Is it Color/Chroma Shift or Bleed or Smearing & can I improve it?

[grannyGeek]

Here I am again, pleading for advice.

My vhs tape has some color problems.
I’m not sure what this condition is called, and since I don’t know how to identify it, I don’t know which filters to investigate.

Attached some cropped screen-grabs from my vhs DV-Avi capture.
(tape is commercial movie vhs, VCR is mid-grade JVC S-8010UM, captured with DVC-90 USB2 device in Pinnacle Studio)

Note the blurring of color around the shrubbery, and down the side of the Hero, bleeding into the blue of the sky.

It’s most noticeable in sequences shot against that deep blue sky
Other frames show strong edge-enhancement white halos.


Is the dark color blurring Color-Bleed or Smear or is it Chroma-shift?
is there any way to clean it up a bit?
Can anyone recommend AviSynth script / filter, or a VirtualDub filter?

Thanks for any and all help ---
Granny


crop-shots.bmp

[edDV]

...

Note the blurring of color around the shrubbery, and down the side of the Hero, bleeding into the blue of the sky.

It’s most noticeable in sequences shot against that deep blue sky
Other frames show strong edge-enhancement white halos.


Is the dark color blurring Color-Bleed or Smear or is it Chroma-shift?
is there any way to clean it up a bit?
Can anyone recommend AviSynth script / filter, or a VirtualDub filter?

Thanks for any and all help ---
Granny


crop-shots.bmp

Among normal noisy VHS issues like smeared 500KHz bandwidth chroma and edge ringing in luma, that picture is showing significant chroma shift right vs. luminance. This is similar to a misconverged Tv and may have been caused in the recording side of the VCR, the playback side or in the capture device. If it is the VCR, you would see this happening on a good monitor connected directly to the VCR. I suspect the capture device.

Since you cropped into the middle of the picture, it is hard to tell it luma is shifted right or chroma left (delayed) or some of both. Show the picture sides against blanking.





You cannot correct this with a RGB filter. The capture must be YUV and you need a filter that will advance or delay (i.e. move horizontally) UV (chroma) vs. Y (luminance).

[grannyGeek]

edDV, thank you so much for responding.
oops, I'm sorry, I didn't know it would limit examination of the info by cropping the screenshot, but now that you say it, that makes perfect sense. If it would be useful, I can post full screen-shots later today/

The vhs tape is a commercial product, not a dub.
So, you are probably right to point to the capture device as the culptit.
Unfortunately, it is all that I have available to capture with.

Would a different codec improve matters?????

If I use VirtualDub to capture instead of Pinnacle Studio (which I think possibly uses its own proprietory DV codec), can you recommend what codec would be a good choice to try.
I have Huffyuv available, or what would be a good candidate to download?
( sorry, I know zilch about codecs)

Thanks in advance for any further advice you might be willing to give.

Granny

[vhelp]

I have my own quams about your capture device, DVC 90 and the codec
it seems to be using. You said it (or you) were using a DV codec.

When I looked closer at your pic above, I saw resemblence of MPEG
macroblocks. Look near the moving hands and you might see where the
image broke away (misaligned) inside the 8x8 transform. These are
displaced pixels -- without going into boring details -- you don't
notice it at regular distace, but when you maginify the image and
analize them, you (and know what to look for) begin to see things
that you know should not be there.

These can mislead you to other conclusions, such as bleeding, etc.,
though you may have a comination of both, or other causes.

If the video is not a personal or priceless value antique, I would
consider encoding it at a very high bitrate that your dvd player will
handle (I use 9000 cbr) and that might improve it's playing quality.

VHS needs a lot of bitrate because of that is perceived as noise to
the MPEG. In short, vhs lack "gradient" texture. The pixels are
not gradient like. What is soppose to be a gradient value to naiboring
pixels is often the oposite -- a darker or lighter color pixel, not
matching the previous or naiboring pixels. Multiply this throughout
the whole video source, frame by frame, and you have MPEG's worst
nightmare. And, there reason why DVD sources come out so much better
is because they don't suffer this phenomina. The source is usually
very clean, and pixel coverage *should* be gradient.

For some strange reason (I haven't put my finger on it yet) VHS and Cable
sources share the same phenomina of noise.

-vhelp 4068

[grannyGeek]

vhelp, thanks for that information.
( slowly, but surely, you guys are educating me )

The movie is absolutely NOT a precious possession, but right now it is my learning tool.
It has a variety of problems that have ended up teaching me about
telecine and the DeComb filter,
dirt and scratches and the DeSpot filter,
and now these little color problems.

As I said earlier, I would like to try to capture again with a different codec and see if color improves
( just experimenting you know, not expecting miracles! )

How about you --- any suggestion for a codec to use with VirtualDubMod or VirtualDubMpeg2?

Thanks
Granny

[lordsmurf]

Ghosting.

[grannyGeek]

Ah. Spooks, I see ......
I am starting to feel a bit haunted by this thing.

Well, I guess I'll just have to bite the bullet and try some AviSynth filters and see where I land.

I've only ever heard of a VirtualDub filter called Exorcist, any chance of someone telling me some other options?


Thanks LordSmurf
BTW, your site is a whole education by itself, thank you for making it available to us.


Granny

[edDV]

S-Video would be the best way to connect. This avoids delays in the Y/C separator in the capture device.

I'd try Huffyuv for a test segment and see if it makes any perceived improvement. The color misregistration is probably an artifact of the hardware. If so, it will appear in all captures. I doubt the Pinnacle studio software has any Y/C delay adjustment filters. They probably exist for Premiere. I found a link for a Y/C delay filter for Final Cut Pro.

Here is a brute force analog technique if all else fails
See this S-Video DA writeup on the problem. http://www.inday.com/PDF/svda-manual.pdf
Each pixel is ~74nsec wide. 75ohm coax delays video by 1nsec for every 0.648ft. so one pixel correction would require 48ft of coax inserted into the luminance path. Good luck

PS: That won't be very practical, better to find a Y/C delay filter or get a better capture device. Y looks to be 15-18 pixels displaced right or ~ 750 ft of coax

[grannyGeek]

edDv, thanks for posting back.
I do use S-video cable, but since the source tape is not S-vhs, I'm not sure how *much* it actually improves things, tho I'm sure it does help a bit.
I will first try capture with Huffyuv and see the outcome.


Thanks for the link, it looks very, ummmmm, interesting.
Last resort indeed! By the time I paid for that much coax, I could buy several new capture cards, *and* pay an expert to install them, and rub my feet afterward.
Or even buy all my vhs movies on dvd.

I'm googling right now for Y/C delay filters, haven't found any free yet, but the day is young.

Thank you for your input.
Some things to sink my teeth into !

[edDV]

edDv, thanks for posting back.
I do use S-video cable, but since the source tape is not S-vhs, I'm not sure how *much* it actually improves things, tho I'm sure it does help a bit.

The idea of using S-Video was to avoid the Y/C separator in that capture device that may be causing the delay problem but it could be the other way around. Try composite and see if the delay timing improves for that little capture box.

[grannyGeek]

gotcha.

Will try some various combinations of cables and codecs, and see what turns up.

I'll post back results after the weekend.


Thank You for your interest.

Granny
she might be getting on, but she does keep going on . . . .

[Gargantua]

I'm googling right now for Y/C delay filters, haven't found any free yet, but the day is young.

Get the Chromashift plugin for Avisynth or "flaXen VHS Filter" for VirtualDub. Both are free.

http://www.geocities.com/siwalters_uk/chromashift.html
http://neuron2.net/flaxen/flaxen.html

[gshelley61]

If the error is embedded in your source video, most professional type TBC's have a Y/C timing adjustment that would probably fix or improve on the problem. The I.DEN IVT-7 is one that comes to mind.

SignVideo makes a standalone hardware device called a Color Registration Corrector that essentially does the same thing as far as I can tell...

http://www.signvideo.com/cr100_color-registration-corrector_video-processor.htm

[grannyGeek]

Hi guys, thank you for responding.

I'm having mixed luck ---
my vcr went on the sick list, so I can't experiment with new captures right now. I can only work with the original capture.

Gargantua, I did try out those filters, and they did improve matters a bit.
The chroma-shift filters can only help to a certain degree, because some segments seem to have bleeding both to the left and to the right, so I can only get a clean edge on one side of objects/people.
But its quite some better than it was!

I also found the DeHalo_alpha script/function, which did a nice job of attenuating the bright white halos (now everything has a soft angelic glow instead ). No, seriously, it actually helped a lot on the halos.

gshelley61, that Color Corrector or an external TBC does look like something to consider.
I wonder also, if I ought to invest my cash in a better capture device as well, this little usb2 gadget might not have been the best choice. I was limited though, using a laptop didn't allow too many options.

I'll keep experimenting, and try new captures when my vcr is fixed.


Thanks so much for your input.
I'll be back to bug you all again when I get more results.

Best Regards to you all ---

Granny

[grannyGeek]

Back again, embarassing myself in public.

It seems that my “bleeding” problem appears in the source vhs tape as well.
I viewed on 27” JVC tv, and the tape looks about the same as the capture, both with my JVC vcr, and with a borrowed old Hitachi vcr. So that’s pretty embarrassing, because I never noticed it before (of course, I haven’t watched it on tv for donkey’s years)
As I noted in original post, the smear effect is most blatant when filmed against deep blue sky, and only a few segments are like that.

For giggles, attached screen grab of Huffyuv capture with DeHalo applied.
I guess the Huff codec might be worth the huge avi file size – seems like a cleaner picture.
DeHalo_alpha dims the halos, but dims all white highlights also, and somewhat softens the picture overall. It’s a tradeoff.

gshelley61, your recommendation is tempting me so, but present budget won’t allow.
I must either put everything on hold until the checkbook smiles again, or tweak the capture as best I can with ChromaShift and DeHalo.
I wish they made “trial” hardware, so I could test a TBC or Color Corrector, and see if the improvement is worth waiting for.
Decisions, decisions . . . . . Anybody want to offer an opinion? Should I wait?
Remembering of course, it’s only a silly hollywood movie, not a priceless family memory.

I have a feeling this movie isn’t done with me yet. I hear it chuckle in the dark “wait till she sees the dvd, heeheeheehee.”
I am *not* paranoid.

My apologies for kind of posting a “false” problem, but you guys can go to sleep tonight with a smile, knowing you gave me some badly needed education. Now I have a fair idea how to handle or prevent certain capture problems, and that’s a very valuable thing.

regards ---

Granny
she's getting smarter in her old age, and YOU helped





huff%20yuy2%20with%20dehalo_alpha.bmp

[edDV]

Normally one would test the process with a known good source tape. Ideally this would be a high quality prerecorded tape without Macrovision. Good sources might be professionaly recorded education or corporate demo tapes, NASA VHS tapes, etc.

A known good tape eliminates the recording half of the VCR as a source of trouble. We are assuming the dub house had their equipment in proper repair. Using this tape the playback VCR, cabling and capture device can be evaluated. If you are lucky the tape will contain a reference color bar that can be used to test levels. Other test signals measure component timing.

NTSC color bar


Pulse and bar


That will be about as good as it is going to get. Next you load your tape and see how it compares to the reference tape.

[grannyGeek]

Ah-so. good information.
If I'm understanding correctly, that will tell me if this problem can be corrected in hardware ???

I will ask around and see if I can lay hands on such a tape.


Thanks edDV, you have a lot of patience. And I do appreciate all the good teaching you are doing here.

Granny

[edDV]

It's just a way to separate hardware issues (vcr or DVC-90 USB2) from issues unique to that tape.

[grannyGeek]

right, I think that was what I was trying to say / type, but it came out really wrong.
That this test establishes if there is a hardware problem.

what can I say, I just washed my hands and I can't do a thing with them.



Thanks
granny