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

[brassplyer]

Why deinterlace? I thought DVD was interlaced or progressive w/3:2 pulldown, and YV12.
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.

I also appreciate you taking the time. Your examples are done with NeatVideo?

The problem with posting stills from video is it doesn't tell you what the moving image looks like. The images you've posted look similar to what the warpsharp filter does in Vdub. Like previous examples, it looks different than the original, not sure it looks better. Can you make the processed video available for d/l?

As far as color issues, hard to make an assessment as to what they should look like when you don't have a dependable reference of what they actually looked like in person. I gather VHS tends to exaggerate reds. I don't have a big problem with the overall color on the video, the biggest issues I see are noise, contrast and lack of sharpness.

I wonder if anywhere there's an example of best possible quality VHS played back on boutique gear.

[jagabo]

I wonder if anywhere there's an example of best possible quality VHS played back on boutique gear.

Of course, the best starts with a line TBC.

[sanlyn]

The images you've posted look similar to what the warpsharp filter does in Vdub. Like previous examples, it looks different than the original, not sure it looks better.

If you can't see bleeding colors. can't tell when shadow detail is crushed, or can't differentiate between white and red, it would be a waste of your time.

NeatVideo was a finishing touch at very low power5 on Video. Not used on Video2. Most of the cleanup was done with QTGMC and aWarpSharp.

[budwzr]

Whatever changes anyone made would have to be keyframed, but the stills give an idea of what to expect. The tools for editing stills and video are the same.

So it's quite reasonable to post a still here.

[sanlyn]

For purposes of comparing color balance and stuff like chroma shift, stills would likely do the trick. But a moving image would better demonstrate rips in frames, other damage, and how much noise is in those clips. Will try to work up a demo a little later.

[brassplyer]

The images you've posted look similar to what the warpsharp filter does in Vdub. Like previous examples, it looks different than the original, not sure it looks better.

If you can't see bleeding colors. can't tell when shadow detail is crushed, or can't differentiate between white and red, it would be a waste of

There may be some color bleeding, I don't know that washing them out looks better. I'll bet a buck the reds are supposed to be more vivid than what your example shows.

I'm including some screen caps off Vdub from the original - the only processing that's been done is a deinterlacing filter. No adjustment on the color or contrast. The reds don't look all that bad to me. The poinsettias look about right - there's very definite distinction between the tone of the roses and the poinsettias - and where they're more brightly lit in some spots than others. Same with the kids' clothes - I see a lot of subtle differences in the red tones. I'd say the wood tones of the viola and bass look pretty spot on. There are also some shots of cup mutes that look like mine.

[budwzr]

So that's you, the trumpet player?

[brassplyer]

Why deinterlace?

For purposes of online viewing or potentially for prepping for playback on an HDTV.

[brassplyer]

So that's you, the trumpet player?

Umm, uh, yeah! That's me!

[jagabo]

I'm pretty sure that Doc Severinsen -- Johnny Carson's bandleader.

[budwzr]

Cool. I remember Doc. So are you still friendly with Ed and Johnny? What a small world.

[brassplyer]

Cool. I remember Doc. So are you still friendly with Ed and Johnny? What a small world.

Maybe at a seance...

[budwzr]

Why deinterlace?

For purposes of online viewing or potentially for prepping for playback on an HDTV.

To upsize to HD is no problem. To look like HD? Not doable.

[sanlyn]

The images you've posted look similar to what the warpsharp filter does in Vdub. Like previous examples, it looks different than the original, not sure it looks better.

If you can't see bleeding colors. can't tell when shadow detail is crushed, or can't differentiate between white and red, it would be a waste of

There may be some color bleeding, I don't know that washing them out looks better. I'll bet a buck the reds are supposed to be more vivid than what your example shows.

I'm including some screen caps off Vdub from the original - the only processing that's been done is a deinterlacing filter. No adjustment on the color or contrast. The reds don't look all that bad to me. The poinsettias look about right - there's very definite distinction between the tone of the roses and the poinsettias - and where they're more brightly lit in some spots than others. Same with the kids' clothes - I see a lot of subtle differences in the red tones. I'd say the wood tones of the viola and bass look pretty spot on. There are also some shots of cup mutes that look like mine.

Like I said, you'd be wasting your time.

[brassplyer]

Like I said, you'd be wasting your time.

If I interpret what you're saying correctly, you're insisting I have inferior sensibilities. ~shrug~

Are you of the opinion that the screen shot of the kids you put up looks natural, that the color is an accurate representation?

As I say, I'd be curious to see what the actual video looks like.

[sanlyn]

I wouldn't say that's a fair interpretation. People make their videos look the way they want. If you were to say a clip looks too blue or too green, and ask for ways to correct it, you'd get several suggestions. The clip with the kids has oversaturated reds. I look at it and say whites aren't white, grays aren't gray, blacks aren't clean, detail was crushed and highlights were clipped during capture. The blue background has coarse grain and noisy gradations. Not an easy clip to fix. Tape aging affects colors in odd ways (in this case, green is fading faster than the other colors). Many tapes defy "accurate" color and levels correction. Video2 has bleed and chroma shift -- most people don't worry about it, they just sharpen. V2 has coarse shadow noise, something I've seen in many captures here. I've tried for some time to figure out which VCR's are doing this (it doesn't appear in my captures, so I have no idea where this coarseness comes from). There are edge and motion problems, low-level motion noise (you can expect that with VHS). There are annoying interlace effects in both clips that I don't know how to fix. I keep looking for answers to that, too, other than to just de-interlace.

Video2 demonstrates a typical VHS difficulty with color correction: there are three different camera shots, each with differences in color balance. Mixed lighting in the scene (colored lights) is one problem. I tried correcting for flesh tones, which is touchy because the singer's skin tones vary among all three shots. I started to address your earlier comment that one wouldn't know how a scene is supposed to look because one "wasn't there" when the scene was shot. That's true. What colorists would do is to first set correct levels and check for crushed darks and blown-out highlights. Then they would set the white point, the black point, and the gray point. You do this by eye and with histograms and pixel samplers. Sometimes there are no proper objects to sample (a singer lighted by colored floods and spots, for example), so you just guess.

Someone mentioned using a line tbc of some sort. I don't see much in terms of line timing effects (edges are straight and don't wiggle), so I'm thinking your camera's pass-thru has a tbc. That's an advantage -- those problems can't be repaired after capture, so you're one step ahead with that camera. Sony's 4:1:1 recorded colorspace was a problem for me (many Avisynth filters don't work with 4:1:1, and converting to RGB looked a little strange at times. Anyone have a take on this ? ? ?).

Some people would agree with a correction, others wouldn't. I see both clips as good exercises in learning to fix color problems. I'd probably spend more time and do some research on why some techniques don't respond the way I wanted. But that's just me. If you like the results, you're getting what you want.

This is what I see in the original Video1:
video1 RGB histogram: (1) crushed darks, suppressed midtones, lack of brilliance. (2) oversaturated red relative to other colors, and red "blooming". (3) suppressed bright green.

[Attachment 14429 - Click to enlarge]

vido1_noise_lowdensity: noise and rough gradations in background. Flat skin tones, lack of definition due to low density in lower part of histogr5am. Sharpening won't help; midtones need to be expanded.

[Attachment 14430 - Click to enlarge]

video1_colors_density: low contrast and suppressed midtones give two-dimensional murky appearance. No detail in highlights; sharpening makes it worse. Dark-haired kids on the right have purple hair (dark blues too bright). Shirts aren't white; they are magenta (bright green suppressed).

[Attachment 14431 - Click to enlarge]

Video1_white_gray: (1) whites aren't white. (2) Shadows on white aren't gray; they're purple.

[Attachment 14432 - Click to enlarge]

Not everyone wants the same thing. The tools are there for whatever we want. I gave it a try, but I'm not real happy with it. Histogram is still a mess, and noisy blue background. But it's a start. I did manage to fix the horizontal rip in frames 127 and 128.

[Attachment 14433 - Click to enlarge]

[budwzr]

Sanlyn, is your monitor a CRT?

Obviously, the reds would be very vibrant in what appears to be holiday apparel. The standard is Santa Claus red.

[sanlyn]

Two versions were posted, one more blue, one more red.
The monitors are IPS LCD's from LG and HP, 38-point calibrated with EyeOne Display 2/Xrite. Not using the CRT right now. The HD LCD and plasma are calibrated the same way, to ISF/D6500 standard . How is yours set up?
Vibrant and blooming are two different things. I didn't know that Santa Claus Red was a standard color. I didn't know CRT's were supposed to be more red than LCD's. It's usually LCD's that are set up with pushed reds adn odd color temps like 9000K.
You should see no tint in any panel of the 22-step ramp below.


[Attachment 14438 - Click to enlarge]

[sanlyn]

I don't eyeball for whites, grays, and blacks. I use a pixel sampler. The "color" of the "white" shirts in the original averages Red-192, Green-155 Blue-183. The histograms show the same problem.That RGB combination is not white. The exact shade of red worn in the clip can have many variations. I think the "more red" post is probably closer to what you mean. I don't think that clip can ever look "perfect", it's pretty corrupt to begin with. One does what what likes. I do too, but I also measure. Video2 was much better, less nosie and more lifelike color. But one shot was still measured more red skin tones than the other two, and it sure looks that way every way I view it.

[budwzr]

So the BL is that YES, you can upsample SD to HD and improve it. But not enough to actually justify or bother doing it. AND, said SD should be commercial quality, not VHS.

[sanlyn]

So the BL is that YES, you can upsample SD to HD and improve it. But not enough to actually justify or bother doing it. AND, said SD should be commercial quality, not VHS.

Agreed, 100%. The candidate for upsizing should be pristine. If it isn't, the work is cut out for you. Average VHS usually looks better at SD, upsampled by a decent HDTV -- it will look better than substandard VHS upsampled in software.

Besides, upsampling VHS is such a hassle.

[budwzr]

OK, that's it then. I hereby elect myself as the winner of this discussion.

[sanlyn]

[vhelp]

baloney!

[vhelp]

i still prefer to convert op's source untouched. i think it looks perfect enough as-is. its all about compression and bitrate. h264 shows a better quality video even at lower bitrate, but not too low, for this format, that it brings out the true nature (asthetics) of vhs. i mean. i still want to feel like i'm watching a vhs content. nastalgia. whatever. anyway. i believe that video sources stemming from vhs should not be processed--but with some a exceptions. noise reduction- leaves the video looking worse. if one gives enough bitrate to the vhs encoder, it will look very good on almost any output medium. as far as archving it to dvd media, for get about it. that medium lost its luster with me in these matters. for vhs content, you really want to go with h264. plain and simple. i'm just saying..

the op's source video does have some (slight) contrasting issues but that can be fixed if it bothers him. sometimes, you can just leave it alone during processing/converting and the player or sw player will handle correctly with some adjustments. vlc has a few configuration settings you can change to reduce the contrast issue. it is a simple matter of testing it on your rig. sometimes it resolves the color issue and sometimes not. it all depends. but to test for yourself, do the following in vlc:

1. right-click on vlc window
2. select \Interface
3. select \Preference
4. click on the \video icon
under the Video Settings, under Display, you have two choices:
5. [ ] Accellerated video output (Overlay), and [ ] Use hardware YUV->RGB conversions
which give you a total of 4 output options, though depending on your video container format (ie mpeg2 vs mkv vs mp4 etc)
6. flick each checkbox and observe the output (color level) quality.

brassplayer, post a much longer vhs clip...with movements..from maybe a favorite act/scene so i can convert it and post it on my youtube channel as mine own the equipment/capture setup you have is (IMHO) excellent. only question i have is, was that a commercial tape or did you record it to tape.

[budwzr]

Poster #52 is the winner! Congrats all around.

[unsubscribed]

[robjv1]

^^ It's actually pretty cool to see all of the different attempts and philosophy behind each attempt. I guess in a way that we all develop our own special little recipe for what to leave in and what to take out. Sensibilities are influenced by your connection to the material as well I think -- I'd probably process my home movies that I've watched a thousand times differently than those belonging to someone else. There are certain color palettes and quirks about some of my VHS footage that would be missed if I took them out, even if the colors were unbalanced to begin with.

[poisondeathray]

I think the answer is NO. You can't increase the resolution. You can only address the typical problems with VHS like noise and bad colors. VHS is crappy - especially the chroma resolution is abysmal.

But I think you can attempt to minimize the damage caused by typical VHS filtering like noise reduction - as it's very easy to go overboard. Typical noise filtering will actually reduce effective resolution

So if brassplyer's original goal (and purpose of this thread) was to optimize clarity and resolution - that's going to be a trade off vs noise reduction . That's a subjective choice. Personally I'm not a fan of the "oil painting plastiky look" , which is the result of overzealous denoising and sharpening. VHS doesn't contain a whole lot of detail to begin with but I would rather attempt preserve whatever fine details in the crappy VHS at the expense of a bit of noise .

I agree with Sanlyn that there is a red color cast, but there is a lot of subjectivity in regards to color and grading . Everyone is going to have a different opinion on what the colors should look like - but I think we can all agree the shirts are supposed to be "white". The thing is , in real life , lights are never 100% white, so shirts are never 100% white. There is always a color temperature to lighting, whether artifical or natural. I think it would be a mistake to match the shirts to perfect equal R=G=B white values. Not only does it look unnatural - this probably means you will reduce the fine detail in the shirts as you clip the high end or stretch the curve too much using an 8 bit format

If you're going to deinterlace, personally I would bob deinterlace to 59.94p using a good deinterlacer. The choice of a deinterlacer is critical since your goal was "clarity / resolution".
(Unless it's for youtube, which only supports 30p)

[jagabo]

I think the answer is NO. You can't increase the resolution.

You can't increase the resolution. Ie, once small details are gone you cannot restore them. But you can sharpen the remaining edges.

[robjv1]

[Attachment 14443 - Click to enlarge]

That's a very nice job there, the color balance there looks just about right to me. Maybe just a touch warmer.