Artifact in U-matic (3/4") capture

[JVRaines]

I've been doing some U-matic videocassette captures via Canopus ADVC110. The DV files look good except for a sort of ringing artifact on low-to-high luminance boundaries. Is this a limitation of DV or the composite U-matic output? Or could it be a problem with the tape deck? I'm looking for some feedback before I go hunting up another antique VCR.

Here's an example. Check out the edges of the tuxedos against the white shirts. The "ripples" look animated in the motion video.

[filmboss80]

It is possible that the tape could be the later-generation U-matic SP (Superior Performance) recording, which would have higher luma and chroma subcarrier frequencies than standard U-matic. The picture would act up in the way you described if played in a standard U-matic deck. See if the cassette label says it's chrome tape. If so, it's SP.

Also, make sure the outputs are terminated. Sometimes when a 3/4" is in a loop-through mode, but not connected anywhere else, the signal will be amped up. (I'm going by 20+ year memory on this, but I think there's a termination switch on the back of most 3/4" playback decks.)

[JVRaines]

Thanks for the ideas. In fact, I'm using a VO-9850 SP machine. None of the cassettes mention SP or chrome.

So far I have just one video cable connected to VIDEO OUT. Are you suggesting I terminate MONITOR VIDEO as well? What about the VIDEO IN section? That's where the termination switch is located, to terminate one of the VIDEO IN jacks when not bridging through. Of course, I don't have any input to supply because I'm just doing playback.

[Bellower]

Hello,

I'm having the same problem by using professional umatic recordings on a umatic BVU-950 (with TBC and DNR boards included) and I only have connected the video-out to my capture board.
The image is great, but i have those "vertical lines" in the boudaries of everything.. like the first attached picture in this post
What's your suggestion in order to avoid those artifacts?

I really doubt that the umatics tapes have that artifacts in the source.. so i think this is regarding the umatic and its configuration or maybe do ia have to put some terminal in every output.. i don't know..

please if someone can give me a hand to solve this issue i will really appreciate it.

Thanks in advance!

regards!

PD: attached you can find a picture from the back of the BVU-950 and also a captured image where you can see the "vertical lines" on every border of the drawing.

[aedipuss]

well i wouldn't compare it to the first picture because hopefully he figured out that they were wearing pleated tux shirts and the lines belong there.
http://www.google.com/imgres?imgurl=http://www.dapperlads.com/images/product/images/PB...UBMAg&dur=1329

as for the lines in yours it looks like standard aliasing that occurs with low pixel counts and bitrates.

[Bellower]

Well, maybe the cartoon image is not comparative in that picture but it is the same "bourdaries" problem that the one of the first picture, there are some notorius vertical lines in the boundaries of the every object across the video.

My quesion is if do i have to put some "terminals" in every output as it was suggested in the second post.

Thanks in advance for your reply.

[henryperu77]

This an example with zoom

[poisondeathray]

Not sure about the 1st one, but the cartoon just looks like poor chroma upsampling (point resized so you get blocky colors edges) .

How did you take the screen shot ?

Either use better/other upsampling (when taking rgb images) or use antialiaser

e.g.
santiag(1,1)




e.g. antialias and fix chroma bleeding (sharpen color border edges)
santiag(1,1)
converttoyv12
mergechroma(awarpsharp2(depth=20))

[henryperu77]

He made the capture with a blackmagic card using bnc connectors from the U-matic player..
What you mean with upsampling?

[poisondeathray]

He made the capture with a blackmagic card using bnc connectors from the U-matic player..
What you mean with upsampling?

Yes, I know, but you're looking at an RGB image (it's not the actual video, it's a converted representation using a different color model or color space)
http://en.wikipedia.org/wiki/Chroma_subsampling

RGB screenshots are upsampled from YCbCr subsampled video (the BM usually captures in YUY2 or 4:2:2 . If it was unsubsampled it would be YCbCr 4:4:4)

The algorithm you choose to upsample can make the color borders appear blocky or less blocky (even though the video is the same)

[Bellower]

Hi.. thanks for your reply..
but this is not about the captured video.. even in teh preview mode (in vdub, amarec, iVCR, media express, etc.. ) I notice the vertical lines in every umatic tape that I played.

I thought in a very beginning that this was a umatic player issue, that's why a change from the sony VO9800 to the BVU-950 (with TBC and DNR board) and these "vertical lines in the boundaries" are always there..

I want to remark that I even notice this before capturing.. so this is not about the final captured file, this is about the singal it seems.

and again, is happening exactly like the first picture posted in this topic, i Just added another sample image form a cartoon... but it happens also in some live footage that i have on umatic tapes.

That's why i posted here because I would like to know if someone has faced this problem and how it was solved.

thanks again for all your help

[henryperu77]

Ok i uploaded a sample of video (no audio)

http://www.mediafire.com/download.php?sb5o8y3djsiy2mb

Its captured with AMV2 encoder you can download the codec here if you dont have it http://www.free-codecs.com/AMV2_video_codec_download.htm

[poisondeathray]

Your cartoon sample is different than what the others are talking about or the 1st screenshot (it's a "red herring") .

It does exhibit the problem to a small extent, but it's overshadowed by the chroma issue . The original problem is a luminance issue

[henryperu77]

The screenshot is from that capture..
The luminance issue occurs on the u-matic or in the PC when its captured?

[poisondeathray]

The aliasing primarily affects Y' signal (greyscale, black and white, not CbCr information) . The cartoon has an aliasing issue. This is NOT the same issue as screenshot #1 . Antialiasing will help the cartoon , but it won't help the problem in #1 . i.e you've posted a bad example and are confusing the issue. These 2 are demonstrating DIFFERENT problems

I don't know where the issue occurs from the information you've provided. It might be the original cartoon source, or upstream somewhere

If I had to guess, I would expect the artifacts in #1 are due to signal processing, maybe bad NR or a sharpening circuit . The artifacts in the hair suggest a some deinterlacing artifacts. They suspiciously look like yadif deinterlacing artifacts to me. The noise pattern suggests other processing has been done to it (it doesn't look an original unprocessed image, and is certainly resized to 640x480)



If you examine the greyscale image (Y'), you will see the vertical "artifacts" more clearly. But look at the spatial distribution - it's very different than in screenshot #1. These are different problems

[henryperu77]

Ok i uploaded another example from a different umatic tape, you can see the anomalies on edges specially on red color (but still present on another color like blue and brown). You can see it specially on the Ant Red Eyes

http://www.mediafire.com/download.php?s5l7e442acqz1bb

[poisondeathray]

Ok i uploaded another example from a different umatic tape, you can see the anomalies on edges specially on red color (but still present on another color like blue and brown). You can see it specially on the Ant Red Eyes

http://www.mediafire.com/download.php?s5l7e442acqz1bb

Recall the initial problem affects high/low luminance areas not color. Again, these are different issues

If you look at the grey scale images, you will see the vertical artifacts, but this sample has other problems that confuse the issue:

1) This cartoon is field blended
2) There are dot crawl artifacts

[aedipuss]

what did you deinterlace with???

[henryperu77]

Ok listen.. the only problem that really is annoying me is the vertical artifact .. the other problems like blended frames , grain, etc i can fix it with avisynth ...
I didnt deinterlace any of this videos yet.. these are the raw captured avis .. 720x488 (is not 640x480) 29i

[poisondeathray]

what did you deinterlace with???

It's not interlaced - it's progressive content (IVTC it)




The vertical artifacts might be related to the dot crawl. The distribution is similar. Your TBC or other signal processing might be getting confused. Have you tried disabling some processing options?

[Cornucopia]

Ok listen.. the only problem that really is annoying me is the vertical artifact .. the other problems like blended frames , grain, etc i can fix it with avisynth ...
I didnt deinterlace any of this videos yet.. these are the raw captured avis .. 720x488 (is not 640x480) 29i

Either you mistyped, or your capture was wrong to begin with. 720x488 is NOT a standard capping rez at all. Being Peru, which is an NTSC country, you should be capping 720x480. If you truly did 720x488, that could be why - your application is having to do realtime resizing from its internal resolution (probably 720x480) to 720x488, which is a 1.0166666 ratio (BAD! non-integer). If so, then I'm not surprised. And by 29i, I assume (hope) that you meant 29.97i, otherwise, you could be having framerate problems too (skips, duplications, blends, etc).

*************

The OP's signal, OTOH, seems indicative of unterminated signal chains. "Image reflections" occur with unterminated signals, so his/her signal path should be un-terminated except at the very end (and beginning, but that's usually self-terminating), where it SHOULD be terminated. If it can be done there with a switch, go for it!


Scott

[henryperu77]

what did you deinterlace with???

It's not interlaced - it's progressive content (IVTC it)




The vertical artifacts might be related to the dot crawl. The distribution is similar. Your TBC or other signal processing might be getting confused. Have you tried disabling some processing options?

Yes i tried it disabling it... How did you separate the Y/CBCR? it is possible to use avisynth to separate and applied filters?

[henryperu77]

Ok listen.. the only problem that really is annoying me is the vertical artifact .. the other problems like blended frames , grain, etc i can fix it with avisynth ...
I didnt deinterlace any of this videos yet.. these are the raw captured avis .. 720x488 (is not 640x480) 29i

Either you mistyped, or your capture was wrong to begin with. 720x488 is NOT a standard capping rez at all. Being Peru, which is an NTSC country, you should be capping 720x480. If you truly did 720x488, that could be why - your application is having to do realtime resizing from its internal resolution (probably 720x480) to 720x488, which is a 1.0166666 ratio (BAD! non-integer). If so, then I'm not surprised. And by 29i, I assume (hope) that you meant 29.97i, otherwise, you could be having framerate problems too (skips, duplications, blends, etc).

*************

The OP's signal, OTOH, seems indicative of unterminated signal chains. "Image reflections" occur with unterminated signals, so his/her signal path should be un-terminated except at the very end (and beginning, but that's usually self-terminating), where it SHOULD be terminated. If it can be done there with a switch, go for it!


Scott

I'm using a blackmagic decklink, it captures SD at 720x488 29.97i

[poisondeathray]

Yes i tried it disabling it... How did you separate the Y/CBCR? it is possible to use avisynth to separate and applied filters?

I would try to fix it in hardware first . It's better to address underlying problem.



But yes, in avisynth you can manipulate Y or Cb,Cr differently and apply different filters to each. Greyscale() will return Y' . You can use the merge() functions to join
http://avisynth.org/mediawiki/Merge

[Cornucopia]

Ok listen.. the only problem that really is annoying me is the vertical artifact .. the other problems like blended frames , grain, etc i can fix it with avisynth ...
I didnt deinterlace any of this videos yet.. these are the raw captured avis .. 720x488 (is not 640x480) 29i

Either you mistyped, or your capture was wrong to begin with. 720x488 is NOT a standard capping rez at all. Being Peru, which is an NTSC country, you should be capping 720x480. If you truly did 720x488, that could be why - your application is having to do realtime resizing from its internal resolution (probably 720x480) to 720x488, which is a 1.0166666 ratio (BAD! non-integer). If so, then I'm not surprised. And by 29i, I assume (hope) that you meant 29.97i, otherwise, you could be having framerate problems too (skips, duplications, blends, etc).

*************

The OP's signal, OTOH, seems indicative of unterminated signal chains. "Image reflections" occur with unterminated signals, so his/her signal path should be un-terminated except at the very end (and beginning, but that's usually self-terminating), where it SHOULD be terminated. If it can be done there with a switch, go for it!


Scott

I'm using a blackmagic decklink, it captures SD at 720x488 29.97i

NOPE. I just check the Blackmagic site. ALL Decklink & Intensity products capture SD using strict NTSC or PAL specs. That means 720x480 (or 720x576 PAL) - maybe 720x486 if they can also do the full 601 spec. If you are getting 720x488, either YOU or your capping application are doing something wrong.

Scott

[aedipuss]

never mind

[poisondeathray]

why is the second video 720x540? not a legal size for pal or ntsc.

"Jerry_Lewis.mpg" ? it's 720x480

Maybe you're looking at square pixels ?

[aedipuss]

too fast a reply - i looked at a screenshot without checking it with mediainfo :-\

[aedipuss]

as for the deinterlacing of the other one.

Originally Posted by aedipuss
what did you deinterlace with???
It's not interlaced - it's progressive content (IVTC it)

it's 29.97p not 29.97i so it can't be ivtc'd to 24p

[henryperu77]

aedipuss it doesnt matter, thats not a problem.., i can ivtc easily.. the main problem was that annoying vertical artifact.. -_-