Well, VHS capturing Mostly cartoons, having some questions:)

[sanlyn]

Me i have worked with rgbeq a bit more, see attached files

Looks better, tm1. Thanks for the script (I keep samples of how rgbeq can be used). Check the bottom of the RGB histograms.

[sanlyn]

Heavens, I got two PM's on why the script replaces frames 12 and 13. Below are cropped images of a disturbance in those frames, blown up 2X from the original capture. RemoveSpots fixed it (I tried deVCR as well, but no go). RemoveSpots softened the clip too much for me, thanks to all the original noise, so I used only the two repaired frames.



Wouldn't you know it, those lines just had to be in the brightest part of the image!

[themaster1]

Looks better, tm1. Thanks for the script (I keep samples of how rgbeq can be used).

Perhaps you will prefer this one:

[sanlyn]

Thanks, again.

[zerowalker]

Oh, why haven´t i gotten a single mail, i just checked in to see if anyone had written anything, and i see like 5 posts;S
I will have to take a look at both themaster1 and sanlyn's script, i have hard time playing with colors, but you seem to be solving the noise and leaking pretty well.

Not sure what you were talking about with the 12-13 frame, nor what the upscaled picture is from (my clip??).
But i guess you did some software TBC or something?


And as i said, i am trying out the Dscaler driver, though it´s sadly just exiting randomly for some reason, but i have gotten some time to play around, and what i can´t figure out is, if i turn the Comb filter Off, there is less dot crawl then with it on. Isn´t it supposed to be the opposite?

I have recorded with it off, though it´s not exactly the same look as my precious capture, as this has more control than Virtualdub, so it´s getting more into the picture, not sure how that works, but i get more width with dscaler (not resolution but from the vhs) then with virtualdub.

Here is two samples, one that is my previous Rewinded clip but a different scene, and one i recorded just now with the Comb Filter off.
Which looks the best, and what are your thoughts.

Thanks

[jagabo]

Are you sure those are named correctly? The "comb filter off" version is much cleaner.

[themaster1]

I will have to take a look at both themaster1 and sanlyn's script

Mine give you this (mpeg2 6Mb):
http://www.mediafire.com/?l1knh46x79xgh3p

[zerowalker]

Are you sure those are named correctly? The "comb filter off" version is much cleaner.

Yes, that´s what i don´t understand;S

Cause in the settings i can choose:

Comb Filter

Chroma Only
Default
Full
Off

On composite Default is "Full" it seems.

So i looked at the picture and changed between Full and Off, and much if not all the weird dot crawl disappeared with it off;S
But i have played with it before on S-Video (the one with bad Y/C), and there the Default is Chroma Only, if i set Full there, i get Alot of dot crawl, if i set Off it goes away, but i get Dot Crawl on some colors, for example Color Bard Blue/Red between them.
So i have to have it on Chroma Only, and i did the same on the Composite clip above, as i didn´t see any difference between off, and chroma only.


The only thing i can think of, is that the Comb Filter is extremely bad and make more misstakes then use.
Or that there is a Comb Filter that´s on by hardware default and that the drivers messing with it turns it bad or something;S



But jagabo, why can i get more stuff into the picture compared to Virtualdub. I know it´s something with Overscan and stuff, but is there really parts of the VHS that´s recorded, but isn´t supposed to be showing?
Does that mean that it has more width than 720 for it´s full use?



Themaster1, thanks will be looking at it and give feeback when i got time to read and understand your script

[sanlyn]

Oh, why haven´t i gotten a single mail, i just checked in to see if anyone had written anything, and i see like 5 posts;S

I wouldn't worry about it. Since I came here in 2004, listing every one of my PM's wouldn't fill my monitor's screen. Probably not more than 20 in 8 years.

I will have to take a look at both themaster1 and sanlyn's script, i have hard time playing with colors, but you seem to be solving the noise and leaking pretty well.

Chroma bleed is still there, but looks minimized by shifting chroma a bit with FlaXen's "VHS" virtualdub filter. I guess there are some good ideas for fixing chroma bleed in doom9 somewhere, but I haven't found a plugin that helps very much. Suggestions welcome.

Not sure what you were talking about with the 12-13 frame, nor what the upscaled picture is from (my clip??).
But i guess you did some software TBC or something?

I described the two images in post #542 as follows:

Below are cropped images of a disturbance in those frames, blown up 2X from the original capture.

I used the Avisynth function Remove Spots() that I got from jagabo. It's a modification of RemoveDirt() https://forum.videohelp.com/threads/323988-Looking-for-a-filter-to-mask-out-white-specs...=1#post2006402 . QTGMC also helped clean up those frames, which were 12 thru 14.

Below is a crop of the area in question from frame 13. The original is on top, my submission below it. Note the dark horizontal lines acorss the bright part in the middle of the image. Also note horizontal stripes in the orange toy object just above it.



There have been some attempts at designing softweare TBC, but none that work yet. Usually a tbc won't fix the problem.

what i can´t figure out is, if i turn the Comb filter Off, there is less dot crawl then with it on. Isn´t it supposed to be the opposite?

I have recorded with it off, though it´s not exactly the same look as my precious capture, as this has more control than Virtualdub, so it´s getting more into the picture, not sure how that works, but i get more width with dscaler (not resolution but from the vhs) then with virtualdub.

Here is two samples, one that is my previous Rewinded clip but a different scene, and one i recorded just now with the Comb Filter off.
Which looks the best, and what are your thoughts.

Thanks

I noticed the difference in image content, too. Below are caps of frame 68: original on top, Combfilter_Off on bottom. Comparing the two images, it appears that the top image has had about 16 pixels cropped off the left and right border. In the combfilter_off image below it, the frame contains more of the image and the clock and the figure are more slender. The bottom image also contains the right-hand black border pixels from the original source, which has some dot crawl along its edge - not unusual along VHS borders.

[zerowalker]

Probably true, but i had gotten the first posts, but i seem to have just clicked on them, then forgetting about it;S

Yeah well the Chroma bleed is pretty severe, mostly on bright parts and alone colors, and it´s true, it´s oversaturated, and weird on some parts in the colors.

But i must say that i think the original (without the chroma stripes), looks more correct, as to distinguishing parts in the bright clothes of the baby, along with the floor.

The floor looks more natural with the yellowish white, sadly i am not sure if it´s supposed to be Truly white or not, but i am pretty sure that the clothes are supposed to be white, but sadly the tape, along with the camera didn´t make a good match.
But as long as the details remain, some bad coloring isn´t that much a problem for me, though chroma leak is something i would like to solve, atleast on parts where it destroys things around the object.


I am confident that i did Not name them wrong.
As the one with it On, is the one i recorded after rewinding.
I recorded with Virtualdub with the original Directshow filters, and there i can´t even choose to turn it off or on, it´s on by default when there's a composite input.

The other one is made in Dscaler, with Dscalers Driver, and there i can change some settings.

But i wonder, how much is supposed to be captured?
As i can move the picture to the left and right, and get some more width (not just resize) if i choose 768 in Dscaler.
But as you see in that picture, the right get´s the Fading Black edge, while the left (which isn´t in the captured file) get a bright vertical line, then there's a black border where you can see the Lines stuttering from the picture.

[sanlyn]

I will have to take a look at both themaster1 and sanlyn's script

Mine give you this (mpeg2 6Mb):
http://www.mediafire.com/?l1knh46x79xgh3p

Colors on the carpet and child coming along well, but how can your script be modified to address the flutter in the upper left and the high noise level in red?

[themaster1]

See attached file if that can help you.
There is room for improvement in this script , the 2x rgb convertion for instance could be reduced to one but it imply to rework on rgbeq/cmill which i will not do...

[jagabo]

But jagabo, why can i get more stuff into the picture compared to Virtualdub. I know it´s something with Overscan and stuff, but is there really parts of the VHS that´s recorded, but isn´t supposed to be showing?

I believe descaler is using their own driver and changing the capture window to include more at the left and right edges of the picture. The BT spec for a 720 wide image specifies that the actual video data is in a 704 pixel wide window with 8 pixels of padding at the left and right edges. In practice the actual picture may be slightly shifted right or left, hence the padding. Capture cards that follow that spec usually deliver obvious black borders at the edge of the frame. See the image in this post: https://forum.videohelp.com/threads/346000-VHS-artifacts?p=2161081&viewfull=1#post2161081 I suspect your capture card's drivers aren't following the BT spec and are delivering only the center portion of the picture. Keep in mind that analog video has no pixels. It is a continuous waveform from left to right within the frame. It can be capture with as many or as few samples, pixels, as an application wants.

[themaster1]

Sorry i have read a little fast (euro soccer cup....) that is the script i used for the mpeg.

"How to address the flutter in red" -> What about fft3d but better fine tuned than in my script, it can target a specific frequency (i have used sigma 2 for all frequencies and UV grouped). You'd have to determine what frequency needs more filtering and work on V . Then if you pick say 2-3 frames radius (temporal) you should get something in the end...me believe

A good thing to do would be to use neatvideo noise sampler to have a precise idea of how much noise there really is

[jagabo]

Chroma bleed is still there, but looks minimized by shifting chroma a bit with FlaXen's "VHS" virtualdub filter. I guess there are some good ideas for fixing chroma bleed in doom9 somewhere, but I haven't found a plugin that helps very much. Suggestions welcome.

Something like MergeChroma(awarpsharp2(depth=24)) helps sharpen the chroma a bit. I wonder if downsizing the chroma horizontally, the using something like nnedi3 to upscale would work better.

[sanlyn]

Sorry i have read a little fast (euro soccer cup....) that is the script i used for the mpeg.

"How to address the flutter in red" -> What about fft3d but better fine tuned than in my script, it can target a specific frequency (i have used sigma 2 for all frequencies and UV grouped). You'd have to determine what frequency needs more filtering and work on V . Then if you pick say 2-3 frames radius (temporal) you should get something in the end...me believe

A good thing to do would be to use neatvideo noise sampler to have a precise idea of how much noise there really is

NeatVideo in my mpg "NV" version submitted earlier cleaned the red noise and tamed the corner flutter as well, without destroying the texture details (I think the man is wearing a velum sweater) -- I usually avoid strong settings with fft3D.

However, the reds in this video are rather a weird crimson. I'll take a look into FFT3D-by-frequency; red chroma noise in YUV is low-frequency noise, I believe. Looks like the VHS reds I used to see with Monster Cable wire. I think I'll try that later with jagabo's MergeChroma idea, though I'm not sure how to bring that odd red into line. It seems to affect skin tones, too. Maybe it's just the way the source is aging.

[sanlyn]

I guess there are some good ideas for fixing chroma bleed in doom9 somewhere, but I haven't found a plugin that helps very much. Suggestions welcome.

Something like MergeChroma(awarpsharp2(depth=24)) helps sharpen the chroma a bit. I wonder if downsizing the chroma horizontally, the using something like nnedi3 to upscale would work better.

Well, whaddya know, I was reading doom9 mergechroma posts on my breaks today and came across similar techniques.

I've added these and several other of zerowalker's clips to my "OMG Video" collection ("Oh my God, how do I fix this?"). I've seen posts from another member that has similarly odd red hues, oversaturated blues, and reddish blacks -- almost identical to these. I wonder if it's the VCR . . . whatever, most of these make good learning clips.

[themaster1]

(off topic)
Here is the reason why i love mergechroma+ awarpsharp:
(original / filtered / filtered+ mergechroma+awarpsharp
http://imageupload.org/en/file/229610/comparaison-final.jpg.html
If you zoom-in you should see the difference

[sanlyn]

Got it, thanks.

[zerowalker]

But jagabo, why can i get more stuff into the picture compared to Virtualdub. I know it´s something with Overscan and stuff, but is there really parts of the VHS that´s recorded, but isn´t supposed to be showing?

I believe descaler is using their own driver and changing the capture window to include more at the left and right edges of the picture. The BT spec for a 720 wide image specifies that the actual video data is in a 704 pixel wide window with 8 pixels of padding at the left and right edges. In practice the actual picture may be slightly shifted right or left, hence the padding. Capture cards that follow that spec usually deliver obvious black borders at the edge of the frame. See the image in this post: https://forum.videohelp.com/threads/346000-VHS-artifacts?p=2161081&viewfull=1#post2161081 I suspect your capture card's drivers aren't following the BT spec and are delivering only the center portion of the picture. Keep in mind that analog video has no pixels. It is a continuous waveform from left to right within the frame. It can be capture with as many or as few samples, pixels, as an application wants.

I remember that i did turn Clipping of, which added bars, or it it just removed some of the right and left, as i thought. Why would i want to remove it?

I have been playing around with my Xbox as it´s easier to work with a stable source than a VCR. And there i seem to get everything within the frame, no black bars(i think), atleast it doesn´t end, the image continues till the end of the frame.

But on the VCR it will disappear.

And yes i know that Analog doesn´t show pixels, so it will resize to anything etc.
But PAL and NTSC is supposed to have 720 "pixels" of width right?
So i shouldn´t capture more than that?
Or is it more than 720 width, but 720 is where the image is, and the rest is just black or jitter/noise?
As i really want to know as i don´t like to remove parts of the picture, even if the spec chooses an overscan or padding just to follow an aspect ratio or something.

As you see on the clips there is pretty much information on the image that is padded on the Virtualdub capture, but not the Dscaler.


sanlyn:

please link the one with the same problem, would be nice if it was the VCR or something.
And themaster1, i tried to learn a bit myself with mergechroma, but i don´t really get it, i do understand that it will try to fit the chroma inside the object using awarpsharp, but i think i had problems with awarpsharp and tried awarpsharp2, but i didn´t work well.

[jagabo]

But PAL and NTSC is supposed to have 720 "pixels" of width right?

PAL and NTSC what? For DVD you can use 720, 704, or 352 widths. Digital cable and satellite use a variety of widths, often in the 480 to 560 range. ITU 601 full D1 video uses 720 or 704. For 720 the actual picture is in a 704 subsection and there are 8 pixels of padding at the two sides. So the 704x480 picture is the 4:3 image, the full 720x480 frame is slightly wider than 4:3.

With DVD there is some debate or confusion about whether the 4:3 image in a 720 frame should be in the inner 704 pixels or the full 720 pixels. If you follow the letter of the law (DVD and MPEG 2 specs), the full 720x480 frame holds the 4:3 (or 16:9) image. But that differs from the ITU spec which says the inner 704x480 comprises the 4:3 image. The difference is generally ignored on commercial DVDs. Many movies use the full 720x480 frame. Most analog tape sourced DVDs are captured to the ITU spec (720x480 frame, 4:3 image in the inner 704x480 portion) and the pixel AR difference ignored.

[zerowalker]

Well, i meant that standard 720x576/480.

So the ITU spec says that there should be padding, or overscan to get the correct aspect ratio?
While otherwise the full size is 720, for general analogue capture?

I don´t want to capture with padding, as i don´t see the use in ignoring a part of the image just to follow some kind of spec, i rather capture all the information that can be captured, which is why i have problems understanding this.

But will will try taking a picture with the Dscaler driver later, with 720 and 768, along with clipping and without, cause if i remember correctly, i get more width with 768, and not resize, i get more analogue width, more information into the frame.

[jagabo]

The ITU padding doesn't ignore part of the picture. It captures all of the picture along with a little of the signal that's outside the active picture range.

[zerowalker]

Oh, so the clip from Dscaler does have ITU padding?
is the ITU when you get the entire picture, no underscan etc?


EDIT:

Okay, i seem to have been wrong about more width, it seems that with 768 it will prevent overscan.
Here are two picture.

You can see by the size difference, but the one with black bars is 720, then one without is 768.

EDIT 2:

Added two picture more, which one do you think has the most appropiate Brightness/Contrast?
I follow the Histogram to prevent it from Ever going to low or to high, along with White Crush.
But it seems to dark, but at the same time the bright parts seem to be more accurate in details.

[sanlyn]

Of the two images with the figures in the room: the top image has a more manageable luma range, which peaks at RGB 215 in the bright window. In the bottom image's window, the brights exceeds RGB 235 (it goes to RGB 254); you can see parts of the lower window's frame structure starting to "burn out", which on a TV will have the effect known as "blooming".

In both images, chroma is too low at the dark end, so shadows look crushed. If you remove all of the black letterbox borders to keep from throwing off histograms and scopes, the upper image of the figures in the room gives the waveform below. The top part of the 4 graphs can be pretty much ignored because of the bright window. The crushing at the lower borders of each section shows crushed dark colors, which makes the darkest shadows look somewhat dense.

[Attachment 12764 - Click to enlarge]

The darkest areas all look too red. If you look at the way the colors appear in an RGB histogram, you'll see why. The darkest shadows don't have enough blue and green, whose values are below zero-black in those areas. There are no clean blacks in the image. Note the small bl;ue coffee pot in the center of the picture. Read its pixel values, and you'll see that the "black" handle isn't black. But, then, maybe it's not supposed to be, I don't know. The only area of the image that's actually close to black are the letterbox borders, but they have some chroma noise that keeps them from reading as pure black.

Otherwise, the darkest darks aside, the images don't look "too dark" to me.

[zerowalker]

Okay, so the Bright is "correct".
Should i increase the brightness a step?
As i look at the Yuv values, and if i increase it one step, it will go from 16(which is the limit) to 17.
It will stay Stable at 17, but i don´t know if it´s stable at 16 or if it goes under it, as it doesn´t show anything below it.

But i guess it´s safer to be one step over the limit then under?


And yeah i guess the handle should be black, can´t see any reason for it to be any other color.
i am able to change:

Hue
Color
Color U
Color V

Is there any of them that can correct it?
And how do you recommend to correct it, as i can´t get a stable picture from the VCR except for the Blue menu with White text, and that isn´t to much help with colors.

The thing i can look at except the Y histogram, is the entire YUV range, Y,U,V.

[jagabo]

Oh, so the clip from Dscaler does have ITU padding?

I'm not sure. Usually with ITU capture there are ~8 pixel wide black borders at the left and right edges. But it's possible for picture information to appear in those borders, depending on the source. Your picture:

https://forum.videohelp.com/attachments/12761-1339703056/TV2012061421391200.png

Is about what you usually see with ITU caps. Your other caps don't seem to have those borders so the borders in that image might be something about the blue screen generator in your VHS deck. Try some other off air recordings. See if you get black borders with those. Other samples of ITU caps:

https://forum.videohelp.com/threads/346000-VHS-artifacts?p=2161081&viewfull=1#post2161081
https://forum.videohelp.com/threads/345350-VHS-Restoration?p=2155871&viewfull=1#post2155871
https://forum.videohelp.com/threads/345350-VHS-Restoration?p=2157210&viewfull=1#post2157210

is the ITU when you get the entire picture, no underscan etc?

ITU caps contain the full broadcast picture in the inner 704x480/576 portion of the 720x480/576 frame.

[zerowalker]

Okay, well i don´t have any paddings as you see, though the Blue menu on the VCR has black bars further into the picture then the Tape (maybe differ from tape to tape).

But i will just go with the standard and record 720x576/480 and let it show what´s supposed to show. No systematical black bars or anything. Not sure if it´s adding bars, as the images you link seem to have the H-lines jittering at the edges, which wouldn´t be there with a added black bar.

[sanlyn]

Okay, so the Bright is "correct".
Should i increase the brightness a step?
As i look at the Yuv values, and if i increase it one step, it will go from 16(which is the limit) to 17.
It will stay Stable at 17, but i don´t know if it´s stable at 16 or if it goes under it, as it doesn´t show anything below it.

But i guess it´s safer to be one step over the limit then under?

Yes, safer to be a little over, but not always necessary. Luma was OK, it's colors that are a bit dark. My critique was posted very late in the AM, I see that I somehow deleted part of my remarks. The part I didn't mean to leave out was that the levels were much better than earlier captures. You weren't far off the mark.

I'll give you more info to work with later today. Working on 2 PC repair customers now. Back later.

[zerowalker]

Ah okay, i think i have it stable know, but i still have problems with a setting called White Crush, everything will go dark with it on eventually, but with it off, the White will bleed everything in it´s presence.

So i have to keep it on, will do a pic with it on and off later.

And thanks, would be nice if i could get the Chroma tweaked, as Luma seems to be within the limits now.