View Profile
View Forum Posts
I ran the script with BOB instead of yadif. Never used stab() with bob before. Cleaned a little more jitter. Yadif "looks" better. Your choice. There are other smart bobbers around.
Don't forget, there are still a handful of other frames that need fixin'.
Yes, yadif does some interpolation. That can be turned off, but it often doesn't look as clean.
+ Reply to Thread
Results 91 to 120 of 197
-
Last edited by sanlyn; 21st Mar 2014 at 18:49.
-
If you want to avoid interpolation with Stab(), you'll actually have to avoid bobs altogether, use SeparateFields.Stab(dxmax = 0).Weave() (maybe with AssumeTFF calls in there too), and modify the Stab() script so that the Depan call has a "subpixel = 0" argument at the end.
This will ensure that the output fields are the same fields as the original clip, just displaced vertically...but restricting Depan to whole-line movements will make Stab() even less effective in terms of fixing jitter. -
In other words, there's no free lunch. I'd try a recapture with a tbc of some kind, first. And repacking the tape would help. There are bound to be more problems elsewhere in this old tape and the others. I used a bobber because it was quick, I'm working on another project at the same time (and it's a nightmare). Thgen again, there was those missing 10 pixels at the bottom; I got tired of the math. That happens when you're not cut out for multitasking.
Last edited by sanlyn; 21st Mar 2014 at 18:49.
-
Instead of Stab after Yadif, separate fields and Stab the even and odd fields separately:
I didn't bother fixing the small black borders that Stab will create when it shifts a field.Code:AviSource("test4.avi") AssumeTFF() AddBorders(0,0,0,10) ConvertToYV12(interlaced=true) SeparateFields() even=Stab(SelectEven()) odd=Stab(SelectOdd()) Interleave(even,odd) Weave() TFM() TDecimate()
Nevermind. It works for some, not for others. I think the capture card is getting confused about which fields are top fields and which are bottom fields. So Stab after Yadif (or QTGMC?) works better.Last edited by jagabo; 27th Mar 2012 at 16:25.
-
I changed the script posted earlier in post #87, to work the borders differently. Borders don't any residual noise now. Remember this test4.avi had 10 pixels removed from the bottom for "illegal" 710x470 frame size. The bottom 10 borders of noise have to be removed anyway, even if they were present. This version of the border stuff got all the hopped fields and a couple more that were missed earlier. All that's left are two problem frames (actually, 3 bad fields).
Code:LoadCplugin("K:\AviSynth 2.5\plugins\yadif.dll") Import("K:\AviSynth 2.5\plugins\stab.avs") AviSource("J:\forum\videeo\test4\test4.avi") ## ------------------------------------------------------------- ## Input is 720x470. Add 10 p[ixels for a border to get 720x480 ## ------------------------------------------------------------- AddBorders(0,2,0,8) ConvertToYV12(interlaced=false) #Crop(0,0,0,-4) #<- changed Crop(0,0,0,-8) ConvertToYV12(interlaced=false) AssumeTFF() yadif(order=1,mode=1) Stab() SeparateFields().SelectEvery(4,0,3).Weave() #Crop(0,4,-2,0) #<- changed Crop(0,4,-2,-2) AddBorders(2,4,0,10) #<- changedLast edited by sanlyn; 21st Mar 2014 at 18:53.
-
Interesting results! This looks like a promising way to clean-up some more troublesome spots. I tried it out on test4.avi. What would need to be added to the script to get it to IVTC?
I am going to look into getting some hopefully reasonably priced hardware, either a VCR with a LTBC or something like a ES10. Is the ES15 ok? I read that the other Panasonic models really don't compare, but they are more available at the moment. -
AViSource("path to file\avi-name.avi")
AssumeTFF()
TFM()
TDecimate()
Pannies? I think their output looks grungy. But people like juiced-up contrast, red push, and illegal blacks. Es10,-15,-20 pretty much the same, and their hard-drive equipped equivalents with similar ES numbers. ES-15 not as good on built-in DNR or built-in encoder, but you want the tbc. Later models, not so great. Toshiba R-D2 thru D5 series (sometimes called "RDK" series), RD-XSxx series. And...oh, man, what 's the model of that Philips ? ? (pretty expensive new, but you never know). Oh, yes -- DVDR3475, DVSR3575. In a pinch, new Toshiba and Magnavox combo's have 'em, they're OK if not quite as good as the older guys mentioned (let's hear it out there for the old guys, pee-puuuuhle!) and still better than the primitive and artifact-plagued ltbc's of burned-out VCR's. Can't vouch for other brands or anything else new.
Anyway, get one of those nice TBC -equipped VCR's and what happens with those scrungy older tapes? They spit 'em right back at you.Last edited by sanlyn; 21st Mar 2014 at 18:51.
-
Just ran IVTC on the repaired clip. No duped frame or motion stutter this time, the jittery version must have been throwing TFM off. But this time 0-frame is still discolored. Sometimes happens on frame 0, which is why you should usually include a few frames precteding the frames you want to ivtc. Still, it can be fixed.
Last edited by sanlyn; 21st Mar 2014 at 18:51.
-
Mini-Me mentioned the DVD3475 in a previous post, and also that it needs to run with a proc amp? I'd like to minimize the additional gear, especially if I have to start hunting for lost remote controls.Originally Posted by sanlyn
A currently sold Toshiba or Magnavox combo (recorder?) has a TBC? Do you happen to know of one that has been verified to have one that works OK?
I was thinking that getting a JVC with TBC and DNR but now you've got me thinking maybe I should just go with a pass-through on a VHS/DVD recorder that has a TBC so I know I have a mechanism that works reasonably well. But can the old JVC VCR's also work in pass-through? That could open up the option of going either way. -
Something to be wary of is that Bob() resets the assumed field order (of its progressive output) to the default BFF. So when you reinterlace you will get BFF unless you add AssumeTFF() before SeparateFields().Originally Posted by sanlyn
(I haven't followed this thread closely and don't really understand what you're doing, so I'm not sure if this issue is relevant here.) -
Just to chime in on a few of your questions:
Regarding the DVD3475 -- you'll only need a proc amp if you're planning on actually capturing with the DVD recorder. If you're just using it for passthrough, you should be fine using the color controls on your capture device. It'll also prevent dropped frames on capture, which yes, the Magnavox DVD recorder previously mentioned above will also do -- not sure about the Magnavox combo decks or the Toshiba, so I won't speculate.
Regarding the JVC SVHS decks -- No, you can't use the JVC SVHS deck DNR/TBC stuff as a pass-through device. -
I wanted to touch on this point a little bit, although it has been addressed somewhat.Originally Posted by videeo
In my experience, almost any decent VCR can capture all of the actual detail in a VHS tape -- it's what it does with that capture that counts.
For example, many consumer decks made in the early 2000s and on will record and playback with a fairly sharp image, but then go and over-sharpen that image to death, presumably so that the relative softness of a tape didn't stand out next to a cable signal. Some people DO like this look and grew up on it, so they feel more or less that "this is what VHS should look like". This wasn't such a huge deal when TVs sold were mostly between 20 and 27 inches, but it can get ugly when you watch that same footage on a big TV these days. If there is no adjustable sharpness control, you can't turn it off when feeding it to a capture card or recorder. You'll have to address it in software, but depending on the deck, it may leave other artifacts behind once the typical noise is cleaned away (ghosting, edge enhancement, etc).
On the other hand, the Pro(sumer) JVC decks (the 9000 series, the 7000 series, the SR-V101US) have a fairly aggressive set of noise reduction filters -- some people love them (because their footage looks noise free on a big screen in almost any encoding workflow) and others hate them because they see the aggressive NR as throwing the baby out with the bathwater and filtering the picture into a gauzy, plastic haze. As mentioned previously, many of them don't like EP tapes, although yours appear to be SP. Some of them also don't track hi-fi audio very well, so that is another consideration if your tapes have a hi-fi audio track (check the box).
You do have some control over the image with most of these decks -- you can always put them in edit mode and get a sharper picture, although you lose some of the NR capabilities. However, be aware that the TBCs themselves (when engaged) can introduce errors into the image, such as vertical jitter for various frames or a horizontal bending of the upper 20% of the picture or so (typically referred to as flagging). So you may find you can correct one type of jitter, but you'll have to then address the TBC induced type.
The SR-W5U/SR-W7U decks (and perhaps some of the DVHS decks, I'm told) are the exception to these rules (minus the occasional vertical jitter caused by the TBC) but you'll pay big bucks to get one ($500 - $800 seems to be the running average these days). They are often untested by the sellers who have no idea what it is, so never buy one without a return policy. They are also frequently damaged in shipping as they are about twice as heavy as the average VCR and have many sensitive internal parts. JVC also may not service them at all at this point. Still, if you have money to burn, they do have a full frame TBC and line-TBC built in and are much better at tracking EP tapes and dealing with hi-fi audio.
As far as the line TBCs in these VCRs -- they don't really impact the perception of true detail IMO, they mostly impact the color purity and the chroma noise, which indirectly may impact the perception of detail and clarity in the image. In my experience, the passthrough DVD recorders I've used do not do anything that resembles this effect. Once again, some people like it because it gives the image a cleaner look, while others feel it slightly smudges edges together in an effort to make them look more uniform.Last edited by robjv1; 28th Mar 2012 at 18:19.
-
I've always used Assume(tff/BFF), probably too often, but I removed it from the last re-interlace statement to check its effect. No change. Thanks for heads-up that Bob() insists on making certain assumptions; I've always used it with Assume(*).Originally Posted by Gavino
We are working with a capture of a stretch of bad hard-telecined tape. Jitter, hopping, broken 3:2 pulldown, etc. We know the clip is TFF by watching blended frames with movement. with TFF motion is smooth, with BFF it's back-and-forth.
The last thing to fix is frame 0. It shows jumbled parts of other frames and 1/3 of the bottom is green. If you bob that frame or separate fields, the first field displayed is a damaged/green image (top/even field 0), followed by a perfectly good image (bottom/odd field 1). So you'd think that with TFF, and separating frames with Bob or whatever, you should be able to replace the bad "0" with a copy of the good "1".
But it didn't work that way. Originally I didn't use FreezeFrame, but used it here just to test one frame. If I use FreezeFrame(0,0,1), I get two copies of the "bad" image. If I reverse it with FreezeFrame(1,1,0), I get two copies of the "good" image.Last edited by sanlyn; 21st Mar 2014 at 18:51.
-
As always, tried to read through these posts and have no bloody clue what is being said....Some of this stuff is way too complicated for me.
Back to the HALF D1, personally do not use it, however also don't use a PC capture program or PC capture. Don't like it and have never gotten good results. Plus it is a pain to code the video.
Did several tests a few years back with converting VHS to Half D1 using 3 hour mode on a JVC recorder. My eye for detail is pretty good, found very little difference in the picture. With Digital Recordings that have all the lines, yea found the picture dull and not as good as full D1. VHS only has 240 lines of res, so it really doesn't matter. Half D1 is like recording in XP mode (cause of the bitrates). With HDMI outputs you get very little to no Macro Blocking in the picture. However that is using one of Lord Smurf's recommended DVD recorders.
When dealing with VHS to DVD stuff, found that recordings done in 4 LP mode vs Half D1 recordings, would rather deal with the Half D1 recordings. Please understand, I get hundreds of recordings in from different sources.
However pretty much no matter how bad the source tape, to get the best results, need the original VHS tape.
(Just did a full restore job on a 30 year old VHS tape, the results turned out amazing cause we had the source tape and set it up correct using the best possible hardware for the recording, if this was done any other way, the recording would have been horrible. Than having to deal with many of the issues which are written up on a daly basis in these forums)
As far as picture tearing at the top of the screen. This is usually caused by the TBC in the VCR or the VCR. It doesn't mean the tape is bad or anything wrong with the tape. It is best to fix this in hardware. You don't need to run scripts on the video or crop the video to fix this problem.
Have read about the ES10 several times in these forums, however never purchased or tested out the unit.Last edited by Deter; 28th Mar 2012 at 11:41.
-
Unfortunately, this part isn't true. You DO need a proc amp if you're using the DVDR3475 for passthrough. Software controls are useless after the signal passes through, because the DVDR3475 causes flickering and internal clipping when the input gain is too high. It doesn't happen all the time, but it will happen in brighter scenes. If you reduce the contrast afterwards in software, you'll reduce the white point, but the flickering will still be there, and the signal will still be irrecoverably clipped (there will be a spike in the histogram where all of the higher luma values were squeezed into...). Once the signal enters the DVDR3475, it's too late to prevent the problems it causes.Originally Posted by robjv1
You MAY be able to get away with using the DVDR3475 without a proc amp if your tapes are recorded within certain levels or VCR outputs unusually low levels, but...I wouldn't count on it if you're coming anywhere near camcorder home movies.
It does. The Vidicraft proc amp is pretty small and cheap though, and even combined with the DVDR3475 it will likely be cheaper than most of your other options. You can try the DVDR3475 without it, but don't say I didn't warn you...it will give you flickering and [internal, irreversible] clipping in bright scenes if you don't reduce the gain first.Originally Posted by videeo
Mini-Me mentioned the DVD3475 in a previous post, and also that it needs to run with a proc amp? I'd like to minimize the additional gear, especially if I have to start hunting for lost remote controls.
You're probably right here: With LP mode, full-D1 input resolution would likely cause too much blocking or blurring from the encoder to retain even any of the extra bandwidth.Originally Posted by Deter
Just a technical note though: Don't put too much stock in the idea that VHS is "only 240 lines." Even if that's the case, a significantly higher sample rate may still be required to capture all of it.
First though, I'm not sure if the limit is really 240 lines: VHS uses about 3MHz of bandwidth for luma, and chroma information is encoded in the slightly higher frequencies. NTSC video has 29.97 FPS, and each frame has 525 lines. The line length is therefore 63.555 microseconds, of which 52.6555 microseconds are part of the active window. Therefore, each line contains 3MHz * 52.6555 microseconds = 157.97 sine wave cycles of the highest representable frequency. Let's round up to 158, which indicates a max of around 316 distinct lines of horizontal resolution (because each oscillation has a peak and a valley) rather than 240.
Moreover, you can't adequately sample 316 lines just by taking 316 samples. Let's pretend for a second we're living in an ideal world, our analog signal is totally bandlimited (it's not; real-world data doesn't work that way) and infinite in both directions for each line (obviously impossible), and that we can use an ideal sinc filter with an infinite window to reconstruct it from the samples. In that case, Nyquist limit would indicate 316 horizontal samples should be sufficient to faithfully reproduce an original waveform with a length of 158 cycles. I assume that wouldn't be the case if we sampled at EXACTLY 2x the maximum frequency and our samples all landed at the zero crossings...but let's ignore that.
Here's the catch: Sampling theory indicates that the Nyquist rate is insufficient for reconstructing real-world video signals, because ALL of the ideal assumptions behind it are wrong. Our video signals are not bandlimited, would which would make an infinite-window sinc filter give the wrong results...and we don't have the processing power for that anyway. For that matter, our lines are finite in length, so the sinc filter couldn't even get an infinite window to work on if it had all of eternity to work on it.
Long story short: We have to compromise with less "ideal" reconstruction filters that work better for real-world data, and these filters rely on only a small window of samples. Because of this, we need significantly more than 316 horizontal samples to reliably reconstruct a VHS signal, and bumping up the sampling frequency by just a few percent (up to 352 or 360 samples) is very conservative/optimistic. Since PC comparisons have shown 352x480 and 360x480 resolutions to capture significantly less detail than 704x480 and 720x480, it would seem we need significantly more than 352 or 360 horizontal samples to capture the whole resolution...and the noisier the signal is, the more you need. Using 704 or 720 might be overkill, but I wouldn't be surprised if the actual number of samples required to capture full detail were closer to full D1 resolution than half D1 resolution.
If you're directly recording to DVD though, the whole game changes, because I'd agree that your biggest worry will often become the limited bitrate.Last edited by Mini-Me; 3rd Apr 2012 at 01:30.
-
Deter, downloading and watching the sampled and fixed videos is an invaluable aid in understanding what's being done in this thread.
Last edited by sanlyn; 21st Mar 2014 at 18:52.
-
So, videeo, put all these scripts together on tst4.avi, and it starts looking OK. I don't know why some shots in there look soft, others are sharp (good ol' VHS!). I'd say it's a good candidate for the restoration area. For those who don;t know what's going on with field-replace script in post #6:
Original frame 0 of test4.avi, no processing:
Using SeparateFields(), this is Frame 0's TopField:
Frame 0 Bottom Field:
Bob the frames (you could use Yadif, as well), and replace top with bottom:
Maybe a couple weeks of worth on overall PQ. Here's a quick effort. Still some work to do, but I think it'll float.
Last edited by sanlyn; 28th Mar 2012 at 15:38.
-
sanlyn, I was in the process of posting, but before I could send, the power went out for 2 seconds and I lost my post... What were your VirtualDub filter settings?
Here is my attempt at combining the different scripts to get the video IVTC and the jitter automatically fixed:
I disable the AviSynth filters so I could play around with some VirtualDub ones. Here is a sample:Code:LoadCplugin("C:\Program Files\AviSynth 2.5\plugins\yadif.dll") Import("C:\Program Files\AviSynth 2.5\plugins\stab.avs") AviSource("G:\VIDEO\test4.avi") ## ------------------------------------------------------------- ## Input is 720x470. Add 10 pixels for a border to get 720x480 ## ------------------------------------------------------------- AddBorders(0,2,0,8) ConvertToYV12(interlaced=false) #Crop(0,0,0,-4) #<- changed Crop(0,0,0,-8) ConvertToYV12(interlaced=false) AssumeTFF() yadif(order=1,mode=1) Stab() SeparateFields().SelectEvery(4,0,3).Weave() #Crop(0,4,-2,0) #<- changed Crop(0,4,-2,-2) AddBorders(2,4,0,10) #<- changed #- ------------------------------------------------------- # Inverse telecine/decimate. Result is progressive video #- in its original state as 23.97 fps movie film #- ------------------------------------------------------- AssumeTFF().TFM() TDecimate() #- ------------------------------------------------------- # SmoothLevels doesn't really change levels here, but helps #- smooth sharp RGB "peaks" that result in subtle banding. #- ------------------------------------------------------- #SmoothLevels() #MCTemporalDenoise(settings="low",enhance=true) #gradfun2dbmod(thr=1.8,str=0.9,strC=0.9,mask=false) #LSFMod() #AddGrainC(var=2.0,uvar=3.0) #- ------------------------------------------------------- # RGB conversion for VirtualDub. Omit if not wanted #- and then save as YV12. #- ------------------------------------------------------- #ConvertToRGB32(matrix="Rec601",interlaced=false)
test4-jumpy-autofix-IVTC.avi
There are still some artifacts in there, but I think it's on its way to becoming a good transfer. Thanks for the help along the way. I ended up buying an ES20 to see if I can get some cleaner signal to my capture card. I'll report the results once I get to try it out. -
I spent an hour looking for a specific post from a new member who was trying to find out if his Magnavox combo had TBC functions. A comparison video was posted (VCR with no pass-thru, VCR with pass-thru) and his unit appeared to clean a few problems line-timing. I can't find the damn post(!). But his wasn't a $90 cheapie, it was four times that price. I wouldn't vouch for any cheapo recorder, they have to cut corners somewhere. Nor do the newest products compete with the older ones (pre-2006) mentioned. New products that do have TBC pass-thru of some kind are not in the same league with earlier models.Originally Posted by videeo
And there's this sweeping statement: https://forum.videohelp.com/threads/343156-How-do-I-ajust-the-sharp-softness-on-my-Pana...=1#post2138484
That post is followed in the same thread by one of the resident experts here who has been playing with every video machine made (apparently) for many years. Discussion of line and frame tbc's with recommendations and caveats, including pro and con concerning Panasonic/Toshiba, about three paragraphs down: https://forum.videohelp.com/newreply.php?do=newreply&p=2138515
You can't use a TBC-equipped VCR as pass-thru. Which way you go is up to you, but the old VCR's had TBC and DNR circuitry that's primitive compared to later DVD machines. Consider that with a pass-thru unit you're not limited to which VCR you can use. Consider that these VCR's have mostly been used to death and no longer have parts or repair support. Consider that newer DVD/VCR combos are no competition for pre-2006 DVD recorders. Certainly, their VHS player sections are terrible anyway, so you'd pay for something you're not likely to use. And as you noted previously, newer combo's have some odd circuitry: your combo's s-video output works only DVD playback, which makes no sense to me. But, yeah, it's probably cheaper to make 'em that way.Originally Posted by videeo
Last edited by sanlyn; 21st Mar 2014 at 18:54.
-
I didn't use VirtualDub filters on test4.avi. No Avisynth plugins, either. The last video I put thru VirtualDub and color work was 2012032300633703_c. The settings for the filters are in the ,vcf file, posted earlier here. Settings for filters like gradation curves are many and complex, but can be imported with a .vcf file. The vcf doesn't just load the filters, it loads their settings as well. Careful, though, if you already have some filters loaded in VirtualDub and you use "Load processing settings..." to load a vcf, it loads the filters but removes filters already there. Bummer.Originally Posted by videeo
The same filters might work with test4.avi (is this from the same video with the space ship shot? The kids are wearing the same clothes). But as I mentioned, you can't use exactly the same filters and corrections for every video. Each video has its own color balance, noise, etc., and with tape you even get wide variations scene by scene.
Good show on the ES20. Mine still works (but it's my second copy). Use that and s-video, you'll see improvements.Originally Posted by videeo
Here is my attempt at combining the different scripts to get the video IVTC and the jitter automatically fixed:
[Attachment 11650 - Click to enlarge]
There are still some artifacts in there, but I think it's on its way to becoming a good transfer. Thanks for the help along the way. I ended up buying an ES20 to see if I can get some cleaner signal to my capture card. I'll report the results once I get to try it out.
Looks as if you have the script running OK. The repair functions didn't address noise or artifacts. It will take some playing around with plugins, usually chosen to address specific problems. MCTemporalDenoise is a multipurpose job, but like most such things it still leaves specific areas to be touched-up. I'd try that by itself first, then check what's left. I didn't see many obvious artifacts in the video, but there's some chroma bleed and what looks like fine-grained dot crawl. S-video, tbc, and MCTD should help with some of that, but nothing can catch all of it.
Time to put that ES20 thru its paces.Last edited by sanlyn; 21st Mar 2014 at 19:25.
-
Good catch there. I thought it just had some issues with the black level -- I wasn't aware about the issue with the internal sensitivity to the gain. Are the others recorders in that line afflicted with the same problem?Originally Posted by Mini-Me
Unfortunately, this part isn't true. You DO need a proc amp if you're using the DVDR3475 for passthrough. Software controls are useless after the signal passes through, because the DVDR3475 causes flickering and internal clipping when the input gain is too high. It doesn't happen all the time, but it will happen in brighter scenes. If you reduce the contrast afterwards in software, you'll reduce the white point, but the flickering will still be there, and the signal will still be irrecoverably clipped (there will be a spike in the histogram where all of the higher luma values were squeezed into...). Once the signal enters the DVDR3475, it's too late to prevent the problems it causes.Originally Posted by robjv1
-
Did you do anything different to this last jpg you posted?Originally Posted by sanlyn
[Attachment 11649 - Click to enlarge]
If I use the rev "C" AviSynth filters + "C" VirtualDub filters, I get a more blown-out background, where a lot of detail is lost:Last edited by videeo; 28th Mar 2012 at 15:57.
-
The more expensive (HDD-equipped) DVDR3575 has the same problem for sure. Desperado2 from Avsforum was using his as a DVD recorder, and he made several posts about the flicker...in fact, he's the one who first realized how to fix it. The DVDR3576 is almost identical to the DVDR3575 too, so I have no doubt they share the same circuits.Originally Posted by robjv1
Odds are, any unit without the problem will also have different TBC circuits too. It's a bother, but at least the fix is simple enough that a Vidicraft proc amp can handle it. It's worth it though IMO, because it gave me the option of using whatever VCR I wanted. (Then again, I should be receiving a W7U tomorrow or the next day, which I'm hoping will make everything else redundant. We'll see.)
As Desperado2's post showed, you don't technically need a dedicated proc amp to fix the flicker/clipping...just something to reduce the gain before the signal enters the DVDR3475. In his case, a Sima unit worked to eliminate the flicker. However, using that might also prevent the DVDR3475 from functioning as a TBC: I'm not sure how much the Sima affects the signal beyond stripping Macrovision, but I know my AVT-8710 rewrites all of the sync information, so it prevents the DVDR3475 from being able to make any correction.Last edited by Mini-Me; 28th Mar 2012 at 16:23.
-
I mounted the video in VirtualDub and fiddled with gradation curves for about 5 minutes. Even at that, those shots need more work. The scenes that follow it wouldn't use exactly the same filters.Originally Posted by videeo
Did you do anything different to this last jpg you posted?Originally Posted by sanlyn
[Attachment 11649 - Click to enlarge]
If I use the rev "C" AviSynth filters + "C" VirtualDub filters, I get a more blown-out background, where a lot of detail is lost:
You can't use the same filters for every video. Often, not even for every scene in a video. Tape playback isn't that accurate or stable. And denoisers and other filters can alter contrast and colors.
This kind of detail is restoration, not capture. Once capture problems themselves are solved and your routines are set, then frame repair, ivtc, jitter flicker, denoising, fixing artifacts, color and such are a different matter. What we've been doing after capture so far have been fairly simple adjustments to see if a video is worth the trouble - to ascertain how much more has to be done. Tape captures will never look like DVD recordings off your digital cable box. They look like captures from tape. Bringing those imperfect captures up to snuff for digital display is what has to happen with old media.
The basic color balance in the jpg is about right, actually. But remember that luminance adjustments are also a part of those "C" filters, designed for a different scene and a different capture. Takes a while to learn to fiddle with color filters. But try this:
Add an extra ColorMill filter to the filters you already have for this scene. You can mount more than one copy of the same filter. In ColorMill, turn on the preview button. On the left of ColorMill's window, click the "Levels" button. Pull the "Darks" slider down about 8 points. Pull the "Middle" slider down 8 points. Leave "Light" as-is. You'll see the mids and lows start to shape up.
Click the "Saturation" button. Raise all three sliders on the right upward, about 12 to 15 points. What you're doing with ColorMill is modifying the effects of the other filters. If you had set up some filters more suitable for these low-contrast low-saturation shots in Test4.avi, you would keep the same filters most of the time but add one or two "tweaks" to modify scenes that require it.
But the basic principle is: you can't use the same filters for everything. Everything doesn't look alike. Especially tape.Last edited by sanlyn; 21st Mar 2014 at 19:25.
-
Understood...I am trying to figure out how you got the improved color without blowing-out the background (well it's a little blown-out on the upper-left grassy area), because I have been trying to figure out how to do it the past couple of days. If you happened to save the settings, it would be great to take a look at your .vcf file. With your recommended addition of Colormill tweaks on top of the rev C filters, it still is a bit off.Originally Posted by sanlyn
Here's another try:
-
That background is blown out no matter what. The bright sunlight back there and the darker shadows under the tree -- it's too wide a contrast range for video. In real life your eyes (and your brain) adjust to that range, but film/video can't do it.
I didn't keep the settings for that single image, mainly because I knew it wasn't quite right and because they wouldn't work in the next couple of shots. Later tonight I'll work with it in more detail. But you're in restoration now, babe. This has nothing to do with capture.Last edited by sanlyn; 21st Mar 2014 at 19:26.
-
Very interesting. I have the DVDR3575 in my living room, but I only hooked it up once in my setup here to see how it performed as a recording deck with VHS material, but I am sure I had my proc amp in the chain. Now I need to bring it back in here to test it for it's ability to correct geometric distortions which somehow I totally missed the boat onOriginally Posted by Mini-Me
The more expensive (HDD-equipped) DVDR3575 has the same problem for sure. Desperado2 from Avsforum was using his as a DVD recorder, and he made several posts about the flicker...in fact, he's the one who first realized how to fix it. The DVDR3576 is almost identical to the DVDR3575 too, so I have no doubt they share the same circuits.Originally Posted by robjv1
Odds are, any unit without the problem will also have different TBC circuits too. It's a bother, but at least the fix is simple enough that a Vidicraft proc amp can handle it. It's worth it though IMO, because it gave me the option of using whatever VCR I wanted. (Then again, I should be receiving a W7U tomorrow or the next day, which I'm hoping will make everything else redundant. We'll see.)
As Desperado2's post showed, you don't technically need a dedicated proc amp to fix the flicker/clipping...just something to reduce the gain before the signal enters the DVDR3475. In his case, a Sima unit worked to eliminate the flicker. However, using that might also prevent the DVDR3475 from functioning as a TBC: I'm not sure how much the Sima affects the signal beyond stripping Macrovision, but I know my AVT-8710 rewrites all of the sync information, so it prevents the DVDR3475 from being able to make any correction.
That's right! I forgot I was supposed to check out my SR-W5U and the histogram for you. Did you win that auction on eBay that seemed to be mislabeled (it seems they couldn't decide if they were selling an SR-W5U or SR-W7U, they had both listed in the description). You'll have to post some videos and your impressions of the deck. Are you receiving any manuals with it by chance? I posted the SR-W5U and SR-W5U service manual up at DigitalFaq, but nobody seems to have come across the equivalents for the SR-W7U. -
Originally Posted by robjv1
Very interesting. I have the DVDR3575 in my living room, but I only hooked it up once in my setup here to see how it performed as a recording deck with VHS material, but I am sure I had my proc amp in the chain. Now I need to bring it back in here to test it for it's ability to correct geometric distortions which somehow I totally missed the boat on
That's right! I forgot I was supposed to check out my SR-W5U and the histogram for you. Did you win that auction on eBay that seemed to be mislabeled (it seems they couldn't decide if they were selling an SR-W5U or SR-W7U, they had both listed in the description). You'll have to post some videos and your impressions of the deck. Are you receiving any manuals with it by chance? I posted the SR-W5U and SR-W5U service manual up at DigitalFaq, but nobody seems to have come across the equivalents for the SR-W7U.
I actually managed to find my W7U on Craigslist for $337 shipped..a great price for a W5U, let alone a W7U! I just hope it's in good condition...it's supposed to be, but the previous owner only used it to demo WVHS on 42" plasmas for customers, and he never tried with VHS. He seems very much on the up-and-up though. He was forthcoming about the WVHS tape having a spotty section, but he was convinced it was just the tape/recording, since it was fixed by recording over it a couple times. He was also happy to overpack. Unfortunately, he didn't mention anything about a manual, and I didn't think to ask. I'll assume it doesn't come with one unless it magically appears in the box...which will hopefully arrive within the hour.
(EDIT: Scratch that: The mailman didn't come bearing gifts today, so I'll have to wait for tomorrow. 
) It's unfortunate that those manuals are so hard to find, but I'll be thrilled just to get a unit without issues.
I also tried to snag a W5U in the auction that ended Monday so I could capture two tapes at a time, and I grabbed a spare All In Wonder card just in case. It's not imperative that I get a second unit ASAP though, so I was the lowest of the three serious bidders at just over $500. It sold for $586.99 plus $15-something shipping despite an inability to question the seller for details...likely because of the generous return policy (guaranteed working condition, 7-day returns for item exchange, or money back if no exchange possible). It seems like the average price has crept back up into the $450-600 range for good units with some form of return policy, but we'll see if that lasts.
There's still an open ebay listing for a W5U though. It's the one with both W5U and W7U in the description (to increase search hits). The seller claims it's in working condition but says "Sold as-is due to its age,no returns." She also originally listed at a Buy It Now price of $1200 I think, PLUS $42.50 shipping. How outrageous is that? She seriously asked for by far the highest amount any W5U has sold for in years and years (AFAIK), for a unit that might as well be described as "probably broken?"
She's been gradually dropping the price in $100 increments, then $50 increments, now $25 increments, and it looks like she's now created a parallel listing on New York's Craigslist for $850 as well (http://newyork.craigslist.org/mnh/pho/2928231438.html - same photo, location, and greed, with "excellent condition" replacing the explicit "as-is" wording). I think she's surprised it hasn't sold yet, since one sold so recently for over $800 without an official return policy...but that seller privately guaranteed excellent condition in ebay-logged private messages (specifically citing ebay's "as described" policy), and it was bought new, used only three times by a company, and stored in a climate-controlled room. That one was a sure win, whereas this one has the fewest guarantees (or room for recourse) I've seen in a while, particularly for such a high price. I haven't messaged her to ask for more details, but "as-is...no returns" can't get much clearer. (Granted, those are the implicit conditions of any Craigslist sale including mine, but it's more of a red flag when it's explicitly stated on ebay.) All that said, she does have great feedback, and she's sold expensive "as-is" items before...so maybe it's working after all. It's still a gamble though.
As a side note, I also love how some of these sellers make it out like they're not asking much for a unit with a > $5000 MSRP...when the vast majority of these units were apparently brought to the US brand new/factory-refurbished for $949 (W5U) and $999 (W7U), and many (most?) have become broken since then.
Last edited by Mini-Me; 29th Mar 2012 at 11:24.
-
videeo: had a PC repair job last nite, bot back mighty late.
The JPG I posted of frame 0 with the quick color work is too red. I toned it down a bit. A vcf to accompany the script for Test4_new1 is attached below.
I combined my scripts as you did, but usually I run scripts like this as separate runs. It goes faster, and if I change something I don't have to run the whole thing over again. This script runs sloooow with all the steps in one. The AVS below has different routines than the one you posted earlier. It's klutzy, I reorganizerd it rather quickly. But I copied your path/filenames into the one shown here:
Test4_New1.avs:
The .VirtualDub vcf calls for ColorMill, which I think you have, and uses VDub's built-in sharpen filter at low default. It also uses Donald's Graft's GUI-enhanced Bright/Contrast plugin, which I have attached below. Open the AVS in VirtualDUb, and before you hit "Save as Avi..." load the .vcf file. You can usually run avisynth and VirtualDub filters in the same run.Code:LoadCplugin("K:\AviSynth 2.5\plugins\yadif.dll") Import("K:\AviSynth 2.5\plugins\stab.avs") AviSource("J:\forum\videeo\test4\test4.avi") ## ------------------------------------------------------------- ## Input is 720x470. Add 10 pixels for a border to get 720x480 ## ------------------------------------------------------------- AddBorders(0,2,0,8) ConvertToYV12(interlaced=false) Crop(0,0,0,-8) AssumeTFF() yadif(order=1,mode=1) Stab() SeparateFields().SelectEvery(4,0,3).Weave() Crop(0,4,-2,-2) AddBorders(2,4,0,10) #- ------------------------------------------------------- # Inverse telecine/decimate. Result is progressive video #- in its original state as 23.97 fps movie film #- ------------------------------------------------------- AssumeTFF().TFM() TDecimate() #------------------------------------ # replace bad field in frame 0 #------------------------------------ file1=last file2=file1 file2 Trim(0,-15) AssumeTFF().yadif(order=1,mode=1) FreezeFrame(1,1,0) #-- ------------------re-interlace ---------------------- SeparateFields().SelectEvery(4,0,3).Weave() file3=last # --- replace only one frame in original source --------- file4=ReplaceFramesSimple(file1,file3, mappings="0") file4 ColorYUV(gain_v=-5,gain_u=-5) MCTemporalDenoise(settings="Low",enhance=true,edgeclean=true,useTTmpSm=true) LSFMod(125) AddGrainC(var=2.0,uvar=3.0) #- ------------------------------------------------------- # RGB conversion for VirtualDub. Omit if not wanted #- and then save as YV12. #- ------------------------------------------------------- ConvertToRGB32(matrix="Rec601",interlaced=false) return last
In the post that follows this I'll show you why the filters for the first part of that clip won't work for the other scenes -- and why I think you should be using your old RCA combo for this video to get much better (and easier) results..Last edited by sanlyn; 21st Mar 2014 at 19:27. Reason: yet another stupid typo
-
This Frame 0 in the test4_new1.avi you get from the above script. All the frames shown ehre are resized to display size 640x480:
The filter settings in the .vcf were set specifically for this scene, which doesn't look the same scene in later shots. It has different luma levels and color balance. The original looks washed out -- but so do all the other scenes in the original capture. This is how those same filters look in other shots from the same clip: (filtered frames 107, 183, 451, 509):
I think you can see what I'm talking about. This is commonly the way VHS plays back. With the filters for scene #1, subsequent scenes are too dark and saturated, and red and blue look like neon. But I think the VCR you used for this capture isn't helping a bit. Here are two frames from the original test4.avi capture, before any processing:
The original park bench scene looks washed out because the VCR's contrast range is so anemic. At the same time, red and blue look oversaturated; they tend to pop off the screen. The kid on the right has pale-looking blue arms. If you increase saturation and contrast to get a more natural look, those two colors look even worse. The image on the left of the kid in denim looks horrible. There's chroma nosie everywhere in red and blue. ANd here';s a closeup of what that VCR does to shadows (it did the same thing to the earlier 20120323000633703.avi). This edited frame is enlarged 2X with bicubic resize:
Those skin shadows are just plain rough and ugly, and really difficult to work with. You also have edge artifacts. These colors are really unnatural, even for a scene shot in the shadow of a tree. In that earlier AVI of this same video with flying saucer shot, this rough blue chroma noise in skin shadows even showed up in warm daylight shots. I didn't see obvious effects like these in the capture of similar scenes that you made with your old RCA. In fact that RCA capture would have required very little correction.Last edited by sanlyn; 29th Mar 2012 at 13:57.
Quote
Similar Threads
-
DVHS VCR performance with EP speed tapes and VHS-C tapes?
By NJRoadfan in forum RestorationReplies: 13Last Post: 8th Apr 2011, 20:36 -
Modify CCD-TRV70 to play older Hi8 tapes
By bkrelectric in forum Newbie / General discussionsReplies: 0Last Post: 11th Oct 2009, 09:41 -
Newbie question: cannot view/capture older DV tapes
By sinbarambam in forum Camcorders (DV/HDV/AVCHD/HD)Replies: 5Last Post: 9th Jun 2008, 07:48 -
Capturing Video 8-tapes
By Mållgan in forum Capturing and VCRReplies: 7Last Post: 26th Apr 2008, 03:07 -
New VCRs Have Trouble with Older Tapes
By homeboy in forum RestorationReplies: 15Last Post: 30th Mar 2008, 19:35