VCR/Hi8 capture tests; help evaluate?

[dianedebuda]

Looks like you'all have had enough of this. Oh, well. I do want to express my appreciation to my mentors for all their time and effort.

Best wishes,
Diane

[sanlyn]

Looks like you'all have had enough of this.

Enough?

I don't know about everyone else, but still tracking in case you have more questions, or problems, or moving on to the next steps, or....whatever......

[dianedebuda]

I was going to post some hi8, but wanted to limit them to just a couple. Didn't hear from anyone about hardware combos. Afraid of "3 days fish" - and this thread is a month old and quite long. Really afraid I've worn out out my welcome.

If not -
For the VHS, it turned out that I was better off turning off all the VCR's "fixes" and running as plain as possible. Wondering for hi8 if it's the same with the TRV480. Several weeks ago, I made DV files with a variety of TRV480 settings. When I look at these, the DV versions have much smoother edges (arms, legs, etc.) than the ones captured via composite. Huge bleeding problem in the DV though. And DV is lossy. So am I better off converting DV for post processing or going with the capture card route?

I've been taught SO much here - you can't imagine how grateful I am.

[sanlyn]

I was going to post some hi8, but wanted to limit them to just a couple.

Thats Ok with us. No reason to overdo it.

Afraid of "3 days fish" - and this thread is a month old and quite long. Really afraid I've worn out out my welcome.

Not a problem. I can refer you to a thread that I and others worked on for 17 months. And that was only one video !

For the VHS, it turned out that I was better off turning off all the VCR's "fixes" and running as plain as possible.

Unless you have a really wacky tape, that's probably best. All you really need are setting decent dark and bright levels. Sometimes reducing saturation a bit can educe bleeding, but it depends on the tape.

Wondering for hi8 if it's the same with the TRV480. Several weeks ago, I made DV files with a variety of TRV480 settings. When I look at these, the DV versions have much smoother edges (arms, legs, etc.) than the ones captured via composite. Huge bleeding problem in the DV though. And DV is lossy. So am I better off converting DV for post processing or going with the capture card route?

Others might differ, but I don't think smoother edges balanced other VHS->DV problems such as halos that looked worse, bad (and sharpened) DCT ringing and edge ghosts, color changes that appear as false contouring in skin tones and a plastic look in objects such as clothing, along with worse line shimmer and interlaced edge "sparkle" than seen in the lossless captures. But give it a try and see.

[dianedebuda]

sanlyn, I'm focusing on hi8, not VHS, when I'm talking now about DV. For the VHS, I'm 95% set on using the 550. Moving on.

[sanlyn]

Oh. Well if your hi8 is digital, stick with DV.

[dianedebuda]

hi8 is analog, recorded with an older camcorder. So question is where the analog/digital conversion is done - internally in the TRV480 or composite/capture card.

[dianedebuda]

The hi8 tests

a and b are DV "captures" using trv480, tbc on & off
c is a composite to 550, proc amp levels set. Actually turned out to be same levels as used in earlier 550 vhs tests

Don't see a whole lot of difference. 550 has jagged edges with motion. All have significant color bleed.

Unless someone notices something else, the bleed seems to be the only thing that I'd attempt to post process for DV. Can the bleed be minimized? Make a difference for this if DV avi vs huffyuv avi? For 550, I'd attempt the bleed control & check about the motion jagged edges.

[Brad]

I haven't checked the forum for several days, so that's the reason for my own non-response.

The only brights that the camera's clipping are the light and the white socks, so I don't think that poses any loss whichever way you go. You want to bring the DV levels down to valid YUV range post-capture though, which is kind of a bummer since it means you can't just play the DV files as they are.

The vertical chroma position is different for TBC On vs Off. Not sure which is better, but you can just move that. For whatever reason, TBC On is less saturated.

550 negative: rainbowing from composite is visible on shoe highlights, shirt details, arm edges... It appears to lose out on horizontal detail. I'd take the DV captures over the lossless composite any day. The digital artifacts don't jump out as much as the analog ones.

Screenshots with adjustments; not much difference without zooming:

V1 = AVISource("a hi8 TapPractice trv480 +tbc, DV clip2.avi",pixel_type="YUY2").ConvertToYUY2(interlaced=true).ColorYUV(off_y=0,gain_y=-25,cont_u=35,cont_v=35).AlignExplode("Sony TRV480 (TBC On)",2,2,0,1)
V2 = AVISource("b hi8 TapPractice trv480 -tbc, DV clip2.avi",pixel_type="YUY2").ConvertToYUY2(interlaced=true).ColorYUV(off_y=0,gain_y=-25,cont_u=0,cont_v=0).AlignExplode("Sony TRV480 (TBC Off)",0,4,0,1)
V3 = AVISource("c hi8 TapPractice trv480 -tbc-dnr, composite 550 vdub B109-C148-Sh0 clip2.avi").ColorYUV(off_y=0,gain_y=0,cont_u=-20,cont_v=-20).AlignExplode("Sony TRV480 (TBC Off, DNR Off) <CVBS> ATI 550",0,4,1,0).Trim(1,0)

Interleave(V1,V2,V3)

[dianedebuda]

Thanks, vapereon. Just watching the captures, the DVs do seem clearer than the 550. But you folks have taught me to not necessarily trust first impressions 'cause post processing is really what counts.

I'd take the DV captures over the lossless composite any day. The digital artifacts don't jump out as much as the analog ones.

Doesn't matter to me whether it's DV or 550 as long as I can get to something with potential that can be worked on in vdub/avisynth. Is there additional loss (beyond the native DV) with ConvertToYUY2?

A lot of my hi8 tapes are of HS bands where the uniform colors are red/white or red/blue and recitals with bright costumes on stage in dark auditoriums. The couple that I looked at are like the TapPractice with significant color bleeds. I'm not sure if the bleed is all/partially caused by the camcorder and actually on the tape (player doesn't matter) or it's the TRV480. When searching for how to handle hi8 bleeds, ran across a 2006 thread that I had read months ago, Capturing Video8 and Hi8 - guides. There swiego is comparing his Digital8 camcorder with a hi8 deck, EV-S7000, connected with s-video to a 550 (coincidence?). In post #13, he says:

Digital8 (camcorder) vs. Hi8 Deck:
1. Resolution - the Hi8 deck captured more detail. I saw this consistently, in the DVD caps as well as the res chart capture. For Video8 material, you won't notice any difference. For Hi8, my EV-S7000 captures more information. Hi8 Deck wins.
2. Image quality - the Hi8 deck exhibits almost no color bleed. It's a huge difference. Color is more accurate to the source. The picture is smooth and artifact free. On the other hand, the D8 produces a slightly less noisy capture. My opinion, Hi8 Deck wins.

A quick look at amazon & ebay suggest that particular deck is really expensive, but there are some EV-S5000 that might be available for < $200. Don't know spec differences yet. Wonder if a deck would even be worth looking into?

Reading Re-Newbie Needs Startup Advice; hopefully a lot of my post processing questions will be answered there since it seems to be detail-rich like oldretiredguy's thread was.

On my ToDo list this week is learning how to read these histograms and waveforms. I've seen a number of them, but can't interprete - yet I know, you've told me I was going to need to learn it before. Busy time of year - TCM 31 Days of Oscar and figure skating at the Olympics. At least I'm not redecorating like sanlyn.

[sanlyn]

I'd take the DV captures over the lossless composite any day. The digital artifacts don't jump out as much as the analog ones.

She can figure out how to clean up the sharpened DV edge noise, posterization, and bad ringing. I'd take the 550 as much easier. By the time you clean up those DV defects, you'll have the 550 and a little less.

[dianedebuda]

Just wanted to add that I am looking at the histogram/waveform "education" links back in posts 14-16. Didn't register much then as to what they were showing, but hopefully will be much better this time around - like a lot of other topics I have reread now.

Trying to get the filters that sanlyn used for the guitar clip fix earlier. Went to avisynth and see stuff like "FixChromaBleeding" listed as external filter with download link. Tried to download, but site requires registration. Ok, did that. Can't download without email verification. Have tried several times on different days, but never receive email. Checked & rechecked that my email is correct & not getting put into spam. Ever run into this?

[sanlyn]

Don't download stuff from sites that require registration. A lot of those filters were originally posted in the doom9 forum and are still there.

The original function is posted here: http://forum.doom9.org/showthread.php?t=77074
.
There was also a modification in that thread, now called Fix ChromaBleeding2. But is uses a different version of the MaskTools plugin sometimes, so I usually use the first one listed above. I've attached avs scripts for both functions.

[dianedebuda]

Is http://avisynth.nl/index.php/External_filters not the "official" site? Then I'm going to need the other avisynth "externals" used too: Vinverse.avs, masktools.dll and mvtool.dll. Actually got a masktools, but it looks like it may be MT.

I don't tread much over at Doom9; most of the discussions are too deep for me and often rather combative - at least that was my impression 10 years ago when I set up my first capture box. I'm a wimp?

[sanlyn]

That is the official site. You'll find the original FixChromaBleeding listed there, download by clicking on the link titled "Script" on the right an copy the text displayed on that page. Another official site has older plugins and earlier versions of new ones (http://www.avisynth.nl/users/warpenterprises/) plus some extra page links at the top of that page. For the past several weeks AVisynth developers have been changing web servers, so many links have yet to be updated.

Doom9 is usually above my head as well and seems angled more at developers. But you can still pick up a lot of tips if you dig in a little, and there are many unpublished scripts that haven't found "prime time" yet in the usual lists. But, yeah, you can spend a lot of time in there and come away in a coma if you don't get into a fist fight before then. There are also German and other international versions of the forum. The two "official" links above should be all you need.

[dianedebuda]

What I was doing was clicking on FixChromaBleeding which took me to a subpage that listed a download; clicking the download gives a "you must be logged in" message. Of course it might have helped if I'd noticed the message continued with "to upload files".

Funny, but as I wrote my previous msg, I was reading Postprocessing video using AviSynth - on Doom9.

[sanlyn]

That postprocessing site has been around since The Flood, but it's still worthwhile. If you have nothing to do for the next 3 months you can go to their capture guide, of which the "post" page is only a small part. Unfortunately most of the capture info goes back some years and is directed mostly at anime . But some of it is still relevant. That is, if you got nothin' to do....

Then there always AMV (Animated Music Video). Lot of stuff, some of it relevant to what you're doing -- but a good page on Avisynth flters and the kinds of artifacts one encounters is here: http://www.animemusicvideos.org/guides/avtech31/post-qual.html . Scroll down the page and you can click on the images to see before/after (Wow! Cheap thrills!). Actually a lot of the filters they demo are in regular use with "real" video, too. The anime images help make the artifacts easier to spot.

[Brad]

Is there additional loss (beyond the native DV) with ConvertToYUY2?

Whoops. That's a leftover that's literally doing nothing, because the source filter is providing YUY2 to it. That one can be removed.

I'm not sure if the bleed is all/partially caused by the camcorder and actually on the tape (player doesn't matter) or it's the TRV480.

Hi8 has only slightly better color resolution than VHS, despite having much improved luma bandwidth. Maybe the deck really would help, I dunno. At some point you're running up against the format's design though.

She can figure out how to clean up the sharpened DV edge noise, posterization, and bad ringing. I'd take the 550 as much easier.

I'm not sure where you're seeing those particular artifacts. My understanding is that there's no fix for composite rainbowing on moving objects with Avisynth, especially not rainbows that already escaped the 550's good 3D comb filter.

[sanlyn]

I'm not talking about rainbows. Frankly, if you folks see "no difference" or "nothing", then at that rate I don't even know why you're advising about filters, because apparently they aren't needed. I'd just advise to lean close to the monitor while this stuff is playing and whisper "You're OK now", then go straight to encoding. Who would ever know?


Meanwhile, I think dianedebuda ought to be cautioned: the same filters can' be used for everything, so don't be concerned if some routines look OK on one video but not on another. Choose filters to match the problem you detect, and don't be afraid to read up or ask about different settings. It doesn't take long to get a feeling about what a particular plugin is doing (or not) if you test in small segments. There are also some coding tricks you can try:

If you want to end processing after a couple of filters or before other filters start to work in a script, insert a line of code where you want to stop-and-check, like this:

return last

If you want to skip a filter or lines of code, put a comment mark (#)at the beginning of a line:

#return last
# skip this one also
#skip
#three
#lines
run this
run this
#skip this

For the DV captures you'll need some stronger anti-alias and edge cleaners, and some softening/deblock/deband filters to keep faces from looking like cartoon masks and to prevent skin shadows from appearing to have "hard" edges with "hard" highlights, or what is known in plasma TV circles as "false contouring" (some call it posterization, others call it clay-face). Some of its effects are similar to over sharpening artifacts. Edge artifacts, ringing and ghosts are tougher to clean with analog->DV; ghosting is more obvious and looks grainy.

For those who don't "see" any of this, the whisper trick will probably work. It doesn't work at this end, howeer. Never did. Maybe if I used a cheaper monitor....?

[dianedebuda]

Having problems getting DV files to open in avisynth - no decompressor. Googled & see 1 common solution is cedoceda. Is that what you use?

Opened my first avisynth histogram with the Tap 550. See "activity" in the low dark range. If I apply offset Y 15, it moves all over to the ok range with the super whites (correct term?) mostly off the chart. Probably can use gain to rein that in. But the question is: when I captured, nothing showed red on the left side, so why am I seeing "stuff" in the "danger" zone to begin with?

the same filters can' be used for everything, so don't be concerned if some routines look OK on one video but not on another. Choose filters to match the problem you detect,

Even if that thought had ever occurred to me, I'm sure a few threads on correcting this and that would have enlightened me otherwise.

My biggest concern right now are these color bleeds. Don't know if/how much they can be fixed in post processing, so don't know if I should look for a deck. The deck idea is expensive and there's not a lot of feedback to give confidence that this is truly a potential solution. Sent a PM to swiego, but really don't expect a reply - it's an old thread that hasn't had replies to later posts. I can hardly remember tests & setups I did 6 months ago let alone 8 years ago, so I'm not faulting him at all.

[jagabo]

Having problems getting DV files to open in avisynth - no decompressor. Googled & see 1 common solution is cedoceda. Is that what you use?

Yes, if you're using AviSource().

Opened my first avisynth histogram with the Tap 550. See "activity" in the low dark range. If I apply offset Y 15, it moves all over to the ok range with the super whites (correct term?) mostly off the chart. Probably can use gain to rein that in.

Or ColorYUV(cont_y=-val) will bring black up, white down. Or use a combination of gain, off, cont.

But the question is: when I captured, nothing showed red on the left side, so why am I seeing "stuff" in the "danger" zone to begin with?

You'll have to provide a sample.

[dianedebuda]

You'll have to provide a sample.

The c 550 one in post 128.

[sanlyn]

To open your posts I have to enable huffyuv in ffdshow and use DirectShowSource instead of AviSource, or I get a scrambled image. I have cedocida installed. The problem is compatiblity with different versions of huffyuv and the way it installs and behaves on different machines. Apparently I'm also using your same version of huffyuv (CCE Patch version, lordsmurf source, since 2004). During post processing I recompress with Lagarith. No problems.

Which histogram are you using? If i enable Avisynth's "Levels" histogram I don't see anything in the red. If I use the ColorTools RGB histogram, I see some left-hand overshoot because of the black borders, and some overshot in the brights. Remember that RGB expands or "enhances" the extremes at each side during YUV->RGB conversion. I don't see levels problems with the "c" post that can't be fixed. It's when detail gets clipped or looks crushed/blown out that you have a problem. If they exist, lowering contrast usually clears it.

[dianedebuda]

That's the huffyuv I'm using.

AVISource(xxx)
histogram()

Remember, I said it's my first avisynth histogram, so no options, settings, etc. Just didn't expect to see anything in the left "danger" area. Seeing stuff made me question if I'd missed something in capture.

[jagabo]

You'll have to provide a sample.

The c 550 one in post 128.

I don't see anything out of range in that sample:

function HistogramOnBottom(clip vid)
{
  top = vid
  TurnRight(vid).Histogram().TurnLeft()
  bot = Crop(0, 0, width, 256)
  StackVertical(top, bot)
}

AviSource("c hi8 TapPractice trv480 -tbc-dnr, composite 550 vdub B109-C148-Sh0 clip2.avi") 
HistogramOnBottom()

Note that Histogram() defaults to showing a waveform monitor, not a histogram. The brown bars demark the illegal luma values. I prefer VideoScope() for this but it doesn't indicate where the illegal levels are:

[dianedebuda]

Histogram default = waveform, not histogram. A planned gotcha for newbies, right?

Have downloaded VideoScope since I've seen you use it a lot, but baby steps here. Got to understand what I'm seeing.

[jagabo]

Have downloaded VideoScope since I've seen you use it a lot, but baby steps here. Got to understand what I'm seeing.

A ways back I gave a link to an explanation of waveform monitors.

https://forum.videohelp.com/threads/361379-VCR-Hi8-capture-tests-help-evaluate?p=229648...=1#post2296488

[dianedebuda]

I tried to say a few posts earlier that I was still aware of the "educational" links so you wouldn't have to repost. Didn't understand them then much at all, but had a real terminology handicap that I've been slowly trying to resolve. Like explaining trig to a preschooler. I started from the point of RYB are the primary colors (kindergarten level) and RGB are color coordinates. Contrast, brightness, hue, saturation, etc. meant almost nothing. But I'm gaining on it - I hope.

[sanlyn]

I started from the point of RYB are the primary colors (kindergarten level) and RGB are color coordinates.

RYB are "printing" colors (subtractive process). I wonder why they didn't include black ("K")? Too early for those days? Actually, its Magenta Yellow Cyan. Never mind.

RGB are colors of light (additive process).

The world prints with CMYK. The world watches TV in RGB.

[Brad]

Wikipedia to the rescue? Those darned kindergarten teachers were out to confuse us all along.