Calibrating luminance levels/color for capture from VHS

[jagabo]

Just went and got my DVD-R out of storage and the S-video IN on my VC500 won't connect directly to the S-video OUT on my DVD-R and I don't have a cord nor the finances to get one at this point

This picture at Amazon shows what looks like a standard s-video plug:

http://www.amazon.com/gp/customer-media/product-gallery/B000VM60I8/ref=cm_ciu_pdp_imag...e=UTF8&index=2

A standard US$5 male/male s-video cable should work.

http://www.amazon.com/Belkin-F8V308-06-6-Feet-Plated-S-Video/dp/B00006HVWL/
http://www.amazon.com/S-Video-Cable-ft-Gold-Plated/dp/B0002MQGK4/

[Cherbette]

Thanks. Unfortunately since I'm out of a job right now I don't have 2 pennies to rub together. I was hoping someone might have an extra. If not, I may just have to wait until I've got some money and go buy one.

[Cherbette]

For now, I will just use the RCA out on the DVD-R. Here are some screenshots from the menu...anything I should change?

[Cherbette]

And here is a sample video with DVD-R as a passthrough using RCA out, and with the settings in the screencaps above:
http://www.mediafire.com/?5y5d6xwfgvp2wr1

[jagabo]

And here is a sample video with DVD-R as a passthrough using RCA out, and with the settings in the screencaps above:
http://www.mediafire.com/?5y5d6xwfgvp2wr1

That cap is much better than the others. The time base is much cleaner.

[Cherbette]

Yeah I noticed that right away. What a useful little tip that I acquired here

[jagabo]

There's also much less of that purple/green chroma noise. I believe that's because of the cleaner time base too.

[Cherbette]

What could I do to improve upon that clip with Avisynth? Should I change anything in the DVD-R settings? Some of it was a bit foreign to me...

[sanlyn]

Just went and got my DVD-R out of storage and the S-video IN on my VC500 won't connect directly to the S-video OUT on my DVD-R and I don't have a cord nor the finances to get one at this point

This picture at Amazon shows what looks like a standard s-video plug:

http://www.amazon.com/gp/customer-media/product-gallery/B000VM60I8/ref=cm_ciu_pdp_imag...e=UTF8&index=2

A standard US$5 male/male s-video cable should work.

http://www.amazon.com/Belkin-F8V308-06-6-Feet-Plated-S-Video/dp/B00006HVWL/
http://www.amazon.com/S-Video-Cable-ft-Gold-Plated/dp/B0002MQGK4/

Folks, I believe that VC500's s-video thingie is an output, not an input. Even with the adapter, the VC500 might not take anything as input on its s-video plug. I couldn't find a VC500 manual to check that, however.

cherbette, big improvement with the pass-thru!

[jagabo]

Folks, I believe that VC500's s-video thingie is an output, not an input.

According to this it's an input:

http://ecx.images-amazon.com/images/I/41FRKiOamyL.jpg

I suppose it could be wrong. But Amazon's description says:

The VC500 can capture video and audio from almost any video device, such as VCR, Camcorder, DVD player, or any device supporting video output through an S-Video or composite RCA connection.

Cherbette, check the capture software and be sure it's set to capture from the s-video port, not the composite port.

[Cherbette]

Well until I get a S-video cord I won't be able to capture through the S-video port. I'm fairly sure it's an input as well...

[sanlyn]

Folks, I believe that VC500's s-video thingie is an output, not an input.

According to this it's an input:

http://ecx.images-amazon.com/images/I/41FRKiOamyL.jpg

I suppose it could be wrong. But Amazon's description says:

The VC500 can capture video and audio from almost any video device, such as VCR, Camcorder, DVD player, or any device supporting video output through an S-Video or composite RCA connection.

Cherbette, check the capture software and be sure it's set to capture from the s-video port, not the composite port.

SHucks, jagabo, you beat me to it (why am I not surprised?). I just found that material.

[jagabo]

[QUOTE=Cherbette;2109398]What could I do to improve upon that clip with Avisynth?
I'd bump the contrast up a bit and maybe use a little noise reduction.

ColorYUV(cont_y=25)
McTemporalDenoise(settings="low", interlaced=true) # or medium

Should I change anything in the DVD-R settings? Some of it was a bit foreign to me...

Leave the 3D comb filter on for composite video. Try the Line-in Noise Reduction and see what you think of it. In my experience most noise reduction filters in DVD recorders is too strong -- leading to ghosting. I suspect the black level control is just to account for the difference between the Japanese and American IRE levels (Darker for USA, Lighter for Japan). You probably don't need to change that but you can try it and see what happens.

[sanlyn]

I don't know if the Panny has a 3d comb filter on composite or if it works in pass-thru (never tried it, but it works when recording to DVD). Be careful; Panny's MPEG noise filter can often overdo it, and artifacts like block noise are common with that filter. Toshiba's is much cleaner, but more sophisticated (and expensive).

Worked with the color a bit and centered the image in AviSynth (it was cropped right and bottom). MPEG attached.

VirtualDub + gradation curves + temp smoother (=3) + 2D by Casaburi (set very low. I hate that 2d filter, it tends to block up highlights). Would have used NeatVideo, but there's no frame in this clip that makes a suitable test patch. Color's still not quite right. You could play with this forever. There's some flicker, you see it in the red background. I cut the clip in two parts, the first part set to plain grayscale, worked with color on the part with Cher. Red's a bit over saturated (that's a Panasonic for ya!). Around frame 14 or 18 there are white spots; they could be removed in AviSynth, but archival footage often looks like that. I'd just leave 'em there. I had to cover the old borders with cleaner ones (RGB 8) in TMPGenc Plus. There's some blue discoloration along the right border.

VHS has uneven color/luma from scene to scene. Very common. That's why it's almost impossible to correct VHS color entirely during capture. You set up a scene "perfectly", the next scene looks like crap.

[jagabo]

I don't know if the Panny has a 3d comb filter on composite or if it works in pass-thru

It's in the first line of the screen cap he posted. "3-D Y/C Separation"

I didn't touch the color because there were so few colors in the sequence. The B/W portion had a slight green cast. ColorYUV(off_u=1, off_v=1) will fix that.

[sanlyn]

Oh, that's right. I'm thinking of my old DMR-ES20. Doesn't have a 3-D Y/C. My Toshibas do. Works good.

[sanlyn]

As captured, the StillCher2 video at 720x480 has left and right pillars. Are these pillars supposed to appeart in the final DVD? If they're removed, the frame size will generate an error from encoders (illegal frame size). So if the pillars are removed, the video will have to be resized horizontally (again), right? I'm guessing that the original source was 4:3 letterboxed. If it wasn't letterboxed originally, then the pillars as well as the letterbox has to be removed, and the frame size to the encoder will not be kosher. So, another resize?

[jagabo]

As captured, the StillCher2 video at 720x480 has left and right pillars. Are these pillars supposed to appeart in the final DVD?

They're on the tape, why shouldn't they be on the DVD. He might want to shift the image left a bit. And if the entire video is letterboxed I'd blacken the borders so they don't eat up bitrate. That will also get rid of the line 21 noise at the top.

Crop(22,40,-4,-24)
AddBorders(16, 32, 10, 32)

Obviously, make sure that works for the rest of the video too.

[sanlyn]

As captured, the StillCher2 video at 720x480 has left and right pillars. Are these pillars supposed to appeart in the final DVD?

They're on the tape, why shouldn't they be on the DVD.

Umm, that's what I'm asking. I can understand the letterbox, but...if the pillars are in the source, then stay they will. I left the pillars, but centered the image in AviSynth.

[Cherbette]

I don't know if the Panny has a 3d comb filter on composite or if it works in pass-thru (never tried it, but it works when recording to DVD). Be careful; Panny's MPEG noise filter can often overdo it, and artifacts like block noise are common with that filter. Toshiba's is much cleaner, but more sophisticated (and expensive).

Worked with the color a bit and centered the image in AviSynth (it was cropped right and bottom). MPEG attached below, or download here: http://dc204.4shared.com/download/VD2O7_5Y/StillCher2_try2.mpg .

image reduced 25%. Original is 720x480.

[Attachment 8866 - Click to enlarge]

VirtualDub + gradation curve + temp smoother (=3) + 2D by Casaburi (set very low. I hate that 2d filter, it tends to block up highlights and shadows). Would have used NeatVideo, but there's no frame in this clip that makes a suitable test patch. AviSynth smoothers would work well, just didn't have time. Color's still not quite right. You could play with this forever. There's still some flickery noise, seen in the red background. I cut the clip in two parts, the first part set to grayscale, worked with color on the part with Cher. Red's a bit over saturated (that's a Panasonic for ya!). Around frame 14 or 18 there are white spots; they could be removed in AviSynth, but archival footage often looks like that. I'd just leave 'em there. I had to cover the old borders with cleaner ones (RGB 8) in TMPGenc Plus. There's some blue discoloration along the right border.

VHS commonly has uneven color/luma from scene to scene. Makes it almost impossible to correct VHS color entirely during capture. You set up a scene "perfectly", the next scene looks like crap.

I think that looks pretty great. I think I have the noise reduction turned to off on the DVD-R if I'm not mistaken...would there be a way for the to correct the over-saturated reds at capture that you pointed out or is that something that is going to be part of the Panasonic's territory. Regardless, I don't see the jitter any more and that (to my eyes) has cleared up many of the issues we all noticed.

[sanlyn]

Unless you have a big-nad problem video that requires a major fix during capture or in YUV (they exist), most color is done post-capture. This one required a lot of tweaking, the histogram is a mess. Again, VHS can change color balance/luma on a dime; set it up one way for a scene, and BAM! It turns green or something, because red was cut earlier. But you did well to keep an eye on darks and brights during capture - that's most important. If you use Panny's denoiser watch it carefully. When it works, it works. When it doesn't, it's painful.

I also ran a MCTemporalDenoise version. Skin colors visibly cleaner, some of that grainy noise still needs work. If you want to try MCT later, I'm still trying to get the components together and zipped.

[Cherbette]

What filter would you guys suggest for adjusting RGB histograms in VirtualDub? Right now I'm using red/green/blue adjustment 1.0 by Donald Graft but I don't really like the way it works...

[jagabo]

What filter would you guys suggest for adjusting RGB histograms in VirtualDub

He's probably using Color Tools.

My general rule is that color changes caused by VHS are YUV based. Color changes caused by fading film are RGB based. Color changes caused by a poor telecine process could be either.

[Cherbette]

Hmmm...so what should I do? I am indeed using Color Tools...

[Cherbette]

I don't know if the Panny has a 3d comb filter on composite or if it works in pass-thru

It's in the first line of the screen cap he posted. "3-D Y/C Separation"

I didn't touch the color because there were so few colors in the sequence. The B/W portion had a slight green cast. ColorYUV(off_u=1, off_v=1) will fix that.

Here's another clip hopefully with more colors...the skin tones on the interview portions of this documentary are very yellow/greenish and based on how the rest of the non-interview clips look in the documentary I would guess that (to a certain extent) it was probably present in the broadcast. Perhaps it was the lighting? Maybe you guys can give me some better feedback on that...

http://www.mediafire.com/?5793nvekk45a7xq

Also I readjusted the brightness and contrast while watching the histograms for this one...so this isn't the same exact capture as last time.

Another thing: I have...I have "sharpen" set to zero in my proc amp. I figured this is something that should be done in post-processing?

Here is a screencap of what I think may address the questions sanlyn had about ratio:


This was a UK documentary and I forget where I got it but it's on a NTSC VHS from what appears to be a digital broadcast. It looks to be letterboxed on the tape itself so maybe that's how it was broadcast or maybe the fact that it was broadcast with PAL dimensions is the issue?

[intracube]

It looks to be letterboxed on the tape itself so maybe that's how it was broadcast or maybe the fact that it was broadcast with PAL dimensions is the issue?

Ooh, ahhh *waves hand in the air* - I have an explanation.

If the programme was made after 1998/1999 it's likely to have been produced in 16:9 SD, then formatted to 14:9 within a 4:3 frame. You want to keep the small bars at the top/bottom of the frame and it should be kept as 4:3.

If you're likely to watch this on a 16:9 TV, you could also crop the black bars off the top/bottom and generate thicker bars at the sides to create a 16:9 frame with the slightly narrower 14:9 image within it.

[jagabo]

Hmmm...so what should I do? I am indeed using Color Tools...

I came close to what he did with Levels, HSV, and RGB Adjust:









I was just adjusting the colors here, not addressing the noise.

[jagabo]

the skin tones on the interview portions of this documentary are very yellow/greenish and based on how the rest of the non-interview clips look in the documentary I would guess that (to a certain extent) it was probably present in the broadcast. Perhaps it was the lighting?

I haven't seen the new upload yet but lighting is always an issue. Looking at a single shot you can't know if the colors are wrong because of the recording or if colored lighting was used and the intention was for those colors. All you can do is look at the entire video for hints. An interview is more likely have normal lighting. Stage performances are more likely to use colored lights at times.

Also I readjusted the brightness and contrast while watching the histograms for this one...so this isn't the same exact capture as last time.

Be careful about using histograms when you have large black borders. You want to adjust for the actual picture content, not the black borders.

Another thing: I have...I have "sharpen" set to zero in my proc amp. I figured this is something that should be done in post-processing?

In my experience the sharpen filter in VHS decks and DVD recorders creates oversharpening halos and enhances noise. If you're going to filter in software later I would minimize filtering by the DVD recorder.

This was a UK documentary and I forget where I got it but it's on a NTSC VHS

I thought I saw some PAL/NTSC conversion field blending in the earlier video but there was so little motion it was hard to say for sure. There are techniques for reducing field blending but we can address that later, after I've seen the new clip.

[sanlyn]

As jagabo shows, you can learn to use any number of tools, in RGB and YUV alike. You do have to get accustomed to reading pixel values and reading what -grams and -scopes are saying.

I made the StillCher2 color adjustments with the VirtualDub gradation curve, http://members.chello.at/nagiller/vdub/readme.html. I use others, in VDub and AviSynth. Similar to the same tool in Premiere Pro, After Effects, etc. But in VDub it can work in RGB, YUV, LAB, and others. You can chain curves together: one curve for major adjustments, then add more curves to fine-tune. If you chain more than 3, your work gets grainy and discolored -- as is usually the case with too much filtering.

ColorTools doesn't change anything, it just measures. With it you can spot problems: crushed/blown-out colors, suppressed midtones, color dominance or deficit, etc. Use histograms and vectorscopes in conjunction with free pixel samplers like CSamp (works like the pixel sampler in Adobe). Your video StillCher2 provides examples:

Frame 14 is below, with ColorTools' histogram. The green arrow on the gal's face shows a shadow area that was sampled with CSamp. ColorTools reads the whole image. Because the image is mostly grays, all colors in the histogram are similarly shaped and distributed. In a color image, colors won't look this much alike.





Above, two 3x3-pixel samples from Frame 14 taken with CSamp. The monochrome image looks a bit green: the CSamp "A (Color)" reading on the left tells you, yes, it's about 5% off from true gray (RGB 95-100-96). If you convert the image to grayscale and strip out chroma, you get the "B (Grayscale)" reading. In this case CSamp reads a true gray in that spot (RGB 98 98 98). Three basic "colors" you should to know about:
black (RGB 0 0 0, or 15 15 15 in Rec601)
white (RGB 255 255 255, or 235 235 235 with much video)
gray (middle gray = RGB 128 128 128)

CSamp (free!): http://www.netreach.net/~gavin/gavsfreeware/csamp.htm . Can read AviSynth output, too, in VirtualDub's windows.

Oops, problem here at home. Continued, next post . . .

[sanlyn]

The image below is Frame 34, the first full-color frame in StllCher2.



The histogram shows why this image looks too yellow-red (and too green in some places). Cher is nearly orange. In the histogram, Blue is shoved off to the left into a small wedge with a big peak at RGB 14. No wonder Cher's eyeballs look pale yellow. It's possible that the producers wanted a warmish image, but I don't think Cher's face should be the color of cantaloupe.



Above, left, a Csamp reading off the bright part of Cher's nose. The histogram is correct! With normal lighting, blue should be 20 or 30 points higher. We expect the red, because Cher is no pale-skinned wimpess. I used the gradation curve to modulate blue, and to fix other colors. Above, right, is the Blue RGB panel of the curves filter. At default, Blue's adjustment line is diagonally straight -- RGB Zero and the darkest points are at bottom left, Blue 255 at upper right. The grid's exact center is RGB-128 Blue. The arrows show what happens to Blue when you re-shape the line.



Above, some early results. The gradation curve control panel is at the left, preview window at right; in the center of the preview window I've placed the CSamp floating panel. Blue's adjustment line is considerably changed. I'd class this as a somewhat extreme adjustment (happens often). The blue is a bit uneven, too much of it in the darker zones, and Cher is turning purple just left of her mouth. The bright spot between her eyes is getting bluish. But we're in the ballpark. Fine tuning comes later, with other curves filters that will be added in the filter list. You can save the settings as an .amp file, or with VirtualDub's "Save processing settings" menu item.

IMHO I have a bit too much blue. General rule: work all day, then come back 3 days later.



Above: a gradation curve "RGB" panel set to limit the Rec601/other videos color and luma range. Well, I cheated: it's really 12-235. I usually add this "fixer" at the bottom of the filter list.

Try Google for online articles about working with Adobe curves, levels, histograms, etc. Most tutorials aim at photos, but for video the adjustment principles are the same. Working with RGB also helps you understand more about YUV. Many color problems can be solved only in YUV.

Take a quick look at John Dickinson's "Curves Quick Reference Guide", http://www.motionworks.com.au/2007/01/curves_guide/. Click on the link in the header article that says "View Guide". The GIF image that displays can be copied and saved.