Chroma mess

[Mr Chris]

I'm capturing from VHS using a Hauppauge USB-Live2 and AmarecTV, passing through a Panasonic DMR-ES10 on the way. Here are two consecutive frames of a random recording, either side of a cut:


[Attachment 51239 - Click to enlarge]

[Attachment 51240 - Click to enlarge]

(It's a promo - don't judge me.)

You'll notice there's an after-image of the younger lady in the second picture. What's going on here? It looks a bit like the chroma is somehow a frame behind the rest of the video, if that's even possible. I don't think it happens with every capture, although perhaps it's just not always noticeable, but it is quite common. I'm at a bit of a loss to know where to start, and really need to be talked through what could cause this. What have I done wrong? What do I need to do to put it right?

I've attached a very short sample below for you to examine (Ut codec) - all it's been through is a 'Direct stream copy' trim in VirtualDub. The clip contains two cuts, and the issue is evident on both.

[jagabo]

The chroma isn't behind the luma, it's temporally blended. Make sure you turn off the noise reduction filter in the ES10. And the VCR and capture device too (if they have one).

[dellsam34]

Looks like you need a VCR with line TBC to clean that line wiggle, also use S-Video and a S-VHS deck whenever possible especially when chroma problems arise. DNR ON in the VCR can cleanup some of that chroma noise.

[Mr Chris]

Thanks very much for the replies.

The chroma isn't behind the luma, it's temporally blended.

Just to make sure I've got my head around this, are you saying that the chroma of one field is essentially getting 'smeared' into the next field, and then that field's chroma gets smeared into the next, and so on and so forth in perpetuity?

I'm going to have to do some test captures, but I know for a fact that both the noise reduction and comb filter of the ES10 are off. Would I also be correct in saying that this is an issue that, if not fixed at the capture stage, there's no way to put it right later? I expect so.

Apologies for the (probably) newbie questions, but I'm trying to fully understand what's happening here.

[jagabo]

Thanks very much for the replies.

The chroma isn't behind the luma, it's temporally blended.

Just to make sure I've got my head around this, are you saying that the chroma of one field is essentially getting 'smeared' into the next field, and then that field's chroma gets smeared into the next, and so on and so forth in perpetuity?

Yes. At the scene changes you can see the blending for at least 3 frames (lighter with each successive field):


[Attachment 51258 - Click to enlarge]

Here you see the U channel, as greyscale (after increased saturation to make the problem more visible), of four consecutive fields. Field 55 is just before the scene change. Field 56 is the first field after the scene change. You can see the heavy blending of field 55's chroma (especially obvious in the window in the background). At field 57 the blending is still there but lighter. Even at frame 58 you can see a tiny bit of the blending.

Would I also be correct in saying that this is an issue that, if not fixed at the capture stage, there's no way to put it right later? I expect so.

It's best to avoid it. But you can probably remove most of it with some filtering an AviSynth. I've done so before.

[Brad]

Hi Mr Chris. Are you able to capture this same clip using some of your other hardware that you used in the Blind Test thread?

I'm curious about the source of the luma AGC effect at the second scene change.

[LMotlow]

Looks like you need a VCR with line TBC to clean that line wiggle, also use S-Video and a S-VHS deck whenever possible especially when chroma problems arise. DNR ON in the VCR can cleanup some of that chroma noise.

Note that the O.P. is using an ES10, which has a very strong line TBC. The edge noise is typical VHS noise and it's everywhere in the frames. DNR in tape players is very primitive, mostly just blurring. It can create ghosting and smeared motion, especially with really dirty old tapes like this. There is much better, more sophisticated noise reduction in post-processing with Avisynth.

I know for a fact that both the noise reduction and comb filter of the ES10 are off.

You can turn off the dnr, but not the comb filter.

The clip contains two cuts, and the issue is evident on both.

I'm betting it was either broadcast or recorded with those effects.

[Brad]

I know for a fact that both the noise reduction and comb filter of the ES10 are off.

You can turn off the dnr, but not the comb filter.

It's a setting on the PAL model, for PAL sources only.


[Attachment 51452 - Click to enlarge]

[Mr Chris]

Yes. At the scene changes you can see the blending for at least 3 frames (lighter with each successive field):

. . .

Thanks for taking the time to do all that - I meant to follow up but got a bit distracted.

If you all bear with me, I'll try to get some samples up at the weekend.

[Mr Chris]

Here we go, then - I'll try and go in as logical an order as possible, but it's still going to be a bit all over the place. Before I start, the following stay constant in all of the tests:

Capture device is a Hauppauge USB-Live2
Capturing software is AmaRecTV
Codec is Ut

Bearing in mind that the issue in question comes and goes, it stands to reason that none of the above are suspects in this case. What else can we eliminate early on? Well, let's see if the chroma blurring is burned into the recording on the tape itself. The 'rawest' capture I can do is from my Sharp VCMH704 consumer-level VHS player directly into the Hauppauge via composite (01chroma.avi in the attachments):

Sharp VCMH704 - [Scart|Composite] - Hauppauge

[Attachment 51536 - Click to enlarge]
01screen.png

That screenshot (and please do click on the screenshot - it's resized when embedded in the post, which doesn't present an accurate picture) is the first frame after the scene change. Ignoring all of the multiple other issues with that capture, there's no sign of any after-image to my eyes (the black vest top from the original post is my main point of reference). I'm satisfied that the tape itself is not to blame.

Let's now feed the same VHS machine through a Panasonic DMR-ES10 (02chroma.avi):

Sharp VCMH704 - [Scart] - PanaES10 - [S-video] - Hauppauge

[Attachment 51537 - Click to enlarge]
02screen.png

Your eyes are probably more sensitive to these issues than mine, but again, I see no sign of the chroma mess in this one, either. So do we agree that it's not the tape, and it's not the ES10?

But that's all been composite so far - let's see if S-video is the problem. Time to fire up my old SVHS machine (03chroma.avi):

Panasonic NV-HS1000 - [Scart] - PanaES10 - [S-video] - Hauppauge

[Attachment 51538 - Click to enlarge]
03screen.png

The NV-HS1000 is in passive mode there, with everything switched off apart from the TBC. No issue that I can see.

All of this leads me to conclude that the problem lies with my previously-unmentioned latest acquisition, a JVC HR-S9600. Here's a capture through my normal preferred chain (06chroma.avi):

JVC HR-S9600 (TBC on) - [S-video] - PanaES10 - [S-video] - Hauppauge

[Attachment 51539 - Click to enlarge]
06screen.png

I've done a lot of toggling of settings (and we may get in a tangle here, because the PAL version gives different names from the NTSC model) and also tried different cables and inputs/outputs, but nothing can shift that after-image. Is this a general problem with the HR-S9600, or is it just my unit? I'd be surprised if I'm the first person to notice this.

There is, however, one final vaguely interesting discovery. The following is a capture through a chain identical to the previous, but that one had the TBC on and this one has the TBC off (07chroma.avi):

JVC HR-S9600 (TBC off) - [S-video] - PanaES10 - [S-video] - Hauppauge

[Attachment 51540 - Click to enlarge]
07screen.png

You have to look a lot closer, but an after-image is still there. However, now I wonder whether that's luma we're seeing rather than chroma. Try opening the video in (say) VirtualDub and rocking back and forth between the two frames following either of the scene changes. It's more of a shadow than a colour artefact, isn't it? Or is it something else?

I have a feeling I'm stuck with this issue, which is particularly disappointing as my captures have otherwise never been so good. Any ideas? All thoughts and theories welcome.

[jagabo]

I've read that (some?) JVC S-VHS decks switch the noise reduction function along with the TBC. Ie, you use neither or you use both. Your captures here appear to verify that claim. And yes, from #7 it looks like the noise reduction in the luma is never completely disabled.

video #7, Y, U, and V channels (top field on top, bottom field on bottom) all contrast stretched:


[Attachment 51548 - Click to enlarge]

AviSource("07chroma.avi") 
AssumeTFF()
SeparateFields()
StackVertical(SelectEven(), SelectOdd())
StackHorizontal(ColorYUV(cont_y=1000, off_y=-100, cont_u=-256, cont_v=-256), \
                ColorYUV(cont_u=600, cont_v=600).UtoY(),\
                ColorYUV(cont_u=600, cont_v=600).VtoY())

There's definitely some residual of prior fields in the luma channel. I don't really see any in the chroma channels. Compare to video #6 with obvious ghosting in all three channels:


[Attachment 51553 - Click to enlarge]

[Mr Chris]

Thanks again for doing that - it seems to confirm what I suspected. The issue is only occasionally noticeable at normal playback speed, so it's not like I can't live with it, but my goodness is it a frustrating 'feature'.

I've read that (some?) JVC S-VHS decks switch the noise reduction function along with the TBC. Ie, you use neither or you use both. Your captures here appear to verify that claim.

Now you mention it, I've just had a closer look, and the button that turns the TBC on and off is actually labelled "Digital TBC/NR", so it really is all or nothing regardless of the menu settings. It seems I've managed to find a bit of a downside to a very well-respected VCR.

[KarMa]

I've read that (some?) JVC S-VHS decks switch the noise reduction function along with the TBC. Ie, you use neither or you use both. Your captures here appear to verify that claim. And yes, from #7 it looks like the noise reduction in the luma is never completely disabled.

Yeah I've read here that when JVC is in the EDIT mode, the DNR is nearly off or atleast much less aggressive.

[KarMa]

double post

[Brad]

My analysis of (US NTSC) JVC S-VHS settings based on samples posted by someone else: http://www.digitalfaq.com/forum/video-capture/8455-comparing-jvc-vcr.html#post64723

Specific VCR there is JVC SR-VS30U. Other models may differ a bit.

There are no shot changes in the sample posted there, so I may have been a bit hasty to say that "there is no luma NR" for EDIT + TBC/NR.

[KarMa]

There are no shot changes in the sample posted there, so I may have been a bit hasty to say that "there is no luma NR" for EDIT + TBC/NR.

If I were to post JVC S-VHS w/ and w/ out TBC samples what do you want? By shot changes do you mean scene changes?

[KarMa]

double post

[Brad]

There are no shot changes in the sample posted there, so I may have been a bit hasty to say that "there is no luma NR" for EDIT + TBC/NR.

If I were to post JVC S-VHS w/ and w/ out TBC samples what do you want?

Just the three setting combinations I analyzed there. The chroma NR is most obvious in very dark scenes, if you're asking about what specific sort of clip I'd like to see.

• - = disabled
• + = enabled
• Cal = Video Calibration
• R3 = Digital R3

1. -TBC&NR, EDIT, -Cal, -R3
2. +TBC&NR, EDIT, -Cal, -R3
3. +TBC&NR, NORM, -Cal, -R3

Do you have Spider-Man? VCR comparison, SP mode [WARNING: auto-load!]

Or certain other Hollywood movie retail tapes; I can PM you a list. I'd want to be able to directly compare your JVC S-VHS captures to my JVC D-VHS.

By shot changes do you mean scene changes?

Yes. To me, "shot change" makes more sense as a technical term for merely cutting between different camera takes. Whereas "scene change" implies a literal change in location or "unit of action / segment of a story in a motion picture."

http://www.cs.nchu.edu.tw/~crhuang/ShotChangeDetection/scd.htm
https://dcmp.org/learn/132-a-note-on-carrying-captions-over-a-shot-change
https://support.syncwords.com/hc/en-us/articles/360018533272-Shot-Change-Detection

https://www.worldscientific.com/doi/10.1142/S021946780100027X
https://www.worldscientific.com/doi/10.1142/S0219622004000957
https://www.worldscientific.com/doi/10.1142/S0219467817500024

https://www-nlpir.nist.gov/projects/trecvid/shots.html

@Mr Chris: Thank you for fulfilling my request, by the way. Forgot to say that in my last post.

[KarMa]

I actually bought Spiderman after seeing your large Spiderman comparison post. Which scene or scenes do you most want?

Second thing, my JVC SR-MV50 (also have a MV45) does not have NORM. It has AUTO/SHARP/SOFT/EDIT. A bit ago I did a Jurassic Park comparison and it looks like the SHARP and AUTO did just about the same processing, while EDIT was visually different. My machines are from 2006 and 2007 so maybe JVC changed the name of NORM to SHARP?

[Brad]

I actually bought Spiderman after seeing your large Spiderman comparison post.

Awesome!

Which scene or scenes do you most want?

Dark alley/swinging immediately after Uncle Ben's death scene: contains saturated red makeshift costume against black and brick backdrop.

• Approx. timecode from Canadian tape start is 0:52:59. (US VHS opening is 30s longer due to Cingular ad, so from USA tape start should be 0:53:29.)
  • Alternatively, approx. timecode from movie start is 0:44:39 (i.e. reset counter to 0:00:00 @ fade-to-black after "This film has been modified" screen).

The Canadian tape has about one frame of bleed-through baked into the transfer to begin with, visible at shot changes throughout the movie. But VCR DNR makes it worse.

Second thing, my JVC SR-MV50 (also have a MV45) does not have NORM. It has AUTO/SHARP/SOFT/EDIT. A bit ago I did a Jurassic Park comparison and it looks like the SHARP and AUTO did just about the same processing, while EDIT was visually different. My machines are from 2006 and 2007 so maybe JVC changed the name of NORM to SHARP?

It's the same as all the other JVCs; if you turn off Video Calibration then AUTO becomes NORM.

SR-MV45U / SR-MV55U manual:


[Attachment 51605 - Click to enlarge]

[Mr Chris]

I've done a bit of experimenting with EDIT mode. The path for both of these captures is JVC HR-S9600 - [S-video] - PanaES10 - [S-video] - Hauppauge.

EDIT mode, TBC off

[Attachment 51610 - Click to enlarge]
08screen.png

EDIT mode, TBC on

[Attachment 51611 - Click to enlarge]
09screen.png

For the same path in NORM mode, TBC on, Digital R3 off, and the PAL equivalent of Video Calibration ("B.E.S.T." mode) turned off, see 06chroma.avi in post #10.

Comparing only the after-image and nothing else, 09chroma.avi is probably the best it's going to get with the TBC on - it's definitely reduced, though unfortunately still clearly visible. So switching to EDIT mode does seem to help the HR-S9600 a little when the TBC is on, which to me suggests a double-helping of NR is applied with a NORM+TBC setup. The luma problem doesn't seem to change too significantly when the TBC is off, whether in NORM or EDIT mode.

Besides the after-image, how do you think 06chroma.avi and 09chroma.avi compare generally? The ITV logo in the top corner and those sort of windowed light panels on the set are good things to look at if you want to line these up in an NLE. I'll have to do a much longer capture in EDIT+TBC mode to see if there are any other major advantages or disadvantages over NORM+TBC, and perhaps try a dodgier tape as well. Or should I bite the bullet and forget about the TBC all together (thereby sidestepping the chroma issue), unless I 100% definitely can't get a stable capture without it? Decisions, decisions...

[KarMa]

It's the same as all the other JVCs; if you turn off Video Calibration then AUTO becomes NORM.

Learned something new and I was able to capture in NORM. I posted my examples over in your old thread. https://forum.videohelp.com/threads/379450-VCR-comparison-SP-mode-%5BWARNING-auto-load%21%5D