VHS frame comparison of sharpness?

[guyburns]

I'm asking for comments on the four stills attached, from two points (c and f) on the same tape, captured by two different cards, but everything else the same. No post-capture alterations. The comparison is between 12c and 38c; and then 12f and 38f. I had them layered in a tiff file, making it very easy to compare, but tiff files aren't allowed as attachments.

What do the differences reveal? Video 38 is sharper than the other. Has it undergone sharpening as part of the capture, or does it reflect what was on the tape? Maybe Video 12 has been subject to softening as part of noise reduction.

I'm just wondering whether it is possible to infer from these two captures, which one is likely to be the more accurate.

[lordsmurf]

I don't play guessing games, I need to know what I'm looking at. (The entire idea of "blind" comparison is actually scientifically very faulty. It allows one-way bias, with the 2nd party unable to call BS.)

What I see in that CIC image is a wash:
- cleaner + darker + AR differs
- noisy + possible illegal luma + dropout detected + AR differs

That Universal image is more stark:
- darker + seems more accurate
- excessive chroma bleeding + lighter + halo/ringing + yellow chroma hot

Neither is accurate. With proc amp, one probably could be.

So what are we looking at?

If you want advice, get advice. Don't try to be coy, cute, and clever about it.

Attach TIFF or PSD inside a ZIP or RAR.

[timtape]

To my eye, in both cases the one with more wanted picture detail also has more unwanted noise. So long as the full signal has been captured, what will determine how we cut the trade off will depend on what we or the target audience value more, either more detail but unfortunately more noise and artefacts, or less noise and artefacts but unfortunately less picture detail.

It's no surprise VCR and TV manufacturers have often provided an easy to reach front panel "sharpness" control, with a middle detent position, allowing each viewer to decide the trade off themselves. The difference between your two sets of comparisons seems mostly as if each had the viewer's "sharpness control" at slightly different settings. That may just be how each card was set up, slightly different in its emphasis of the fine detail.

For me in this context it's important to use the hardware which results in, in engineering terms, the maximum signal to noise ratio: maximum picture information off tape, but with minimum added noise from the hardware itself. The same applies to the audio.

So changing capture cards may do little more than mildly alter the "sharpness" setting. Unfortunately it doesnt improve the signal to noise ratio, which is what we really want. The exception is when the wanted picture detail is less fine grained than the fine grained noise. In this case, we can sometimes reduce the more fine grained noise without compromising the less fine grained picture detail.

[guyburns]

Keep the comments coming, everyone. I'm trying to learn how to interpret these captures, from people who know more about it than I do. But, I'd rather not have biases creep in because of perceived quality of the equipment.

Quote:

Originally Posted by aramkolt

The true test I think would be to post unlabeled samples of different popular and some less popular chains all using the same starting VCR/tape and have people determine for themselves what is visually the best to them so that any bias on equipment choice is removed.

Quote:

Originally Posted by timtape

The exception is when the wanted picture detail is less fine grained than the fine grained noise. In this case, we can sometimes reduce the more fine grained noise without compromising the less fine grained picture detail.

You might have to run that by me again. Are you saying: if the image has a certain of blurriness, covered by tiny bits of noise, then you can use filters to remove some of the noise without really degrading the image?

[lordsmurf]

Quote:

Originally Posted by guyburns

But, I'd rather not have biases creep in because of perceived quality of the equipment.

Again, this is not correct.

You may even think you're unbiased, but you're not. The simple act of blinding often creates (and then shields) bias.

https://en.wikipedia.org/wiki/Blinded_experiment

This experiment contains too many variables (such as the unknown source tape in use), and not enough constants. Still images are also not representative of video.

Furthermore, most gear has certain knowns, which can explain what you/we see, and why. It will also unblind the quality of the source, which is a variable, essentially doubly blinding us.

So you're getting random feedback here. It's fairly worthless without context. You may think it's useful, but it's just not.

[timtape]

Quote:

Originally Posted by guyburns

...You might have to run that by me again. Are you saying: if the image has a certain of blurriness, covered by tiny bits of noise, then you can use filters to remove some of the noise without really degrading the image?

Why not test it out yourself? With the same image pairs you uploaded, gradually soften (blur) the sharper one and compare. Let the facts/results guide you.

[guyburns]

After I decide which half dozen captures are worthy of further consideration, that's when I'll be adjusting them in Premiere, as part of this testing process, to find out if the poorer ones can be made to look as good as the best one. That will mean adjusting colour, contrast and sharpness. I doubt I'll be blurring a capture to improve it's looks. Sharpness is the primary objective. Audiences will react negatively to a blurred image, but they are quite forgiving of colours inaccuracies. Thousands of slide scans that have ended up on my Blu-rays have taught me that. Sometimes a friend may wander in while I'm starting work on an image. "Oh, that looks really good". And it's the raw scan before correction. And it does look good, but I know it can be significantly improved. They have no reference, and neither have I. So what I do after I've finished working on an image, is to run an action in Photoshop which I've called Slides (Curves 3), which adds an additional colour-correction layer to the two already present. It uses one of the automatic colour corrections offered by PS. Usually the auto correction is poorer, but every now and then, PS shows me how my corrections could be improved; it will have generated a better looking image. I now have another opinion, albeit an automated opinion, that shows me what to aim for. I never use the PS version. Curves 3 is turned off, and I go back to Curves 2 and try again myself; learning. How come this dumb PS can do a better job with this slide than I did?

I'm doing a similar thing with this thread; asking others what they see and what I may have missed; judging an image by what's on screen, not how it was obtained; trying to work out – Is that sharper image real, or has it been sharpened during capture? Or vice versa: does that not-as-sharp image mean the capture process used is inferior.

I'm trying to find a capture process that reliably gives a high-quality image, an image that has undergone the least processing; that simply captures what's on tape. I can sharpen and colour correct as much as I want in Premiere, but I want the capture to be as raw as possible. And I want to learn, with the help of others more experienced, what to look for when judging images from tape.

[aramkolt]

Oh, believe me, I get the frustration. It is hard to know what the VHS source really looks like and to tell what each card did to it. Ideally you'd want to capture it as it actually is so you can decide how much to alter it later.

Best case for source material would be to use a DVD source which you can actually go back and look at to know what the starting source looks like as it's already digital. With the DVD, you also won't get variations from playback to playback as tape can wear out or alignment can slightly vary from one test to the next.

I still think capturing test patterns from something like the "Video Essentials" DVD would give you a pretty good idea of what each card is actually doing. Those can be had for like $8 shipped on ebay. Video can look different as well, so you'd probably want to capture some test patterns, and some actual video. Some here suggest cartoons as there's lots of high contrast areas that you can look for artifacts around.

At least with that, others can post how their capture chain does using an identical starting source. Everyone's DVD player will be slightly different, but odds are that DVD output is going to be fairly similar across if playing the same DVD.

Most here would also prefer short video clips over screenshots, or at least in addition to screenshots as they can give more info about levels clipping etc.

[BW37]

@LS

While I agree that some of the opinions given here may be little more than random feedback, I've got to disagree as your interpretation of that Wiki article. It states pretty clearly that blinding is useful and often necessary because of the bias of all involved. Not the other way around as you seem to be saying.

We are ALL susceptible (if not down right beholden) to confirmation bias in spite of our protestations to the contrary. In every undertaking I've ever done, work or hobby, I've seen it and been guilty of it. It's simply "human nature". One of the most useful ways to minimize this problem in experiments is to "blind" at least the observer.

When I was working as an engineer in a manufacturing environment our group was called on to perform "designed experiments" to evaluate the combined effects of various product and process changes on the performance of a (often new) product. Invariably, the lab (product design) personnel would interpret the raw data results to confirm their assumptions even though the actual statistical analysis of the overall design said otherwise. Of course the lab was higher on the totem pole than us lowly "plant people" so we all to often had to modify products and processes after the fact because of their insistence on overriding the actual results with their pet theories. As the saying goes, "that's nice in theory but..."

BW

[latreche34]

In the entire conversation no mention to what workflows are involved in the two captures, If you keep it blind, fine, but we have to know what we are blindly voting for. It's like complaining to a mechanic about a mechanical fault but not telling him anything about when it happens, at what speed, in cold or hot engine condition...

[guyburns]

Thanks for all the replies.

Quote:

Originally Posted by latreche34

If you keep it blind, fine, but we have to know what we are blindly voting for.

I'm not actually asking for people to vote, I'm asking "What do the differences reveal?

Re videos: I have my doubts about judging captures by watching videos. There's just too much happening too quickly. That's why most people don't see the difference between DVD and Blu-ray movies; they're concentrating on the action, not the technical aspects. I know that from experience. About a dozen times a year I invite people around for a movie night, some classic, pre 1970s, that has received glowing ratings from Blu-ray.com or DVD Beaver for the quality of the transfer. I always get up and introduce the movie, and usually read something from the technical review. The friends are well aware of my interest in high-quality images. But a few times I've wanted to show a movie that wasn't available on Blu-ray, so I've put on a DVD. No one's made negative comments -- and I do ask afterwards if they've noticed anything different.

This is an example of what I'll be reading out before our opening movie for 2024, my most recent purchase:

It is unknown whether additional restoration was performed for this Blu-ray, but the results on Paramount/Warner's 1080p, AVC-encoded Blu-ray are simply astonishing. Details, densities, black levels, textures and colors are all revelatory. I literally felt that I was seeing a film I had never seen before. It's not just in the obvious scenes, such as the mountain vistas around the Starrett farm, that this Blu-ray's image reveals its treasures. It's also in the subtler shadows of the day-for-night sequences (what Stevens called the "Rembrandt lighting"), such as the encounter between Ryker and Joe Starrett after the Fourth of July celebration, where the shadow detail is just sufficient and the shades of black and blue layer over each other in just the right proportions to create the sense of depth and danger that Stevens and Griggs intended.

The film's grain pattern is natural, fine and undisturbed by digital manipulation. The impressive bitrate of 29.69 Mbps helps explain the lack of any compression issues.

I want to be able to confidently analyse VHS captures in that sort of detail. I can spend ten minutes peering at two stills, jumping back and forth, working out what the differences are and where they come from. When I finally rank the captures the Matthew has done for me, and added some of my own, the final test will be in our lounge room, with several people watching short excerpts from maybe four captures, and being asked to rate them. I reckon I'll have to run through them 5-6 times as people get to know the videos, and how to judge them, refining their ratings as we go along.

If the video results are similar to the stills results, I'll know that I've got the answer.

[mrmuy97]

Quote:

Originally Posted by guyburns

I'm just wondering whether it is possible to infer from these two captures, which one is likely to be the more accurate.

Video clips are badly needed here.
But I only see it as "which one produces the better quality captures."
As for "What is the better quality capture?" timtape put it well:

Quote:

Originally Posted by timtape

To my eye, in both cases the one with more wanted picture detail also has more unwanted noise. So long as the full signal has been captured, what will determine how we cut the trade-off will depend on what we or the target audience value more, either more detail but unfortunately more noise and artifacts, or less noise and artifacts but unfortunately less picture detail.

This can vary per capture and per client. The 12s will be "better" to some. The 38s to others.
You can show those 4 comparison stills to 100 people and 47 may say the sharper ones are better while 53 say the softer ones are better.

Quote:

Originally Posted by timtape

For me in this context it's important to use the hardware which results in, in engineering terms, the maximum signal to noise ratio: maximum picture information off tape, but with minimum added noise from the hardware itself. The same applies to the audio.

This is a key, major point on this forum.
And brings us neatly to:

Quote:

Originally Posted by lordsmurf

most gear has certain knowns, which can explain what you/we see, and why

Another key, major point on this forum.

Quote:

Originally Posted by guyburns

I have my doubts about judging captures by watching videos.

You totally lost me there.

Quote:

Originally Posted by guyburns

There's just too much happening too quickly. That's why most people don't see the difference between DVD and Blu-ray movies; they're concentrating on the action, not the technical aspects.

Isn't this thread specifically about the technical aspects? On the DigitalFAQ "Capture, Record, Transfer" forum??

Quote:

Originally Posted by guyburns

the final test will be in our lounge room, with several people watching short excerpts from maybe four captures, and being asked to rate them. I reckon I'll have to run through them 5-6 times as people get to know the videos, and how to judge them, refining their ratings as we go along.

If the video results are similar to the stills results, I'll know that I've got the answer.

[lordsmurf]

Quote:

Originally Posted by aramkolt

Oh, believe me, I get the frustration. It is hard to know what the VHS source really looks like and to tell what each card did to it.

This gets back to my "known knowns" comment. This conversation has attempted to remove known knowns, replacing it with unknown unknowns. That's just bad science, it you're trying to understand how, why, etc. All you get back is random noise.

For example, one image is darker, the other lighter. Which is correct? Context will give hints, or outright answers. Without context, who knows?

Quote:

Originally Posted by BW37

@LS
I've got to disagree as your interpretation of that Wiki article.

Well, I wasn't overly relying on the Wikipedia entry for specifics, it was more about the general fallacy of blinding, as mentioned in part of it. Blinding mostly functions in areas of pure opinion (taste tests) where there is no wrong answer, or where unknowns exists (medicine). When you want to analyze something that has known facts, where you attempt to blind the facts, that's where the testing bias can enter.

In this thread, no meaningful feedback can be given, and the OP is seemingly taking it out of context to "prove" something -- and we don't know what that even is (ie, enter the bias).

Part of this is also me being ahead of the conversation, because I've seen this play out for decades in the realm of video. Both as a hobbyist, and as a pro (working at a studio). 99% of the time, it ends in a "WTF?" conclusion, because the whole methodoly was bogus BS.

Quote:

Originally Posted by latreche34

It's like complaining to a mechanic about a mechanical fault but not telling him anything

Yep.

I'm not even sure what we're commenting on here.
- the source tape?
- the source before the tape?
- the capture card?
- what, exactly?

Quote:

Originally Posted by guyburns

I'm asking "What do the differences reveal?

That's the problem here. Nothing. I can tell you A is darker, B is lighter -- but it mean literally nothing. This is where analog video gets squirrelly to test. You must have test patterns for some things, and cannot have test patterns for others. In this exact thread we need more facts. And we need clips, not stills. This is video, not photography.

Quote:

That's why most people don't see the difference between DVD and Blu-ray movies; they're concentrating on the action, not the technical aspects.

Yes and no.

How old are you? There are people that cannot remember if certain movies were B&W or color. There are B&W movies out there -- not color, never colorized -- that people swear they remember seeing the red blood in scary scenes. So forget about DVD vs. BD, it can be a stark as color vs. B&W!

Video cannot have major defects, or distractions, and content must be good. It's really fringes that insist on 4K, or must win an Academy Award (or whatever award). That's always been my schtick, quality without problems/distractions. We simply needs to be enjoyable, and quality must allow it. When it comes to videotapes, TBCs correct major flaws, and that's most of the technical quality battle for analog tape conversion.