places to go wrong though -- subpar equipment, the wrong combination of equipment, improper capturing and encoding techniques, and not having the ability or awarness to correct seemingly unfixable problems. Another big one that often gets overlooked is capturing and applying filters to a video on your computer monitor, but forgetting to check it out on a TV set with overscan or seeing if the colors look too hot/crushed on a CRT.

If you want to push the limits of the format, you do have to invest some time and money and depending on your footage -- possibly a lot of both. There is no CSI enhance magic bullet that is free or cheap, everything is on a case to case basis. NeatVideo is probably the closest to a miracle potion as far as noise removal, but it will cost you at least $50 (but worth every dollar), is quite slow, and can potentially ruin video if you don't learn how to manually configure it. Once you learn it though, it becomes pretty indispensable with certain types of tapes.

You also have to contend with other limitations -- the quality of the original recording equipment and production values will limit how far you can push it. I have SP recordings from lesser productions that are very soft on even my sharpest VCR -- not much you can do with the sharpness at that point.

One thing I've found over and over again is everytime I think I've done something the best it can be, I learn something new, get a new piece of equipment, or find someone who can do it better than me. Then you have the great dilemma -- is it good enough or is it back to the drawing board?

Keep in mind that with a little bit of experience, time and solid equipment (at a relatively reasonable price) you can probably get 80% of the way there. After that, improvements become much more incremental, while price and time investment increase at more like an exponential rate. Getting that extra 10% might cost you another 30 hours of your time for a two hour tape, plus a $1000 of your money if you don't already own the necessary equipment. There's that dilemma again.

Not to be overly dramatic, but I've been working on my parents home movies for a long time, redoing captures, and trying to get things perfect to really impress them. I kept putting off showing the footage to them, even though my mom keeps bugging me about it. Well, last week my mom unexpectedly died -- and now I am really kicking myself for holding them back and missing out on reliving them with her, all in some vain effort to clean up superficial stuff that nobody but me would probably notice.

[robjv1]

Do you feel that with typical consumer VHS there's a lot better image that can be gotten out of it than with a decent consumer VCR?

It depends on the equipment and the tape in my experience.

In terms of sharpness, some tapes are already maxed out on sharpness, even when they don't seem very sharp to begin with. Anything but the slightest high frequency boost brings out the coarseness and the grain. On the other hand, some commercial SP masters seem to come to life with a little boost.

In terms of color properties, maintaining a stable picture, suppressing or correcting common VHS picture and sound issues, and suppressing video noise -- there is huge variation in quality of playback among decks. In other words, you may have a tape with A+ quality footage on it, but if your VCR has a C- transport and electronics, you'll never see it at its full potential. For example, some decks pump up the contrast too much, so right from the start you're losing detail on both ends.

Choosing the right VCR is the most important thing in the whole process -- although once again, there is no magic bullet, you're better off owning a few different decks and running tests on each tape if it is crucially important to you to make the best possible recording.

[jagabo]

milk moustache

VHS needs to be sharpened along the horizontal axis only. Ie, it's blurry along the horizontal axis, not the vertical axis. If you sharpen equally on both axis you generate that type of artifact.

[lordsmurf]

Do you feel that with typical consumer VHS there's a lot better image that can be gotten out of it than with a decent consumer VCR?

Absolutely. I've been saying that for 10+ years now.
It's also the kind of work I do on a daily basis. (That and other video/media work.)

I've never much liked NeatVideo. It's "neat", yes, but I find other methodology yields more life-like quality.
NV has a bad habit of plasticizing your image, even when using mild settings.
But it does have a place, a time, and a use.

[jman98]

brassplyer - While I'm not in any way disputing what lordsmurf says, you do need to understand that he has access to equipment that most people don't have and the experience AND time to do complex restoration work. There are limits to how much time I'm personally willing to spend on this kind of thing and how much money I'll spend on equipment to restore what is actually a pretty crappy storage media (VHS tape) and I just settle for "good enough" rather than perfection. We've got one guy who about every month or so reactivates a thread he started over a year ago about capturing from his VHS tapes and the last time he posted, he had accomplished absolutely nothing in more than 1 year because every time he started to capture, he got pissy about the quality and started all over again. So being obsessive to the point that NOTHING satisfies you is clearly not the way to go either. You just have to strike a balance between what you're willing to spend in time and money to do this kind of thing until you get results you can live with. And trying to get "lordsmurf results" is probably not a realistic outcome for the vast majority of us.

[brassplyer]

Do you feel that with typical consumer VHS there's a lot better image that can be gotten out of it than with a decent consumer VCR?

Absolutely. I've been saying that for 10+ years now.
It's also the kind of work I do on a daily basis. (That and other video/media work.)

Are there any examples of the kind of thing you've done you can post or point to online? Curious to see a before/after comparison.

[budwzr]

Was that Carol Burnett clip recorded from a VCR, or camera?

[brassplyer]

Was that Carol Burnett clip recorded from a VCR, or camera?

That isn't Carol Burnett, it's Carrot Top.

[budwzr]

milk moustache

VHS needs to be sharpened along the horizontal axis only. Ie, it's blurry along the horizontal axis, not the vertical axis. If you sharpen equally on both axis you generate that type of artifact.

Hmmm....there's a control for that in Vegas, but I'm no expert.

[budwzr]

Was that Carol Burnett clip recorded from a VCR, or camera?

That isn't Carol Burnett, it's Carrot Top.

I thought Carrot Top is a man? Comedian right?

[brassplyer]

Was that Carol Burnett clip recorded from a VCR, or camera?

That isn't Carol Burnett, it's Carrot Top.

I thought Carrot Top is a man? Comedian right?

I thought you were being facetious with your original comment. It's singer Maureen McGovern.

[budwzr]

Was that Carol Burnett clip recorded from a VCR, or camera?

That isn't Carol Burnett, it's Carrot Top.

I thought Carrot Top is a man? Comedian right?

I thought you were being facetious with your original comment. It's singer Maureen McGovern.

No, you had it right. hahaha.

Anyway, the comment offered regarding horizontal sharpening has me intrigued. I'm gonna play with it and see. Others here are more knowledgeable than me, though.

[brassplyer]

Now if someone can just explain why conversion from DV to Huffyuv throws off the audio sync by about 200 ms.

[budwzr]

Just using Unsharp Mask alone is an improvement, but it adds grain. I think you will need to do a multipass solution.

Convert the audio to PCM first, or too.

[jagabo]

The only AviSynth sharpener I know that works in a single dimension is the built in Sharpen(). A sequence like

BilinearResize(416,480)
Sharpen(0.7, 0)
LanczosResize(720,480)

will sharpen in only the horizontal dimension. The image was scaled down to something near VHS's real resolution before sharpening because Sharpen() only sharpens relative to the immediate neighbors. Sharpening at full horizontal resolution would only increase noise. You can play with different scaling algorithms and sizes, and different sharpening strengths.

A smarter sharpener would work better and wouldn't require the scaling but I don't know of an UnsharpMask-like AviSynth filter that works in only one dimension. Maybe someone else knows of one.

I the absence of a smart sharpener that works in only one dimension you can try doubling the height of the frame, running the 2D sharpener, then downscaling again. Something like:

BilinearResize(720,960) #double height
UnsharpHQ(3.0)
BicubicResize(720,480) #restore original height

That will prevent too much sharpening along the vertical dimension. Of course the video has to be progressive for that script.

[mammo1789]

What about MSU smart sharpen http://www.compression.ru/video/smart_sharpen/index_en.html it has many settings that can benefit VHS mud I use it sometimes i saw this presentation http://www.youtube.com/watch?v=4xlrMG0hjCI&feature=gv and I think it looks great what do you guys think

[brassplyer]

Cropping 4:3 video to a 16:9 dimension is a bad idea IMO. Not only do you end up expanding a less sharp image thereby making the perceived clarity even worse, the shots were framed with the 4:3 AR in mind. You're going to end up chopping off parts of the image that are supposed to be there.

[budwzr]

Fine, then let it play as such. Give it a touchup and screen it.

[sanlyn]

Why deinterlace?
In Video1 darks were crushed during capture. Both videos have problems with red (seriously over saturated). If darks or brights are wiped out during capture, forget about retrieving them later (they won't exist any more). Dot crawl visible, mostly in Video2. I didn't address all the problems (blue background in Video: serious banding and blocks). Mainly used QTGMC ("fast"), ColorYUV, some effort at reducing chroma bleed and shift. VirtualDub: ColorMilland and gradation curves. NeatVideo on Video1 (about 30% power), no NV on Video2.
Had a hard time with cyan chroma bleed and shift in Video2. Finally just ran out of time.
I don;'t know what the hell others are trying to do with color in these 2 clips.


[Attachment 14398 - Click to enlarge]


[Attachment 14399 - Click to enlarge]