VHS capturing - luma/chroma crosstalk? What do you think?

[Slennox]

Hi, guys.

I'm in the process of digitize a huge VHS tape collection. My setup is as follows:

• VCR: Panasonic NV-FJ625. Standard VHS. Only output to SCART.
• SCART to composite cable adapter
• Capture device: Matrox MXO2. Input from Y composite cable (plus 2 RCA for audio).
• Capture software: Adobe Premiere CC 2014 (OS X)

With every tape I try to capture I can see these lines which are not present when I play them on TV/monitor (without using neither MXO2 Mini nor the SCART->Composite adapter):




These artifacts seem to be related to luma/chroma crosstalk. What do you think?

• A defective SCART -> composite adapter? An incorrect one? I read there are different types of SCART cables (raw sync?) and an incorrect one may produce these artifacts.
• If the MXO2 Mini isn't able to properly separate luma and chroma from the composite signal that would be worse as I would need to find an S-VHS to replace my current deck and those are certainly expensive... In any case, any recommendation to get rid of these problems?

Indeed this problem seem to be quite common on capturing from analog sources...


Any guidance would be appreciated.

[jagabo]

luma/chroma crosstalk?

Yes, the chroma subcarrier wasn't completely removed from the luma. The capture device's comb filter is should do a better job. Make sure you haven't disabled the comb filter.

Decrease the resolution to 360x576. Then increase back to 720x576. It will be mostly gone and there won't be much loss of sharpness because VHS only has about 360 lines of resolution. I used Spline36Resize(width/2, height).Spline36Resize(width, height) in AviSynth:

[Slennox]

I've contacted my capture device vendor (Matrox) whether it's possible to tweak the comb filter or the luma/chroma separation process of MXO2 device. But I wouldn't expect much as the configuration options have been very limited on every firmware release.

So we can rule out a defective SCART-composite adapter causing these artifacts as it is just a passthrough of the signal, right? Then the remaining path to follow is to invest on a S-VHS equipment that can output an S-Video signal where luma and chroma are physically separated.

Finally, thank you providing the resizing workaround, very nice! However I would like to see my options first on capturing a cleaner signal (well, I suppose it's just one option...)

[jagabo]

I've seen a few people claim a different composite cable cleared this up for them.

It's a little difficult to say exactly what's going on with SCART. Since it can carry composite, s-video, and RGB people sometimes misconfigure their adapters. But I suspect that's not where your problem lies.

An S-VHS deck and s-video chain should get rid of it. It's always best to avoid the problem rather than fix it later.

Beware that there are some regular VHS decks with s-video output -- those usually don't work well because the s-video converter is an add-on to their regular composite VHS deck. They simply convert composite to s-video with a cheap circuit which gives output just like you're seeing.

[pandy]

Try defreq or similar frequency http://avisynth.nl/index.php/External_filters#Frequency_Interference_removal domain filter can be useful (you can try to trap frequencies around 4.43MHz i.e. with 720 pixels where luma sampling clock is 13.5MHz this gives you omega=0.328) .

[PuzZLeR]

Decrease the resolution to 360x576. Then increase back to 720x576. It will be mostly gone and there won't be much loss of sharpness because VHS only has about 360 lines of resolution.

Yes, I love this trick. I use it all the time.

An S-VHS deck and s-video chain should get rid of it. It's always best to avoid the problem rather than fix it later.

True. Unfortunately, I haven't really liked any of the near-half dozen S-VHS decks I've owned, so would prefer to use this with a VCR that I like, which just so happens to have crosstalk, but it's easily corrected.

Yes, it would be better corrected at the capture level, however, since this is VHS, the penalty for this trick is so minimal due to the low resolution. Also, VHS never needed s-video to begin with.

(However, I'm sure you know that if you try this trick to remove crosstalk from high quality DvD source, the softening of the video will show lots more.)

Beware that there are some regular VHS decks with s-video output -- those usually don't work well because the s-video converter is an add-on to their regular composite VHS deck. They simply convert composite to s-video with a cheap circuit which gives output just like you're seeing.

To the O/P, also avoid those "adapters" that "convert" composite to s-video. Sure they do the job at the interface level, transferring the signal, but they do not remove the crosstalk. In fact, I believe they make it worse.

I've contacted my capture device vendor (Matrox) whether it's possible to tweak the comb filter or the luma/chroma separation process of MXO2 device. But I wouldn't expect much as the configuration options have been very limited on every firmware release.

I'd love to know what they say here. But from my experience with capture devices, and I've bought over two dozen over the years, when they have a good comb filter they have AGC issues, or vice versa. I have yet to see a device that can handle both problems.

I actually went to school a while ago with one of the founding members from Matrox (yeah, I say this all the time... ) - maybe we should chat about this.

[Brad]

It's a little difficult to say exactly what's going on with SCART. Since it can carry composite, s-video, and RGB people sometimes misconfigure their adapters. But I suspect that's not where your problem lies.

I would actually blame the SCART adapter first, personally. Several threads with this same crosstalk ended up coming down to SCART wiring.

Though usually it's people trying to use SCART -> S-Video.

It should be simple enough to determine whether the MXO2 is at fault simply by connecting an actual composite source like a DVD player directly.

[Slennox]

I think the SCART->composite adapter it's definitely not the culprit. I can't see the bands playing the tapes directly on TV: VHS -> SCART to composite adapter -> TV (composite input).

So the problem looks like to be the Matrox device.

I'd love to know what they say here

You can follow up this thread (I pointed out the resize trick there). They recommend using S-Video...

[Brad]

I can't see the bands playing the tapes directly on TV: VHS -> SCART to composite adapter -> TV (composite input).

I interpreted this to mean that you had only used a direct SCART connection to your TV:

With every tape I try to capture I can see these lines which are not present when I play them on TV/monitor (without using neither MXO2 Mini nor the SCART->Composite adapter)

[Slennox]

I tried the two possibilities (my VHS deck has SCART output only):

• VHS deck -> SCART-to-composite adapter -> TV (composite input)
• VHS deck -> TV (SCART input)

Can't see any bands. Only seem to happen when I use MXO2 Mini as capture device.

[Skiller]

Does it happen with other Composite video sources as well? Try connecting a DVD-player, a set-top box or similar to your capture card and see if the patterns are still there or not.

[Slennox]

I continue with the restoration process. After looking around for high-end S-VHS decks (e.g. JVC HR-S9600) I decided against buying one as they are quite pricey and I'm pretty satisfied with the result you get with the resizing trick. My capture device (Matrox MXO2) already has some features of these high-end S-VHS players (TBC, DNR, etc.)

I'm planning to use this Avisynth script for editing the original capture:

SetMTMode(5,8)
FFVideoSource("capture.mkv")
SetMTMode(2)
ConvertToYV12()
AssumeTFF()
QTGMC(Preset="Slower", EdiThreads=8, SourceMatch=3, Lossless=2, MatchEnhance=0.75, TR2=1, Sharpness=0.1)
Spline36resize(360,576).Spline36resize(720,576)
Crop(12,2,-10,-10) .AddBorders(12,2,10,10)
SetMTMode(1) 
GetMTMode(false) > 0 ? Distributor() : Last

Then set up ffmpeg as frame server and x264 as encoder through piping. I'm on OS X and I'm forced using wine for the Avisynth stuff but the encode runs natively:

wine ffmpeg.exe -i restoration.avs -f yuv4mpegpipe - | x264 --stdin y4m --preset veryslow --tune film --crf 16 -o edited.m4v -

I think the results of QTGMC are very good. I know you always lose detail de-interlacing but I'll keep the original lossless capture archived in the cloud anyway. The progressive encoded video would be the live version.

I've upload 1 min sample of the original captured footage (224,7 MB) and the edited video (27,6 MB).

I'm pretty new to restoration filtering so I would be grateful if you could check the footage and point out any improvements to the process.

Thank you.

[themaster1]

you have forgotten to specify interlaced=true ; like this: ConvertToYV12(interlaced=true)
I sugest you use a a color denoiser like CCD and a sharpener at the end such as finesharp

SetMTMode(5,8)
FFVideoSource("capture.mkv")
SetMTMode(2)
ConvertToYV12(interlaced=true)
AssumeTFF()
ConverttoRGB32(matrix="rec601",interlaced=true)
separatefields()
LoadVirtualDubPlugin("C:\Program Files (x86)\VirtualDub\plugins\Camcorder_Color_Denoise_sse2.vdf","CCD",1 )
CCD(10,1) # 0 - 100 # Default =30 /
weave()
ConverttoYV12(matrix="Rec601",interlaced=true)
QTGMC( Preset="fast", edithreads=8,SourceMatch=3, Lossless=2, Sharpness=0.2, TR0=2,TR1=2,TR2=3,Rep0=1,Rep1=1,Rep2=1,SubPel=4,Ch romaNoise=true,DenoiseMC=true,NoiseTR=2,truemotion =true,MatchEnhance=0.6)
Spline36resize(360,576).Spline36resize(720,576)
FineSharp()
Crop(12,2,-10,-10) .AddBorders(12,2,10,10)
SetMTMode(1)
GetMTMode(false) > 0 ? Distributor() : Last

[jagabo]

Instead of downsizing, upsizing, then QTGMC, try downsizing, QTGMC, then upsizing. QTGMC will process the smaller frame ~twice as fast. It won't make as big a difference overall since you are using such slow encoder settings. Encoding with x264 at the slow preset I got 12.5 fps with the former, 21 fps with the latter.

That video is in really poor shape. It contains interlaced frames but it's encoded progressively (this has blurred the chroma channels together). Every field is a blend of three or four of the original fields (so it looks like a multiple exposure). Is that your original video or has it already been reencoded? How are you capturing it? You should look into fixing these issues before investing a lot of time filtering and encoding.

Normally, interlaced sources should use ConvertToYV12(interlaced=true). But since your video was encoded progressively that isn't necessary here.

Those x264 settings are pretty extreme. Be aware that you may not be able to play the resulting video on anything but a computer.

Some other filtering you can consider once you've fixed your capture problems:

1) dehalo_alpha() to reduce the VHS oversharpening halos (since you probably can't disable the sharpening filter in the VHS deck).

2) Sharpening the chroma with MergeChroma(last, aWarpSharp(depth=x)).

3) Shifting the chroma with ChromaShift to put it back in the right place).

4) Other mild sharpening.

I did this and got slightly better results:

SetMtMode(5,4) # quad core CPU
Mpeg2Source("capture.d2v", CPU2="ooooxx", Info=3) 
SetMtMode(2)
ConvertToYV12()
Spline36resize(360,576)
QTGMC(Preset="Slower", EdiThreads=8, SourceMatch=3, Lossless=2, MatchEnhance=0.75, TR2=1, Sharpness=0.1)
Dehalo_alpha(rx=2, ry=2, lowsens=10, highsens=90)
Sharpen(0.3, 0.2)
MergeChroma(last, aWarpSharp(depth=20))
Spline36resize(720,576)
Sharpen(0.2)
ChromaShift(l=-2)

I used your QTGMC settings though I don't think the lossless mode helps with this video. I didn't bother with the cropping for this test.

[Slennox]

That video is in really poor shape. It contains interlaced frames but it's encoded progressively (this has blurred the chroma channels together). Every field is a blend of three or four of the original fields (so it looks like a multiple exposure). Is that your original video or has it already been reencoded? How are you capturing it? You should look into fixing these issues before investing a lot of time filtering and encoding.

It was captured using Matrox Vetura Capture, a simple free application provided with the capture device (MXO2 Mini). I selected "PAL 8-bit" as input resolution and "Matrox MPEG-2 I-frame" as capture codec. I think this codec always encodes progressive although the frames might be interlaced. I thought that shouldn't be a problem. Default configuration settings of the codec are:



I adjusted them for the capture to:



I've captured again a shorter 20-second sample of the original tape with Adobe Premiere, all default settings. Indeed Premiere doesn't allow you to select any capture codec. The audio seems it doesn't play well on VLC but it's fine in Premiere. Does this capture look better?

These are the tools I have setup right now for my workflow as I am on OS X as stated earlier.

[jagabo]

The new cap doesn't have the problem with the blended chroma channels. On the left is the new cap, on the right the earlier cap:



Note the blue tinge in front of the car on the right. That's the color from the next field blended with the color from the current field.

The video still has all the blended fields but that's probably on the tape. Probably a NTSC to PAL conversion. Or even worse, a PAL to NTSC to PAL conversion. Can you try recording a short segment of a native PAL video just to be sure? A movie would probably be good.

[themaster1]

Scan pattern= Alternate i believe that's the option for interlaced fields in your codec's options
Dc precison: 10 bit (if you have that option it's better)

[jagabo]

As I understand it, the alternate scan patter is more efficient for interlaced video but doesn't require it. In any case, it's worth a try. Still, I'd go with Premiere's uncompressed capture if you can afford the disk space.

[themaster1]

from mainconcept reference encoder help file:

Scanning Order: Specify one of two entropy scanning patterns which define the order in
which quantized DCT coefficients are run-length coded. Set to 1 for the alternate
scanning pattern or 0 for the zig-zag scanning pattern. The alternate scanning pattern is
considered to be better suited for interlaced video where sophisticated forward
quantization is not enabled. This setting can be specified independently for each frame
type (I, B and P). See ISO/

[Slennox]

Ok, I'll try next cap with the alternate scan pattern and see if that helps with the blending of the chroma channels or if it is a specific problem of this codec for this particular source. I would like to use an intermediate high quality codec between lossless and lossy as I plan to archive the cap. I read that this codec would be suitable for this purpose, but it looks like I'll need to look other options. Although I can afford a lossless capture for editing and encoding, I can't keep it for archive. It would took ages to upload a single cap in the cloud (~600 GB for a 3 hour cap!).

Anyhow, now I'm much worried with what jagabo said:

The video still has all the blended fields but that's probably on the tape. Probably a NTSC to PAL conversion. Or even worse, a PAL to NTSC to PAL conversion. Can you try recording a short segment of a native PAL video just to be sure? A movie would probably be good.

I couldn't cap a movie today sorry, but I'm pretty sure the source of these caps was a PAL broadcast (they're from an spanish TV channel). However I can see a sticker on my VCR (Panasonic NV-FJ625) that reads "NTSC P.B. ON PAL TV"... Does this means it is an original NTSC deck that can output a PAL_60 signal through the conversion you were suggesting? Every cap will have all the blended fields then, won't it?

A lot of issues related with my VCR. Maybe I'll need to invest on an S-VHS deck in the end and get rid all of these problems.

[jagabo]

I suspect "P.B." stands for "play back". Ie, it's a PAL deck that can play NTSC tapes as PAL60 -- pretty much standard for PAL VHS decks. Since you have 576 line caps your tape is probably PAL. But I suspect video was converted from an NTSC tape before being broadcast as PAL. I doubt the capture device or software converted PAL60 signal to a PAL cap, though that's not impossible. Definitely try some other tapes and see if some are free of the blended fields.

An S-VHS deck is no panacea. As you know, finding one in good working condition is hard and they're expensive. They don't work well with tapes recorded at lower speeds. The line TBC found in many can be helpful but they don't always work well. If you can't get clean caps of anything with your current VHS deck you might try a few other VHS decks -- maybe some friends or relatives will let you borrow theirs for a quick test.

[PuzZLeR]

An S-VHS deck is no panacea. As you know, finding one in good working condition is hard and they're expensive. They don't work well with tapes recorded at lower speeds. The line TBC found in many can be helpful but they don't always work well. If you can't get clean caps of anything with your current VHS deck you might try a few other VHS decks -- maybe some friends or relatives will let you borrow theirs for a quick test.

If I may add - I agree totally (as I've brought up in post #6)

These professional S-VHS decks, IMO, are not only overrated, they can be terrible. Yes, they are expensive, but I honestly don't care for the money as I did for the headaches of using one.

Aside from the electrical failures they are more prone to than other decks, any onboard TBC they may have can cause much more harm than good like color banding, extra jitter (yes, true), wrong field order, etc, which can be disastrous. The only thing "helpful" that I've seen these TBCs may do is on very minority tapes, like bringing out that odd detail in a dark scene, or improving upon a badly damaged tape (that is still not corrected anyway) - not worth it for the definite side effects and other artifacts. And it's not lonesome me that feels that other processing, like DNR, or sharpening, can melt or fry the video.

All you need, IMO, and at least for the very sake of minimalism, is a clean ordinay RCA/composite deck that can produce sharp detail and has good tracking, which can be expected from a good branded, simple, black, dedicated (no combo), 4-head, HiFi unit, along with a line based external TBC.

Yes, you do get side effects from a composite deck, but, as was demonstrated here with crosstalk, can be easily corrected with software. Same goes for color correction and noise. And any extra playback quirks, or any extra playback snow that a cheaper deck may produce, which is usually random, is also easily repaired with median methods.

[dphirschler]

[*]SCART to composite cable adapter

Avoid using composite, which is combining the signals into one line. S-video and component will keep the lines separate which will reduce/eliminate crosstalk.

Darryl

[jagabo]

Is it common for PAL VHS decks to output s-video or RGB at the SCART connector?

[Brad]

Although I can afford a lossless capture for editing and encoding, I can't keep it for archive. It would took ages to upload a single cap in the cloud (~600 GB for a 3 hour cap!).

A lossless capture in YUV rather than RGB shouldn't be anywhere near that big.

[PuzZLeR]

Although I can afford a lossless capture for editing and encoding, I can't keep it for archive. It would took ages to upload a single cap in the cloud (~600 GB for a 3 hour cap!).

A lossless capture in YUV rather than RGB shouldn't be anywhere near that big.

Yes, YUV would be, more or less, half the size of RGB. But even 600GB for RGB at full SD for 3 hours still won't be anywhere that big either.

Then again, forget this cloud nonsense. Just get an external drive for archiving this sort of stuff.

[jagabo]

Uncompressed RGB 4:4:4: 720 x 576 * 3 (bytes per pixel) * 25 (frames per second) * 3600 (seconds per hour) ~= 112 GB/hr.
Uncompressed YUV 4:2:2: only 2 bytes per pixel so 75 GB/hr.
Uncompressed YUV 4:2:0: only 1.5 bytes per pixel so 56 GB/hr

VHS is pretty noisy so lossless compression will get you about half those sizes with no loss of quality.

An hour of DVD quality stereo audio is less than 1 GB/hr. Container overhead is maybe 1 percent with AVI or MOV.

Be careful with long term storage and lossless codecs. Will Huffyuv, UT Video Codec, or Lagarith still be available on the computers we'll be using 30 years from now?

[Brad]

Be careful with long term storage and lossless codecs. Will Huffyuv, UT Video Codec, or Lagarith still be available on the computers we'll be using 30 years from now?

It's lossless; just batch re-encode to the new lossless codec every 15 years.

[PuzZLeR]

Be careful with long term storage and lossless codecs. Will Huffyuv, UT Video Codec, or Lagarith still be available on the computers we'll be using 30 years from now?

I agree to be careful, and yes, they will likely be replaced in 30 years time by better, but I don’t worry about it.

And I did give this some thought since much important content of mine, mostly from VHS, even from tapes disposed of, is archived in HuffYUV – an old, and even rather proprietary, codec no longer developed, maybe even theoretically dead, that has also been succeeded by several others better at compressing lossless.

Again, I don't worry about it, because lossless is lossless in any lossless format, and here's how I look at it:

I don't know how much better the algorithm can get for lossless video in reducing file sizes further, but it SHOULD get at least a bit better in the next 30 years. But the older codecs SHOULD be decodable on a computer, and there SHOULD always be a lossless current format available, and conversion to the new lossless codec SHOULD be possible, and conversion SHOULD be lossless.

I capped the "SHOULD"s to point out that I’d bet money on them, not that I was yelling at anyone.

However, on the contrary, if I am wrong about any of my "SHOULD"s, I also SHOULD sense such a disaster coming, and handle it.

I always say to keep the (digital) source, but with lossless, that rule is flexible and transferable. And if I can transfer it to a better lossless codec in the future, all's well.

I would worry more about archiving in a lossy format. Then again, you can always convert it to the latest and greatest lossless with the HUGE HDD capacities we are, and will be, getting.

EDIT: Oops. Had to go back to edit and change the word "PC" to "computer" because in 30 years time we may not have what is currently known as a "PC".

[PuzZLeR]

Be careful with long term storage and lossless codecs. Will Huffyuv, UT Video Codec, or Lagarith still be available on the computers we'll be using 30 years from now?

It's lossless; just batch re-encode to the new lossless codec every 15 years.

I'm putting it on my schedule.

But I think I'll be doing it somewhat sooner.