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

[themaster1]

In avisynth matrices (which a Y'CbCr RGB conversion is occurring): Rec matrix means RGB 0,0,0 ; Y' 16 CbCr 16 ; and RGB 255,255,255 ; Y' 235 CbCr 240 PC matrix means RGB 0,0,0 , Y' 0 CbCr 0 ; and RGB 255,255,255 Y' 255 CbCr 255 The bottom line is Y' 16-235 CbCr 16-240 (with excursions)

Speaking of the 16-240 range for chroma components , any idea why when you generate colorbars in avisynth and analyse with coloryuv the maximum color intensity reported is 212 ? example: colorbars(width=720,height=480,pixel_type="YV12") ColorYUV(analyze=true)

[jagabo]

Because that is the max value of U and V in the chart:


U from the blue bar, V from the red bar.

[themaster1]

Can i conclude that when we're talking about 16-240 in YV12/rec601 it's false or you have something to defend that notion perhaps ?

I'm trying to understand where this 16-240 come from afterall

[jagabo]

Just because the valid range is 16-240 doesn't mean every image has to use that entire range. The test image only has a handful of colors.

You can convert the entire RGB cube (0-255) to YUV and see what the min and max resulting values are.

[sanlyn]

AGC should never be used during capture

I am talking about the White Crush there, are you telling me i should have it disabled?

AGC should never be used during capture.

So you recommend me to check out those and try to calibrate my screen?
How come it isn´t "Perfect" from the start?

Why would it be?

[themaster1]

i don't see which colors you're refering to jagabo frankly

[jagabo]

i don't see which colors you're refering to jagabo frankly

The blue bar has U=212, V=114. The red bar has U=100, V=212. 212 is the maximum value of U and V in that entire image.

[jagabo]

So you recommend me to check out those and try to calibrate my screen?
How come it isn´t "Perfect" from the start?

No display is perfect. Beyond that there is significant variation in manufacturing. And, for TVs, manufacturers pump up the colors and contrast to make their TV stand out in the crowd at the store.

[juhok]

"In an SMPTE color bar image, the top two-thirds of the television picture contain seven vertical bars of 75% intensity."

http://en.wikipedia.org/wiki/SMPTE_color_bars

[jagabo]

Just a note, intensity in that case refers to RGB values, not Y (as in YUV).

[poisondeathray]

Can i conclude that when we're talking about 16-240 in YV12/rec601 it's false or you have something to defend that notion perhaps ?

I'm trying to understand where this 16-240 come from afterall

As mentioned above, the SMPTE color bars are "75%" color bars. This is chosen on purpose because that's the limit of "broadcast safe" colors . Those coincide with the 75% amplitude markings on the vectorscope . Similarly 100% colorbars would coincide with the 100% amplitude markings on the vectorscope (not broadcast safe) . 100% color bars would show Cb and Cr 16-240

For 75% color bars, when converted to RGB values using Rec601, the R,G,B values should be
grey 191,191,191
yellow 191,191,0
cyan 0,191,191
green 0,191,0
magenta 191,0,191
red 191,0,0
blue 0,0,191


Similarly, for 100% color bars , the RGB values when converted with Rec601 would be the same as above, just replace the 191's with 255's .

[zerowalker]

AGC should never be used during capture

I am talking about the White Crush there, are you telling me i should have it disabled?

AGC should never be used during capture.

So you recommend me to check out those and try to calibrate my screen?
How come it isn´t "Perfect" from the start?

Why would it be?

But if i disable Whitecrush, it´s over brightened, and i can´t solve it, even if i decrease Contrast or Brightness, nothing happens as the image is already "destroyed".

And i looked up the calibration site, but i don´t get the Sharpness thing, it says it should look like boxes, but it doesn´t do that if i so take my sharpness to 0 (i got DVI).

[sanlyn]

Oh, yes, that sharpness test patch -- I think lagom should update that one. Today's monitors respond somewhat differently to that pattern. The contrast, bright, and gamma patches are the basic ones. That's where I first started, years back, with an older and cheaper LCD (and I mean really cheap!) that I've since replaced. Many consumer LCD's can't pass some of those tests, some of which require at least semi-pro performance or special controls. I included that site because it does have some basic contrast/brightness patches; the rest of it demonstrates how difficult it is to adjust monitors by eyeball, even with a test patch. On top of that, the color and gamma controls that come with consumer monitors (even very expensive ones) are not very sophisticated. But something like the EyeOne and its software, well, in 10 minutes or so it can make even a cheapie monitor much more usable (you'd be surprised at the differences).

I wouldn't get obsessive about it at this stage. It's the sort of thing that becomes more clear as you work with video min more detail. Anyway, you can't do everything at once.

I'm not sure what you mean by the "White crush" control. Is that what it's called (I can't say I've heard of that one). Or are you saying that you have an Auto Gain feature built-in somewhere and it can't be turned off?

[sanlyn]

As mentioned above, the SMPTE color bars are "75%" color bars. This is chosen on purpose because that's the limit of "broadcast safe" colors . Those coincide with the 75% amplitude markings on the vectorscope . Similarly 100% colorbars would coincide with the 100% amplitude markings on the vectorscope (not broadcast safe) . 100% color bars would show Cb and Cr 16-240
...........
Similarly, for 100% color bars , the RGB values when converted with Rec601 would be the same as above, just replace the 191's with 255's .

Good discussion, folks, makes my own PC and TV calibration discs and software make more sense. Tech sites advise that 75% test patches are preferred when adjusting a TV's primary and secondary saturation levels (if you have controls for that, which most consumer tv's don't have -- my plasma and many LG's have it). HCFR software works with several color systems; its online help gets touchy about making sure you've set up parameters for the system you're working with.

ED: But that's another story. I'm amazed that one has to go to TV forums to get info on PC monitor calibration. Someone who knows this subject more thoroughly would be welcome to post a guide here (simplified, of course, for dummies like myself). Oh, well...maybe it'll happen. Noise reduction and fixing bad frames is important, but so is making blue look like blue.

[zerowalker]

Oh, yes, that sharpness test patch -- I think lagom should update that one. Today's monitors respond somewhat differently to that pattern. The contrast, bright, and gamma patches are the basic ones. That's where I first started, years back, with an older and cheaper LCD (and I mean really cheap!) that I've since replaced. Many consumer LCD's can't pass some of those tests, some of which require at least semi-pro performance or special controls. I included that site because it does have some basic contrast/brightness patches; the rest of it demonstrates how difficult it is to adjust monitors by eyeball, even with a test patch. On top of that, the color and gamma controls that come with consumer monitors (even very expensive ones) are not very sophisticated. But something like the EyeOne and its software, well, in 10 minutes or so it can make even a cheapie monitor much more usable (you'd be surprised at the differences).

I wouldn't get obsessive about it at this stage. It's the sort of thing that becomes more clear as you work with video in more detail. You can't do everything at once.

I'm not sure what you mean by the "White crush" control. Is that what it's called (I can't say I've heard of that one). Or are you saying that you have an Auto Gain feature built-in somewhere and it can't be turned off?

Oh okay, well that explains why it doesn´t really work for me.
the pixel thing works do, my screen fails on 4b i think, it flickers.

But i will let it be for now i guess?

And the White Crush is a setting that i all Analogue Capture devices seem to have, but not all can Tweak it. (i may be wrong here).

What it does is, decreasing the White so it doesn´t destroy other parts in the Luma (as you may know).

It´s different from normal Bright/Contrast, though i don´t know how it works.

And yes i can turn it, off, on and adjust how it works (how fast it should decrease etc), though i don´t really like it, as it´s adaptive, i don´t really know what settings to use, as the default ones are really bad, if it changes to a bright seen fast, it won´t act at all, and the frames will be to white and the details are destroyed.

[sanlyn]

And the White Crush is a setting that i all Analogue Capture devices seem to have, but not all can Tweak it. (i may be wrong here).

What it does is, decreasing the White so it doesn´t destroy other parts in the Luma (as you may know).

It´s different from normal Bright/Contrast, though i don´t know how it works.

And yes i can turn it, off, on and adjust how it works (how fast it should decrease etc), though i don´t really like it, as it´s adaptive, i don´t really know what settings to use, as the default ones are really bad, if it changes to a bright seen fast, it won´t act at all, and the frames will be to white and the details are destroyed.

That's why you don't use AGC controls during capture. All analog capture devices don't have them; the better ones don't.

You find out what the Brightness/Contrast controls are doing by using it and watching how the image changes. They work somewhat differently on different devices, but here is the overall principle:
Brightness is used, ironically, to control black levels. If you raise the brightness control until deep blacks look light gray, then brightness is set too high.
Contrast is use to control white levels.
The two controls do interact somewhat; you have jiggle each back and forth a bit to get them balanced. Once they are set, they don't change.

The last video you posted a few days ago showed signs of AGC interference every time one of the figures moved.

[zerowalker]

I will do a capture so you can see the white crush active and inactive, as it isn´t like the Chroma AGC which just, Over Saturated everything without reason.

And that everything changed when the figures moved is in the movie, it´s always like that, i think they change background when the guy is moving, so it goes bright and then back.

[themaster1]

Thank you for confirming what i thought poisondeathray, 16-240 that's for full scale (aka" full swing") yuv

rec.601 (SD: TV, DVD,VHS) / rec.709 (HD) = broadcast safe (studio swing) = 16-235 / 16-212

Similarly, for 100% color bars , the RGB values when converted with Rec601 would be the same as above

How do you generate a 100% color bars pattern in the first place ?

[jagabo]

For the color bars only:

ColorBars(pixel_type="RGB32")
Crop(0,0,-0,-160)
RGBAdjust(r=1.67, g=1.67, b=1.67, rb=-32, gb=-32, bb=-32)
ConvertToYUY2()

Or you can just build them with an image editor in RGB and ConvertToYUY2().

[zerowalker]

Here is a sample, on the clip i am turning White Crush, on and off, you will see that when it´s off, it will bright upp imediatyl, and when i turn it on, it will go down in about 1 sec (tweaking settings makes this faster/slower).

As you can see from the Histogram, it doesn´t break the YUV range, so Brightness and Contrast won´t help.

PS: I got an option called "Full Luma" which as it says, allows me to use the entire Luma, it seems to help in bright places to show more details then just, Whiteness.


EDIT: Okay "Full Luma" didn´t seem to help when i did more tests, it just seemed like it i think.

[sanlyn]

ds014.avi doesn't look so great, and parts of it are wrecked. I don't refer to the over-white frames; those are obviously a disaster. That auto brightener shoves even the midtones off to the right and severely clips the brights. Ut also comprtessed the midones and brights into a narrow band that crushes much detail and makes vague mush out of many contours. Unfortunately the AGC doesn't stop the clipping or the compression; it merely darkens what the whitener compressed and clipped (i.e., crushed

Image 186 is one of the AGC-corrected frames. It is washed out, and has lost saturation and depth. You don't need a calibrated monitor to notice the results. The histograms and waveforms tell the story:



YUV histogram, vectorscope, waveform from original frame 186:


Upper left: the YUV histogram shows luma (and colors) compressed into a small area on the right-hand side; midtonesm whiuch should be near somewhere near the middle, are over-brightened and crowded into a tiny column (which is another form of clipping).

right-hand side (YUV waveform): midtones would usually lie along the green line down the middle of the column -- darks to the left, brights to the right. You can see that midtones are in a small peak near the middle portion, along with some of the darks and lower midtones that have been over-brightened. This is another bway of saying that there is very little midtone or dark detail where it's supposed to be. The image looks overexposed.

lower left (YUV vectorscope): Image data is compressed into a little blob.

RGB histograms and 'scopes, original frame 186:


left (rgb histogram): compression and clipping are shown onn the rightr of the histogram. Right (rgb waveform): the thick bands show the affected areas. Details rapidly dissipate toward the bottom of each section.

frame 186 original, rgb vectorscope: AAs with the YUV vectorscope, color and luma are compressed into a small low-saturation area.


An attempt to restore some color and luma range tp "auto-corrected" frame 186 and make it appear more natural. Rather difficult, becaue stretching midtones does restore some midlevel detail, makes the grass on the left look more like grass, makes the distant boat and standing figure more distinct, and reveals sky and cloud details that were compressed and smashed. The RGB histogram (from Photoshop) shows how luma and color were stretched into a more natural configuration. Banding and mottling from the compression are apparent in the sky and water.



Frame 663 is the original Auto-corrected frame. The rgb histogram to its right shows compressed midtones, and on its far right you see the high peaks of clipping. AFC darkened them, but the brights are still clipped (look at the sand). Midtones and shadows in the trees are too bright.


AGC level changes will affect the filters used to expand the image. The image below is frame 663 using the same correction filters used for frame 186 at the start of the clip. Here, the same luma and color adjustments make the shadow detail more obscure. You can fix them, but the adjustments will also affect frame 186, making its darker objects too bright again. With AGC changing levels unpredictably, you can't use the same filters throughout the same camera shot. And the brights, although slightly darker, are still clipped. Your own histogram in the image shows the clipping point.

[zerowalker]

Wait, did the UT codec wash the colors?

AFC darkened them, but the brights are still clipped (look at the sand). Midtones and shadows in the trees are too bright.

And that is in the whole movie, i tried fixin the sickly bright stuff, but my conclusion is that this is in the Tape, perhaps the Camcorder was very bad.
Cause it changes when the Camera focuses on the White, and if it´s focused on something else, the surrounding turns "Whiter".

Image 186 is one of the AGC-corrected frames. It is washed out, and has lost saturation and depth.

Yeah i know that the "White Crush" makes it, well, Lacking Life so to say.
But you see what happens if i turn it off, it totally clips the contrast, even though it´s within the YUV range, and i can´t control it with anything else.

Though White Crush isn´t your normal AGC i think, it´s a must have in capture devices that captures Analogue i think, though mine is badly configured on default.

[sanlyn]

Can't advise you on that, because I don't know how the original video looked without the affects of capture filters.

AFAIK I don't see how any compression codec would cause level changes.

[zerowalker]

Can't advise you on that, because I don't know how the original video looked without the affects of your other filters.

AFAIK I don't see how any compression codec would cause level changes.

I haven´t used Any filters on that clip, i just played and recorded, whit turning on and off White Crush, Nothing more.

Oh, i thought you said it did, well, it should look the same, except for some Small, color change.

[jagabo]

Obviously, you need to keep the white crush setting on for that segment. I found the definition of that function in the BT878 datasheet:

http://www.ele.uva.es/~jesman/BigSeti/ftp/Perifericos/Captura%20video/BT/bt878.pdf

As usual with datasheets, the meaning is somewhat cryptic. You have to experiment with it to figure out what it really means. But it has something to do with the automatic gain control and the sync level.

[sanlyn]

[QUOTE=zerowalker;2182739][QUOTE=sanlyn;2182738]
I haven´t used Any filters on that clip, i just played and recorded, whit turning on and off White Crush, Nothing more./QUOTE]
I seem to remember you sying that you were turning at least one autom filter on and off, and I had then idea you were actually using two of them.

[zerowalker]

Yeah, the BT878 cards got it aswell, and well, it´s as you say, there isn´t really a good explanation, like:
If you get this, set it to that etc.

There is 4 settings for White Crush:

1: White Crush(On,Off)
2: White Crush Up (1-63) = This is at 15 by defauly, and i kept it there.
3: White Crush Down(1-63) = It´s normally at 63, but for me that´s waaay to slow, it takes to long time for it to react, so i changed this to 15.
4: White Crush Majority Point (Automatic, 1/2,1/3,1/4) = I did a test, and i can´t see any real difference, though i think that 1/4 is the "best".
5: White Crush Per Frame(On,Off) = I think if it´s off it changes to Per Field, and is alot faster at reacting. ( I use it Off)

I seem to remember you sying that you were turning at least one autom filter on and off, and I had then idea you were actually using two of them.

Oh, maybe i did say it, but i didn´t use anything on that Beach clip except turning of White Crush (with the settings above) on and off.

[sanlyn]

@jagabo, looking at the PDF I'm looking at White Crush "Up" and "Down" Registers. Is that what we're referring to? (Not that what I'm looking at makes much sense, anyway). I'd request a cap of that video as-is, or none of us will know what we're talking about.

[sanlyn]

OK, so if you tried various settings we still don't know how the video looks that you're trying to correct. But when you say the control is "slow", you mean that it's slow to react to changes, not slow to feed frames to your capture (I hope).

I think I get a new impression here, you're saying more clearly now that when the video is at its brightest points, that's what it looks like without any capture processing at all?

[zerowalker]

OK, so if you tried various settings we still don't know how the video looks that you're trying to correct. But when you say the control is "slow", you mean that it's slow to react to changes, not slow to feed frames to your capture (I hope).

I think I get a new impression here, you're saying more clearly now that when the video is at its brightest points, that's what it looks like without any capture processing at all?

look at my clip, when it goes from "normal" to Super White, that´s when the White Crush is Off = Nothing changes the picture.



And when i say it´s slow, it´s when the image is white, it shouldn´t take 10 secs or more to "fix" it, but with the default settings it does.

The werid thing is though, that when it´s down at the "good level", it never really goes up, so i can pretty much have 0 on both Up and Down, the only thing that can ruin is if it flashes to white, (when i change from Composite to S-video for example), it will go Super dark, but that fixes with turning it on and off.

Though, in the end, using it with the settings i have now, seems the best, it doesn´t Over darken, and it isn´t slow.

Though i would be glad to understand what the stuff really does, i mean what should it be doing?
I am only changing what i think is good, i don´t know what to "calibrate" from;S