Red color bleeding in Avisynth ffmpeg edit of VHS capture

[rarend]

Hi all,

Following several posts, guides and trial and error I'm trying to digitize a VHS-recording from TV as best as I can using AviSynth and ffmpeg. I'm reasonably satisfied with the results so far
but now I am running into an issue that the red in the video is bleeding outside the areas in which it is supposed to be. I've added screenshots of this. It looks a bit like ghosting but then
only with the color red. Is this chroma bleeding? If not, what would you call it? And how can I counter it?

To capture the analogue video I use an old Sony camcorder that I put in between my S-VHS VCR and my PC using a high-end S-VHS cable and a high end fire wire cable. I do the initial
capture using Pinnacle Studio, to 720x576 DV avi, and then edit the initial result using AviSynth and ffmpeg, directly to mp4.

This is a screenshot of the dv avi:
http://imagizer.imageshack.us/a/img923/1381/kgDOPK.png

This is a screenshot of the same frame in the mp4 after processing:
https://imagizer.imageshack.com/v2/1280x720q90/921/P2IlUY.png

This is a WeTransfer link to the first 2+ minutes of the raw DV avi (457MB, no sound). The screenshot was taken at 2:05.480:
https://wetransfer.com/downloads/01e88c60d4f4a05cf79797081ca83b7920190116223404/9faedc6bdc980d064bfd9aa71fc67f8a20190116223404/fe8670
(Valid until January 23rd 2019)

If you look at the shoulders of the person in the screenshot of the processed mp4, you'll notice that the red is bleeding out above the shoulders, which it doesn't do in the DV avi.

Any idea what might be causing this?

I use the script below, in which I remove VHS flickering using Cnr2 and MDegrain2i2, de-interlace using QTGMC, remove additional noise using fft3dfilter and then sharpen the video using
LSFMod. After this I crop and resize the video to give it a 16:9 compliant height. I then use GradFun3 to smoothen transitions and finally I box the video to give it the 16:9 width that
corresponds with the height.

Thanks in advance for any suggestions and tips.

Script:

# 1. import plugins

LoadPlugin("E:\Program Files (x86)\AviSynth\plugins\Avisynth 32-bit Plugins\2.5x Plugins\mt_masktools-25.dll")
LoadPlugin("E:\Program Files (x86)\AviSynth\plugins\Avisynth 32-bit Plugins\RemoveGrain.dll")
LoadPlugin("E:\Program Files (x86)\AviSynth\plugins\Avisynth 32-bit Plugins\mvtools2.dll")
LoadPlugin("E:\Program Files (x86)\AviSynth\plugins\Avisynth 32-bit Plugins\nnedi3.dll")
loadplugin("E:\Program Files (x86)\AviSynth\plugins\fft3dfilter211\fft3dfilter.dll")
LoadPlugin("E:\Program Files (x86)\AviSynth\plugins\removegrain\RemoveGrain.dll")
LoadPlugin("E:\Program Files (x86)\AviSynth\plugins\removegrain\Repair.dll")
Import("E:\Program Files (x86)\AviSynth\plugins\LSFmod.v1.9.avsi")
LoadPlugin("E:\Program Files (x86)\AviSynth\plugins\mvtools-2.6.0.5\avstp.dll")
LoadPlugin("E:\Program Files (x86)\AviSynth\plugins\mvtools-2.6.0.5\mvtools2.dll")
LoadPlugin("E:\Program Files (x86)\AviSynth\plugins\dfttest-1.9.4\dfttest.dll")
LoadPlugin("E:\Program Files (x86)\AviSynth\plugins\dither-1.27.2\win32\dither.dll")
Import("E:\Program Files (x86)\AviSynth\plugins\dither-1.27.2\dither.avsi")
Import("E:\Program Files (x86)\AviSynth\plugins\dither-1.27.2\mt_xxpand_multi.avsi")
LoadPlugin("E:\Program Files (x86)\AviSynth\plugins\Chromashift\ChromaShift.dll")
LoadPlugin("E:\Program Files (x86)\AviSynth\plugins\WarpSharp\warpsharp.dll")
LoadPlugin("E:\Program Files (x86)\AviSynth\plugins\RemapFrames\RemapFrames.dll")
# Chroma Stabilizer for analog sources. It's a temporal filter and it's very effective against two chroma artifacts: Stationary rainbows and huge analog chroma activity.
loadPlugin("E:\Program Files (x86)\AviSynth\plugins\cnr2_v261\Cnr2.dll")

# 2. inline functions:

# MDegrain2i2 found on https://forum.videohelp.com/threads/308840-old-VHS-restoration/page2 correct old vhs tapes
function MDegrain2i2(clip source, int "blksize", int "overlap", int "dct")
{
Vshift=2 # 2 lines per bobbed-field per tape generation (PAL); original=2; copy=4 etc
Hshift=0 # determine experimentally
overlap=default(overlap,0) # overlap value (0 to 4 for blksize=8)
dct=default(dct,0) # use dct=1 for clip with light flicker

fields=source.SeparateFields() # separate by fields

#This line gets rid of vertical chroma halo
fixed_fields=MergeChroma(fields,crop(fields,Hshift,Vshift,0,0).addborders(0,0,Hshift,Vshift))

super = fixed_fields.MSuper(pel=2, sharp=1)
backward_vec2 = super.MAnalyse(isb = true, delta = 2, blksize=blksize, overlap=overlap, dct=dct)
forward_vec2 = super.MAnalyse(isb = false, delta = 2, blksize=blksize, overlap=overlap, dct=dct)
backward_vec4 = super.MAnalyse(isb = true, delta = 4, blksize=blksize, overlap=overlap, dct=dct)
forward_vec4 = super.MAnalyse(isb = false, delta = 4, blksize=blksize, overlap=overlap, dct=dct)

MDegrain2(fixed_fields,super, backward_vec2,forward_vec2,backward_vec4,forward_vec4,thSAD=400)

unsharpmask(60,3,0) #not sure whether to put this before or after the weave.

Weave()
}

# 3. open file

SetMemoryMax(700) # Optional line.
SetMTMode(3, 1) # See below for value X, could try 5 instead of 3 for non-standard source-filter/avisynth combinations
### SOURCE
source=AVISource("R:\NEWCAPS 20180326\raw dv avi\System Of A Down live Lowlands (20010825)\20180325 system of a down live lowlands (raw dv)_cropped_without_commercials.avi") # DGDecode_mpeg2source, FFVideoSource, AviSource, whatever your source requires

# 4. Force the field order

# This line is important as AviSynth usually guesses field order incorrectly! Change to AssumeBFF() if working with bottom-field-first video.
# Read more: http://www.digitalfaq.com/forum/video-restore/3729-qtgmc-how-deinterlacing.html#ixzz5BeJJyP7Z
# According to MediaInfo, the source is BFF
source2 = AssumeBFF(source)

SetMTMode(2,0)

# 5. remove flickering typical of VHS

# https://forum.videohelp.com/threads/308840-old-VHS-restoration/page2 correct old vhs tapes
# Only use chroma restoration for analog source material
chroma=source2.Cnr2("oxx",8,16,191,100,255,32,255,false) #VHS

#Set overlap in line below to 0, 2, 4, 8. Lower number=better, but slower
#For VHS, 4,0 seems to work better than 8,2. Most of difference is in shadows
#However, 8,0 is good enough and MUCH faster. 8,2 doesn't seem to make much difference on VHS.
output=MDegrain2i2(chroma,4,0,0)

# 6. deinterlace video - this outputs a video with double framerate

deinterlacedVid=QTGMC( output, Preset="Slow", EdiThreads=1 ) # Choose preset based on overall speed/quality you want. See below for value Y

# 7. remove additional noise

additionalDenoisedVid=fft3dfilter(deinterlacedVid, sigma=0.7, bt=5, bw=32, bh=32, ow=16, oh=16)

# 8. sharpen after denoising

sharpenedLSFVid=LSFMod(additionalDenoisedVid, strength=40, defaults="slow")

# 9. crop (must be even values)

# left 12 right 6 -> 702x576
# top 82 bottom 74 -> 702x420
cropped = Crop(sharpenedLSFVid, 12, 82, -6, -74)

# 10. resize

# After the cropping is done, resize to a near 1280x720 resolution (1204x720) and afterwards add borders to make the result truly 1280x720
resizedvid=Lanczos4Resize(cropped, 1204,720)

# 11. apply gradfun at the very end to smoothen gradiations

gradFunVid=GradFun3(resizedvid)

# 12. add borders to get 16:9 dimensions

borderVid = AddBorders(gradFunVid,38,0,38,0)

return borderVid

The version of LSFmod that I use is v1.9, which can be found here:
https://forum.doom9.org/showthread.php?t=142706

[LemMotlow]

Gosh. If you knew you had ugly noisy analog video and you'd have to do a lot of cleanup with Avisynth and ffmpeg and re-encoding, why did you capture to lossy compression-riddled DV????? Everybody knows that lossless is the way to go for this kind of work.

Too late now. You'll get plenty of answers and solutions, but you sure made it tough on yourself.
Try FixChromaBeeding (http://avisynth.nl/index.php/FixChromaBleeding) for starters to dull some of that red while you're still in YUV color. Then ChromaShift (http://avisynth.nl/index.php/ChromaShift) and MergeChroma(aWarpSharp2(depth=30)), for starters. That will still leave you with a way to go, but it's a beginning. Capturing noisy analog to noisy DV only added more noise than you started with.

[FranceBB]

According to your sample, I noticed that there's not just chroma bleeding, but some colors are completely off as well, so you might wanna play a bit with GamMac. It's a filter written by Stainless; I've used it in the past and I've got good results for chroma.
When it comes to chroma bleeding, though, I've seen many different approaches, including using WarpSharp only on chroma, but the result was never optimal.
As LemMotlow suggested, I would go with FixChromaBleeding as well. Besides, I noticed that you are using cnr2 which itself attempts to fix common analog artifacts, including chroma bleeding a little bit already.

[StainlessS]

If GamMac dont work well, maybe worth a try with Bernardd's Automatic white balance with RGBAdapt plugin :- https://forum.doom9.org/showthread.php?t=174411

[LemMotlow]

There's also a very basic problem with signal levels, a lot of illegal luma and chroma levels throughout. That's kind of a rudimentary principle with capturing, and I assume you know what crushed blacks and clipped brights are. Also some very visible chroma ghosting between interlaced frames. You might want to give us some specs on the capture hardware and software, which can help. Otherwise if you don't get a handle on proper capture, you'll work yourself to death and end up with garbage.

[rarend]

Thanks for your replies. There's a lot of interesting remarks in them that I'd like to discuss.
I haven't got time to do so right now, but I will definitely get back to you later on today.

I'm somewhat of a novice that is still learning how to do this by just giving things a try, so please bear with me when I do so.
But I'm very much convinced that your answers will help me a lot to understand how to do a capture like this in a more proper way.

[rarend]

Thanks again for your input LemMotlow, FranceBB and StainlessS. And sorry for not responding sooner, I was preoccupied with some personal stuff.

The main point I got from your posts is that I should improve my way of doing the initial capture. I should do it in a lossless format like LemMotlow suggests (which makes total sense) and I should perhaps take into account more the manufacturer of the camcorder with respect to the used codec. When done right issues such as the chroma bleeding, color shifting and illegal luma that you point out might become less prominent.

I still have the analogue VHS-tape and I'm very willing to capture the footage again to get better initial results. As for the filters suggested, I will surely check them out for situations in which I can' recapture.

For capturing directly to a lossless video format, is there open source / community software that you can recommend? I'm not familiar with any, nor am I with doing so using any of the programs I use.

For this VHS-capture via a DV camcorder I used Pinnacle because I had it installed and because it is very easy to use. I figured that since the source was DV avi, using DV avi also as the capture format within Pinnacle made sense. But indeed, it is not lossless. Afaik Pinnacle can't capture in a lossless format directly.

Would capturing to raw AVI perhaps also be a better alternative still?

As for the hardware I use to capture, I would very much like to stick to the camcorder setup instead of a capture card. I find the visual output quality very good when compared to the analogue source and I find the setup very convenient. Also, the output resolution is higher than the native resolution of VHS so with regard to the resolution the setup is more than sufficient.

And while this particular troublesome footage is from a private VHS recording, I have had excellent results with capturing retail VHS-tapes this way. I got the idea for using a camcorder from VideoHelp (https://www.videohelp.com/dvanalog).

The hardware specifics of my setup are as follows:


VCR: JVC S-VHS HR-S6711
Camcorder used for digitizing via analog-to-digital passthrough: Sony DCR-PC120E PAL
PC firewire card: VIA 1394 OHCI Compliant Host Controller
OS: Windows 7 Ultimate (Intel Core i7-2700K @ 3.5 GHz / 16GB RAM)
VCR to camcorder video cable: Vivanco PROFIGOLD Interconnect Cable (S-VHS)
VCR to camcorder audio cable: Vivanco High Performance Interconnect Cable (L/R tulips to mini jack)
Camcorder to PC firewire cable: VW-1 80⁰C 30V SPACE SHUTTLE-D IEEE 1394 (USB mini to 1394 firewire)


I use a Vivanco cleaning cassette before each run.

Might there be a specific Sony codec preferred to capture and do you know where I might find it?

And as for the chroma, luma and color issues, do you think capturing in a lossless format with better software and perhaps using a Sony codec would prevent such issues? Or might they perhaps partially be caused by other factors as well?

Thanks in advance for your replies.

[LemMotlow]

When done right issues such as the chroma bleeding, color shifting and illegal luma that you point out might become less prominent.When "done right", whether you use lossless or not, you need a means of controlling the input signal. A proc amp will do it, but most people use capture software that allows them to hook into the proc amp controls that come with capture drivers from your hardware. Most DV capture devices and methods have no such controls. That's because they're really designed for original digital sources that aren't plagued with the kinds of wild voltage and levels variations that characterize analog media and playback devices. The idea behind lossless is that (a) you avoid the first step of image degradation by avoiding DV compression and chroma loss, compression artifacts, and added artifacts that result from analog noise, and (b) besides not having DV disadvantages to clean up you avoid further degradation by using lossless compressors for each subsequent stage of post processing, until you make your final lossy encode to your desired output format(s), which would be the very last step in your total processing chain.

For capturing directly to a lossless video format, is there open source / community software that you can recommend? I'm not familiar with any, nor am I with doing so using any of the programs I use.VirtualDub, AmarecTV. Those are the most popular. If you've never seen either, VDub represents a completely different world of detailed control over more aspects of video capture than you ever imagined. Likely you'll learn a great deal more about video than you would with DV and Pinnacle.

For this VHS-capture via a DV camcorder I used Pinnacle because I had it installed and because it is very easy to use. I figured that since the source was DV avi, using DV avi also as the capture format within Pinnacle made sense.Your source, as you described it, is not DV AVI. It'sanalog tape. Completely different systems. Most of us outgrew Pinnacle many years ago. And DV source itself is never captured. It's transferred 1:1 with Firewire and is never re-encoded as a capture. That's two lossy strikes against your so-called "quality" from the very start. However, if convenience and the push-button route is major priority, you're better off with DV and Pinnacle. Priorities of quality will have to wait for another day.

As for the hardware I use to capture, I would very much like to stick to the camcorder setup instead of a capture card. If your camcorder can't output analog, you can forget about lossless if you continue using it. There are a few good capture cards around that are optimized for analog source and can be used with software mentioned above for lossless capture. The Diamond multimedia VC500 has been a favorite for many years, is still made today, and sells for a fraction of of its original price. Or you can go full hog and try for building an All in Wonder System, which of course would be a huge headache and expense. Either way, lossless will run circles around DV when you get to your final output. Then again, maybe your camera has an analog pass-thru that you can use for lossless capture.

I find the visual output quality very good when compared to the analogue source and I find the setup very convenient. Also, the output resolution is higher than the native resolution of VHS so with regard to the resolution the setup is more than sufficient.No, your DV output has digital artifacts and noise that your tape doesn't have, besides losing about 20% of the original data and 50% of the chroma resolution. How is that "better?" VHS is normally captured lossless at 720x480 with VHS' original color resolution at YUY2 (4:2:2: instead of YV12 4:2:0 or 4:1:1). And you end up with only one lossy encoding step instead of two or even more with additional processing steps. With DV if you use color correction, denoising, special effect overlays, or any other form of image modification, your DV video is lossy re-encoded in its entirety for even more accumulative degradation and noise. DV audio is also a less precise YouTube-level 44.1KHz, whereas VHS is usually captured at uncompressed DVD-compatible 48KHz.

And while this particular troublesome footage is from a private VHS recording, I have had excellent results with capturing retail VHS-tapes this way. I got the idea for using a camcorder from VideoHelp (https://www.videohelp.com/dvanalog).. Yes, I'm very familiar with the noisy DV crowd over there. And that's where you were taught to go through two lossy capture stages with Pinnacle? Sounds like the blind leading the blind.

Looks like your camera has what you need for lossless. but it won't last long if you keep using that abrasive "cleaning" tape.


Might there be a specific Sony codec preferred to capture and do you know where I might find it?Again, capture VHS to 720x480 using YUY2 color and 48KHz PCM audio. The SONY DV codecs are not lossless. DV is always lossy, period.
-Best CPU efficiency usage, for lossless YUY2 real-time lossless compression: huffYUV 2.1.1 (). Comes in 32vbit + 64bit., use 32-bit for VirtualDub or AmarectTV capture. https://www.videohelp.com/software/HuffYUV
- 2nd best, slightly higher CPU usage but slightly smaller compresses files: Lagarith lossless compression. Easier install for Win7 and Win10 (if using either, you have our condolences). https://lags.leetcode.net/codec.html
- There is also the more geeky ut video coded and magic, optimized for digital source, Make up your own mind there. Media players can all handle huffyuv and Lagarith. Some of the other codecs are not so widely supported.

And as for the chroma, luma and color issues, do you think capturing in a lossless format with better software and perhaps using a Sony codec would prevent such issues? Or might they perhaps partially be caused by other factors as well?They're caused by many factors, from your playback unit to the condition of the tapes, to your levels and saturation controls.....it's an endless list of variables. How much control do you have over those variables with the DV system you're using? None.

Time for the DV crowd to chime in with their usual, which we've all heard and seen since Windows 3.1. Have at it. There's, nothing more to add here. Offhand it sounds as if you're looking for ease and convenience, an absence of any need for cleanup, and a perfect capture that looks like DVD instead of like VHS tape. It won't happen. Without work, VHS always looks like VHS. DV just makes it look a little worse and makes it easier to blame the source for the results.

Good luck. My guess is that you'll stick with DV. It's not better. It's just easier.

[rarend]

Thanks LemMotlow for your extensive and informative reply. I am now looking into the Diamond multimedia VC500 video card you recommended. I still have some questions though.

For this VHS-capture via a DV camcorder I used Pinnacle because I had it installed and because it is very easy to use. I figured that since the source was DV avi, using DV avi also as the capture format within Pinnacle made sense.Your source, as you described it, is not DV AVI. It'sanalog tape. Completely different systems. Most of us outgrew Pinnacle many years ago. And DV source itself is never captured. It's transferred 1:1 with Firewire and is never re-encoded as a capture. That's two lossy strikes against your so-called "quality" from the very start. However, if convenience and the push-button route is major priority, you're better off with DV and Pinnacle. Priorities of quality will have to wait for another day.

In this context, by source I meant the output of the cam corder to my PC. But I probably misunderstood the concept of the pass-through. I thought that by using a DV cam corder pass-through, the cam corder would convert the analogue signal to DV avi quite like as if it would be recording, and output that DV avi signal to my PC as if It’d be playing a DV cassette.

Is it correct or incorrect to state that the cam corder does output DV avi? Or, if not analogue, is it more of a format-less on the fly digital interpretation of the analogue input, which is inaccurate? In either case I take it from your reply that the cam corder doesn’t do a good job because it doesn’t deal with interpreting the messy analogue VHS-signals very well.

I will look into whether the cam corder is capable of outputting analogue, just to check.

For capturing directly to a lossless video format, is there open source / community software that you can recommend? I'm not familiar with any, nor am I with doing so using any of the programs I use.VirtualDub, AmarecTV. Those are the most popular. If you've never seen either, VDub represents a completely different world of detailed control over more aspects of video capture than you ever imagined. Likely you'll learn a great deal more about video than you would with DV and Pinnacle.

I'm playing around with AmarecTV 2.3.1 now together with Lagarith, because many people recommend it for capturing to losless YUY2. However I haven't been able to capture with it yet, I keep getting 'Can not support colorspace by video codec'. It shows the visual throughput and Lagarith is recognized fine so I'm still looking for the cause, maybe it's because of the cam corder.

I've used VDub often for re-encoding but it always eluded me that you can also capture with it. I will give it a try for sure. Quite a few people on forums prefer AmarecTV though, saying VDub causes audio sync issues and produces ‘flaky’ videos.

Offhand it sounds as if you're looking for ease and convenience, an absence of any need for cleanup, and a perfect capture that looks like DVD instead of like VHS tape. It won't happen. Without work, VHS always looks like VHS. DV just makes it look a little worse and makes it easier to blame the source for the results.

As for the convenience, for me this is mainly about doing the initial capture. I don’t mind long re-encoding sessions afterwards (even days for a single one). But getting out my VCR and hooking everything up is pretty invasive to my living space so I'd like to keep that time as limited as possible.

But well, once I have a capture card that wouldn’t necessarily differ a lot from using the cam corder. A long time ago I tended to use a Hauppage TV card to capture, but that always was a hassle and the quality disappointing. But a card like you suggested might be less troublesome.

There are a few good capture cards around that are optimized for analog source and can be used with software mentioned above for lossless capture. The Diamond multimedia VC500 has been a favorite for many years, is still made today, and sells for a fraction of its original price.

This card I take it is a genuine card and not this
Diamond VC500 USB 2.0 One Touch VHS to DVD Video Capture Device (https://www.diamondmm.com/product/diamond-vc500-one-touch-video-capture-edit-stream/)? I’ll search for a card some more because the device I linked looks rather cheap with low-quality connectors and I doubt it is the one you mean. But of course maybe I am mistaken.

Looks like your camera has what you need for lossless. but it won't last long if you keep using that abrasive "cleaning" tape.

The cleaning tape is for cleaning the heads of my VHS, because my understanding is that old tapes can leave a lot of dirt and tape particles on those heads. Should I not use it? I have to put three drops of cleaning fluid in the cassette, play it, and then wait for a couple of minutes before playing an actual tape. I always wait ten minutes just to be sure so that all leftovers of the fluid have evaporated.

For as long as I don’t have a capture card I will continue to hobby some more with the cam corder. But when I get one I will get back to you with the results. Thanks again!

[Emulgator]

Capturing to 10bit 4:2:2 uncompressed .avi via USB3.0: Blackmagic Intensity Shuttle USB 3.0
Picky, buggy drivers, but I found a working combination.

https://forum.doom9.org/showthread.php?p=1847255#post1847255

[rarend]

Capturing to 10bit 4:2:2 uncompressed .avi via USB3.0: Blackmagic Intensity Shuttle USB 3.0
Picky, buggy drivers, but I found a working combination.

https://forum.doom9.org/showthread.php?p=1847255#post1847255

Thanks for your recommendation and the work-arounds in the other thread, I'm looking into it.

It seems highly recommended by people that manage to get it to work properly, but according to this review on Amazon (https://www.amazon.com/gp/customer-reviews/RTR9LQ88D9XF2/ref=cm_cr_getr_d_rvw_ttl?ie=UTF8&ASIN=B003WSQTWU) it only works on X58 motherboards.

My mobo is an Asus P8P67 Deluxe which has an Intel P67 chipset, are you using it on a pc with an x58 chipset yourself?

[LemMotlow]

Capturing to 10bit 4:2:2 uncompressed .avi via USB3.0: Blackmagic Intensity Shuttle USB 3.0
Picky, buggy drivers, but I found a working combination.

https://forum.doom9.org/showthread.php?p=1847255#post1847255
OMG not that BM crap again!
10-bit5? for analog tape? LOL!

Dream on, fellas. All that money and hype for so little. Some myths never die.

[Emulgator]

Lem: Well I was beat into 4:2:2 10bit uncompressed here at doom9 by manolito, grinning,
me coming from DV-AVI, being content with it for years when having to capture crappy VHS,
knowing that most VHS chroma is off anyway and being less than 1/4 resolution,
accepted 4:2:0 and compensated any chromashift in Avisynth.
Back then I had to spend 300€ or so on the Canopus ADVC-300 for finally proper-looking and block-free DV-AVI.

Now scrutinising the first 10-bit uncompressed VHS results from my overhauled S-VHS NV-HS950,
finally seeing every chroma align properly:
Even with the worst sources, Postprocessing from 4:2:2 gets a better picture.
As a matter of fact I will go through all my VHS captures again, if time allows and sources are still there.
To me the BM was worth it, 198€ spent well. Histogram tells it, densely populated after colour correction.
BTW, I do that professionally, so why stick to something that can be improved...

Still agreeing with the wrath about the unfinished product here and there.

rarend: I just read the amazon report.
The writer was referring to PC tower mainboards only,
so on Nov 23, 2010 was of the opinion exclusively X58 would do.
Good advice at given time within that range I guess.
I am with him in that regard that only Hi-end and onboard Chips will offer full USB 3.0 bandwidth on negotiation.
And it takes the proper driver to acknowledge the BM bandwidth negotiation handshake.
Nevertheless there will be other Hi-end mainboards that employ these TI TUSB7xxx bridges.
As far as I know Renesas will not do, only implement a part of bandwidth (or are the drivers not releasing ?)

The motherboard I use for that is a notebook Clevo P570WM,
the underlying USB 3.0 chipset is a TI TUSB7340.

P.S. An Expresscard34 to USB 3.0 Adapter (Renesas) did not give enough bandwidth on negotiation
(visible only on first connection for a brief moment) and BM did refuse to work.
See post I linked to first.

[johnmeyer]

As a matter of fact I will go through all my VHS captures again, if time allows and sources are still there.I would love to see a few seconds of your original DV captures compared to what you are getting now. Even a few still frames would be useful. I've never seen a really good comparison showing what sort of differences you can get with a full-color, better compression capture chain compared to DV. I fully understand DV's limitations, but have always wondered when starting with VHS, 8mm, or Beta whether those differences would amount to much compared to the limitations in the source media. What I do know -- and you have just described it -- is getting the better capture chain working is often impossible for all but the most dedicated and informed user (that would be you in this case), whereas DV capture is pretty easy to set up and get right.

[LemMotlow]

No, VHS to DV doesn't get it "right", it just gets it "easier". Since you've admitted that with all of your expertise you've never been able to achieve a lossless capture of anything anytime anywhere using gear optimised for analog source, this doesn't offer us much. I'm always namuseds at the failed-at-lossless crowds, they never fail to jump in with "advice" about something they have no experience with. They want more "comparisons", but people have been posting tests and comparisons since Windows 3.1 and as far back as I can remember since Bm and the Canopus sites starting muscling in on analog capture. Why would one think that if the same thing is repeated, somehow the outcome will be different?

But you're correct. if your skill level at capture isn't so great and your overall knowledge of video properties and restoration islikewise, you're better off with DV capture. Why ride a proper bike when you can use the three-wheel trainer and learn so much less? Then the lossy crowd can advise how to clean up the bleed as well as the crushed darks and blown-out brights.

[johnmeyer]

No, VHS to DV doesn't get it "right", it just gets it "easier". Since you've admitted that with all of your expertise you've never been able to achieve a lossless capture of anything anytime anywhere using gear optimised for analog source, this doesn't offer us much. I'm always namuseds at the failed-at-lossless crowds, they never fail to jump in with "advice" about something they have no experience with. They want more "comparisons", but people have been posting tests and comparisons since Windows 3.1 and as far back as I can remember since Bm and the Canopus sites starting muscling in on analog capture. Why would one think that if the same thing is repeated, somehow the outcome will be different?

But you're correct. if your skill level at capture isn't so great and your overall knowledge of video properties and restoration islikewise, you're better off with DV capture. Why ride a proper bike when you can use the three-wheel trainer and learn so much less? Then the lossy crowd can advise how to clean up the bleed as well as the crushed darks and blown-out brights.I have never once seen any results posted from you. Just endless attitude and bitterness. I've posted hundreds of scripts and countless videos. I try to help people, not berate them.

[Emulgator]

I try to help people, not berate them.
This is what I am trying too here, thanks johnmeyer, for your research, countless suggestions and precious time here and in other forums !

I shall prepare a comparison test across a few source tapes/players/cameras/systems/capture variants I have here,
time permitting.
At the moment it is a batch of 40mg tiny coil pairs I have to glue together, strip X-},
and glue in place, solder, measure and hope my eyes hold sight...

rarend: I vaguely remember to have bought & tested a few (3?) DV-AVI capture boxes as these became available.
Their differing results were straight dependent of the actual codec implementation, not the fact that it would be DV.
One did it right !
(The DV 4:2:0 chroma reduction being acknowledged beforehand and not further discussed withing this post)
I had to drop 2 of them, one was the Pinnacle.
Those both had poor hardware implementations of DV compression, giving visible blocks.

I seem to remember that a NEC DV chipset used in Canopus ADVC300 did it properly. (have to open it again)
(300 for me because of the added tweaking possibilities, the ADVC50 and ADVC100 used the same DV hardware encoder device I guess.)

Next faults came in by poor DV decoding.
So system players did DV chroma interpolation wrong.
I remember to have seen a player plainly repeating YV12 chroma samples for both neighbored fields.
Then came fixing the wrong chroma kernel, fixing of wrong chroma placement.
Then I found NLEs that used DirectShow (EditStudio) to deliver blocks.
Well, the 8bit-engine did its damage too, but out came blocks that the source did not have.
Reason: The implemented DV algo in XP quartz.dll indeed added blockiness.
I went to Vegas and finally that issue was solved. Native decoders.
(Later an quartz.dll update came in and with Win7 quartz.dll I could not see these blocks anymore.
Me guessing here: To avoid DV patent liabilties with Sony and to be on the safe side the MS implementation came out imperfect ?
I remember the Apple side of things (Quicktime) had the same quirks around some codec implementations,
DV being one of them, besides intentionally re-mapping sample values.
Later silent updates fixed that...Patents expired ? Could be that simple.
I did not research around those timings, somebody else may.)
Some software DV implementations come into my mind: Sony, Mainconcept, Microsoft, Cedocida, FFmpeg, Quicktime
And DV hardware: Sony, NEC, and some more.
Not all DV implementations get everything right, but it is well possible now.

Bottom line: If DV (non-pro): Avoid the Pinnacle Box and related DV hardware encoders.
(Plug-in Stick MPEG-2/MPEG-4 hardware encoders are not worth considering within the scope of this thread)

If the source chroma is poor and/or 4:2:0 anyway and one is content with 4:2:0 DV (non-pro)
and one tolerates the CUE with interlaced YV12 one may find a canopus box for cheap
(or use a DV camcorder like Sony PC-100, or the mentioned PC-120 in passthrough, same quality codec-wise.
Which the OP did.)

If one does not tolerate the CUE because he wants to get best chroma out of a interlaced source that is worth 4:2:2
and the budget allows for 200USD/€ AND USB3.0 is fully implemented on your Motherboard:
Why not have access to a 10-bit 4:4:2 uncompressed intermediate ?
You will not want to miss once you got it. Huge files of course, but we don't complain about that.
The system ate money, the learning curve ate precious lifetime already.
So with the invaluable help of a community which helps to develop Avisynth and its filters and scripts,
we intend to chain fine sampling, preprocessing,
(For interlaced sources I go the QTGMC route to double framerate almost everytime,
and now we are blessed with fluid and non-leaking chroma)
editing, compositing, postprocessing, and after all that effort: Quality was our goal, I think.

P.S: Looked at the OP's captured .avi file once again: Just viewing in MPC-BE now
Since the OP's chain is reported to have delivered good results:
And while this particular troublesome footage is from a private VHS recording, I have had excellent results with capturing retail VHS-tapes this way.
I seem to be barking up the wrong tree here...
The source simply seems to have been recorded by/rerecorded from a poor deck.
To me it simply looks like it would be a 2nd or 3rd generation loss caused by 99€VHS to 149€VHS (to 199€VHS?),
going the composite route maybe via SCART or even worse:
PB 1st Deck PB -> Composite -> UHF modulator -> CH39 Antenna output ->
REC 2nd Deck Antenna input -> CH39 Demodulator -> Composite -> REC
so the chroma bleeding faults should be already woven-in.
Well, DV25's YV12 is adding to that by smearing chroma just across one more line.
Here I won't blame the OP's capturing chain exclusively anymore...
Playback Paused: the interlaced combs looks suspicious while reds are on the drumset !

Now Avisynth Fieldcheck in Double-PAL 50,000fps:
Within a frame 2 consecutive fields show 2 advancing moments in time.
Next frame, first field shows the same moment in time.
Now a simple temporal fieldshift mends that and delivers the 2 temporals field in one frame.
Still the content doesn't match spatially, jagged edges.
I hope to see the untouched capture here... TMF9, NL under that that Logo 1996..2002 ?
TMF9 Logo and content don't match spatially. When the logo is round, the content is jagged.
Either the shooting was in 25p and on editing shifted by one field or something went indeed wrong afterwards.
If I fieldshift temporally I get chroma field bleeding across edit borders.
Here we cannot have 4:2:0 anymore.
This develops into a nice showcase where capturing in 4:2:0 DV-AVI adds damages before repair.
A 4:2:2 capture would definitely help here !
Can be mended, no time now...
Quickglancing script: I wouldn't have come up with that one. OP, please report ;-)

[johnmeyer]

This is what I am trying too here, thanks johnmeyer, for your research, countless suggestions and precious time here and in other forums !Thank you very much for that. It helps a lot. I don't usually don't get bothered by forum flotsam, but once in awhile, late in the day, I let it get to me. I guess I'm getting too old for this.

Not all DV implementations get everything right, but it is well possible now.Your long discourse on DV is spot on and reminds me of something I used to point out: all DV is not created the same. When I first got started, not long after DV changed the entire video editing scene by letting you use common hardware to edit video (no SCSI RAID arrays required), I was using Pinnacle Studio and the big issue was the Microsoft DV encoder built into Windows which was absolutely pathetic. Everyone was complaining, and half the posts in the Studio forum were about how to replace it.

When I upgraded to VideoFactory (the little brother of Vegas) and then upgraded to Vegas, everyone was marvelling at the quality of the Vegas DV codec. One guy actually did a test where he made a copy of a copy, making a trivial change so the video had to be re-compressed. He did this 100 times and then compared the 100th generation to the original. They were visually almost the same. Here's a link to that test from back in 2003:

Multi-Generation DV Test (https://www.vegascreativesoftware.info/us/forum/multi-generation-dv-render-test--22807/)

The MainConcept DV encoder, which is what I use for other apps (because the Vegas encoder is not installed in a way that can be used by other apps) is almost as good.

So when people start yapping about DV's compression artifacts, I no longer argue, because of course it does have artifacts, like any other lossy compression, but they are not as horrific as some of these uniformed people claim. My guess is that most of the people who complain are not using a professional encoder.

This same thing is true of other encoders: MPEG-2 quality from Procoder or MainConcept is vastly better than from freeware versions. H.264 is actually the other way around: the freeware encoders in Handbrake and MeGUI seem to be better.

I've spent countless hours comparing all sorts of codecs, restoration scripts, and other things that change video, always looking for artifacts, both within each frame as well as temporal artifacts between frames. Some things hit you between the eyes and are to be avoided. DV compression artifacts, when done with a good DV encoder are not that big a deal compared to all the other problems you see in video.

[LemMotlow]

I have never once seen any results posted from you.
Then you haven't looked.

[johnmeyer]

Then you haven't looked.Actually I did, but if I missed them, why don't you provide a link?

[LemMotlow]

Actually I did, but if I missed them, why don't you provide a link?

Offhand, from what I can readily recall on short notice, from here and elsewhere:
https://forum.doom9.org/showpost.php?p=1641322&postcount=3
https://forum.doom9.org/showpost.php?p=1686972&postcount=16
https://forum.videohelp.com/threads/371909-How-to-adjust-the-brightness-and-contrast-of-a-video-in-Sony-Vegas#post2390907
https://forum.videohelp.com/threads/365289-Rookie-Help-Getting-Rid-of-Artifacts-General-Quality/page2#post2329439, continued here: https://forum.videohelp.com/threads/365289-Rookie-Help-Getting-Rid-of-Artifacts-General-Quality/page2#post2329583

I don't post that often these days. I usually figure out solutions for my own work by searching and reading posts by people who know what they're doing. People such as you, on a few occasions. Unlike you, I don't post my unfounded opinions on things I have no experience with.

Video cleanup is being supplanted by the "conversion" craze, whereby decent digital video originals are screwed up (aka "reformatted", "upsampled", etc. ) to look like the bad VHS of yesteryear played on cheap Vizio tv's and YooToob. Doesn't interest me. Or, to take a modern example, bad analog-to-DV caps that aren't cleaned up at all. What do I learn from that? How not to work with video. Learning what not to do is a valuable tool.

[johnmeyer]

Thanks for the links. I stand corrected.

I also love the new cheap shot. You just can't help yourself. <sigh>

[rarend]

Hi guys, I've read your replies with great interest and I will get back with a more in-depth reply soon.

Meanwhile, I've ordered both a Blackmagic Intensity Shuttle USB 3.0 (€199) as well a Diamond VC500 (€29 plus €17 shipping) so I can compare them myself. The Diamond will take two weeks to arrive from the US but once I have them both and I find time to capture I'll posts samples of each of them.

[rarend]

@Emulgator,
I've gotten my Blackmagic Intensity Shuttle USB 3.0 to work on three different systems but on each of them I'm now facing the black flickering (https://www.youtube.com/watch?v=ccR-J8mwH6k) issue that is being reported by many users of the device while capturing analogue VHS.

One system is my pc with the Renesas chipset and two are HP laptops with an Intel USB 3 eXtensible Host Controller (one of which is a recent HP Z Book).

A comment underneath this (https://www.youtube.com/watch?v=VZY_gtp9Rak) Youtube video says that the reason for the flickering is that the reels of a VCR spin at fluctuating speeds, which the Blackmagic Intensity Shuttle can't deal with.

Two possible solutions for this problem are given:
1. Get a VCR with a built-in Time Based Corrector, or
2. Pass the analogue signal through a DVD player (via scart), or through a Camcorder like I used to do, and then into the Blackmagic Intensity Shuttle.

Before I decide to invest into option 1, I have a few questions about option 2 that I'd like to know your opinion on.

- Quality-wise, would there be any advantage to using a pass-through via scart over a pass-trough via a cam corder?

- Would my capture process benefit in any way from hooking up the shuttle in between my camcorder (or a DVD player) and my pc, compared to connecting the camcorder to my pc directly via firewire like I did before?

@LemMotlow
I haven't received the Diamond VC500 yet, but I'm still planning to give that a try as well.

[LemMotlow]

LOL! I figured you'd be another sucker for Black Magic's marketing hype. No surprises here. After you figure out the scanline and frame timing problems that others have always known about, figure out how to control analog input signal levels for legal digital luminance values and avoid dropped frames. Let us know and post the results. The glitches you have now were solved by analog source capture hardware a long time ago.

The analog community has advised against BM and their lookalikes for decades. It's another case of the triumph of hype over experience. All that money and effort, and the captures still look like bad tape from 1995. We've seen it all before. Many times. Many, many times.

You need a line-level tbc and a frame-lvel tbc. Analog and digital inputs are not the same things, and neither are capture cards made for analog or for digital sources.

[Emulgator]

rarend, you need a S-VHS deck with TBC, and one that doesn't smooth too much.
The non-S-VHS decks (Sony up to SLV-E1000) I tried brought chroma 1 line offset, and/or were too smoothened.

Here I am using a Panasonic NV-HS950 which has TBC and 3DNR built-in.
Not having been into VHS at all I paid whopping 1699DM for it back then and it slowly became worth it...today.
I have its service manual here, and was inside for repairing loading gear pulley.
It is a huge bergwerk, 136 pages...maybe someone likes to read up here in this thread ;-) ?
Page 65 of the block schematics shows in section 3-18 referring to 3-51 ..3-54
the max what one can do with poor old VHS.
The video processing signal chain is an-in-all dream team:
(Head amp and following usual processing omitted for clarity)
IC2: Motorola SC371024CFU S-VHS Y processor
MN are Matsushita ICs (Panasonic is only a brand name under Matsushita):
IC3: MN67152 Video 8-bit ADC/DAC.
IC4: MN7A097B2C Digital 3D Y/C Separator.
It has a internal line memory interface talking to chip-internal line memory.
Further it contains a frame memory interface using 8 address lines
and sending/receiving 8bit lines to/from
IC7: MN47V77S (256KiBx8 external memory)
IC6: TI F432536PH Time Base Corrector with Burst Replace and Sync Replace.
Just found one for sale. ~53€ at the moment

All these chips are discontinued, with the demise of analog videotape there was no reason anymore to make such expensive rarities in sufficient quantities.
So the chance of such vintage & elaborated video signal chain being contained in recent and affordable BM is rather.. well... sparse ?
It is called Blackmagic for a reason ;-) I haven't opened it yet. But now I am curious. I will now.
Best is to let a dedicated hi-end S-HVS deck take care of that delicate task.

This is what you want.
It keeps all detail, including graininess from source/tape.
Later you can avisynth it all away if you like so.
Additionally I experienced the 3DNR helping with full-frame flicker.

- Quality-wise, would there be any advantage to using a pass-through via scart over a pass-trough via a cam corder?
SCART? I can't talk about that.
As even the most expensive SCART connectors continued to pop out of their receptacles
I got rid of them, didn't use it at all.
If you fare better with a deck delivering composite or component to capture:
You got to try and find out which device has the better filters implemented, deck or capture card !
With a mini-DV camcorder the black flicker might go away, see TBC. Sony's mini-DV Camcorder have a perfectly working TBC built-in, IMHO...But I what happens during the conversion ? You might trade in 4:2:2 chroma and get back 4:2:0 I guess. And it is 8 bit to start with. The shuttle would be unnecessary because the mini DV did the (down)conversion already.
- Would my capture process benefit in any way from hooking up the shuttle in between my camcorder (or a DVD player) and my pc, compared to connecting the camcorder to my pc directly via firewire like I did before?
If the shuttle does not apply TBC then I see no benefit of inserting if you intend to capture per DV-AVI anyway.

P.S.: If all this is for your fieldshifted 25p sample: You need to capture 4:2:2 and keep it such until it hits Avisynth !
No 4:2:0 DV-AVI in-between !

And this is why the S-VHS deck with line-and-frame TBC becomes unavoidable and irreplaceable anyway.

[Emulgator]

Opened my BM Intensity Shuttle:
HDMI I/O: AD9889B HDMI1.3 165MHz: ok
YCbCr In: ADV7604 Triple 12bit Video ADC 170MHz: nice
YCbCr Out: ADV7393 Triple 10bit Video DAC 216MHz: nice
CVBS In: ADV7180 10bit ADC, 8 bit Output only, Mini-TBC... Ah there it is: Mini: line length tracking only !
Indeed, something missing to be prime.
CVBS Out: ADV7393 Triple 10bit Video DAC 216MHz: nice
Processing: Xilinx Spartan FGPA XC3S1200E
So CVBS In is to be avoided, I did so anyway.
And the YCbCr Ins demand a Timebase-stabilized input which I can supply via the NV-HS950

[WaxCyl]

http://www.digitalfaq.com/forum/video-restore/6551-leitch-dps575-vs.html

Hi Emulgator. Latreche34 at the end of my thread (above) said he may have a Brighteye BE75 for sale.
This device has a full frame TBC, but also allows the the signal to remain digitised after time base correction and can output 10bit 4.2.2 lossless (with 12 bit internal processing)
I had many interior home recorded Hi-8 videos that were very dark and noisy. These type of video clips benefit greatly from 10 bit analog capture.
10 bit video has 1024 grey levels (cf 8bit 256 levels). Capturing analog at 10bit resolution therefore eliminates banding when large adjustments are made to levels.
Your entire processing chain can benefit from having a higher bit depth source. You would also need a Blackmagic SDI capture card.

[Emulgator]

Many thanks for pointing to the BE75.
Just now saw the gutshot: comparable to the BM Intensity Shuttle, but: Plus memory !
I see 4 chips around the Xilinx Spartan which could make a sufficient frame memory.
Now SDI..., well the OP should know what's best for him.

[LemMotlow]

http://www.digitalfaq.com/forum/video-restore/6551-leitch-dps575-vs.html

Hi Emulgator. Latreche34 at the end of my thread (above) said he may have a Brighteye BE75 for sale.
This device has a full frame TBC, but also allows the the signal to remain digitised after time base correction and can output 10bit 4.2.2 lossless (with 12 bit internal processing)
I had many interior home recorded Hi-8 videos that were very dark and noisy. These type of video clips benefit greatly from 10 bit analog capture.
10 bit video has 1024 grey levels (cf 8bit 256 levels). Capturing analog at 10bit resolution therefore eliminates banding when large adjustments are made to levels.
Your entire processing chain can benefit from having a higher bit depth source. You would also need a Blackmagic SDI capture card.

Great. You and Emulgator and latreche34 can show the OP (and the rest of us) how to correct the crushed blacks and blown-out brights with that BM hookup. And for the OP's original problem in the title of this thread, you can show him (and us) how to clean up the chroma shift and bleeding colors.

It should be very interesting.

[rarend]

@Emulgator
Thanks, greatly appreciate your detailed answer!

I did start this thread thinking that there might be just a relatively simple solution to the bleeding in the sample. But I find it very informative how the discussion has broadened to preferable setups, I'm learning a lot from it. And this includes some of the flaming ; )

Capturing VHS is something I really enjoy, especially when it is rare material. So if changes to my setup (within limits and budget) enable me to do so better, I am keen to give them a try. And hey if in the process this prevents the bleeding then we're all still on topic.

I saw that your Panasonic is still being sold second-hand for reasonable prices, I'm considering getting one.

P.S.: If all this is for your fieldshifted 25p sample: You need to capture 4:2:2 and keep it such until it hits Avisynth !

Cheers, I'm going to try that with my current camcorder pass through (as I don't have a deck with TBC yet and the BM requires one).

[rarend]

LOL! I figured you'd be another sucker for Black Magic's marketing hype. No surprises here. After you figure out the scanline and frame timing problems that others have always known about, figure out how to control analog input signal levels for legal digital luminance values and avoid dropped frames. Let us know and post the results. The glitches you have now were solved by analog source capture hardware a long time ago.

Ha! Unfortunately the delivery of the Diamond VC500 went wrong and it was shipped back to the US. But I paid for shipping costs yet again and it is again being send to me. You're not running a scam on these cards now are you?

More seriously: just checking, does the Diamond VC500 also require a VHS with TBC, or does your remark on analog source capture hardware imply that it may also work with my VCR without TBC?

Cheers.

[Emulgator]

Cheers, I'm going to try that with my current camcorder pass through (as I don't have a deck with TBC yet and the BM requires one).
rarend: This is where you came from. The Sony DV-AVI camcorder will capture DV-AVI 4:2:0 and your source will be destroyed before avisynth.