Help me color correct this?

Comments

NickHope wrote on 10/21/2017, 10:29 AM
Zero is always gonna be sitting on - zero. 🙂

Indeed. But from your initial posting I thought that Sapphire might be shifting it up from zero, which is why I commented that it actually isn't :)

Musicvid wrote on 10/21/2017, 10:34 AM

Adis-a,

Well not quite. Zero in this case is an integer scale value that is a reference, not a null value.

Same as 0 volts. Just a reference point.

Marco. wrote on 10/21/2017, 10:50 AM

You might do some testing:

Use that same gradient and add that Sapphire plug-in twice. In the first one select a gamma value of 2.0. In the second one select a gamma value of 0.5.

Does this perfectly restore the linear grey scale?

MikeLV wrote on 10/21/2017, 11:57 AM

nvidia control panel the right place to be looking? Here's some color settings, dont' see anything out of the ordinary here?

For Vimeo: here You go, Mike! 😎

Wow, lots of replies overnight, I'll try to reply to everyone. Adis-a This is what my nvidia settings look like on that tab. Should I change it to match your screenshot?

MikeLV wrote on 10/21/2017, 12:05 PM

Mike, I have arrowed the points on your histogram that ideally should align to 16 or 235. If you forget about white balance and just look at the overall RGB graph, it's pretty much spot on. Everything labelled "N" is mostly just noise and you should not attempt to retain it. This shows why the histogram can be a little dangerous; that noise can look like real data that you want to retain. However that noise barely shows up on the waveform or RGB parade, so in many ways those are better tools, and of course they also let you see where on the X axis your levels are. I must admit, I still use the histogram sometimes when I really should be looking at the waveform and/or RGB parade. I think it's just habit. But also helped by the dots on the waveform not being very visible compared to other NLEs.

Also, you should set the SeMW plugin to "Computer" if you want a preview of how the final look on the web will be (in most scenarios).

Nick, how can you tell what is noise and what isn't? I now understand what musicvid said about the endpoints on the histogram. I always thought the endpoints were where the 16 and 235 should be, but like he said, that makes the picture flat and less contrast. But how can you tell exactly where 16 and 235 should be. How do you identify this so called roll-off point?

Musicvid wrote on 10/21/2017, 12:20 PM

"Exactly" doesn't exist.

Even Photoshop Autolevels suck compared to a good pair of trained eyes.

Eyeball the points where the highs and lows start to drop off. Nick's histogram shows the roll-off points clearly.

The add a little slop on either side. That's a starting point. Then add a computer RGB temporarily to preview playback levels and tweak. Then REMOVE the computer RGB and render.

If you are looking for an exact formula, it doesn't exist. Your eyes will tell you if the played-back video has enough shadow and highlight detail, or if it is too flat for your tastes.

In this case, the white sign and dark jacket are what will guide you.

Or, shoot everything just the way you have been, and forget everything that's been said here. Your original levels are very close to being just right for Vimeo.

MikeLV wrote on 10/21/2017, 12:30 PM

Or, shoot everything just the way you have been, and forget everything that's been said here. Your original levels are just right for Vimeo.

I'm going with this - thanks for the help!

Musicvid wrote on 10/21/2017, 12:50 PM

Post a link to the video on Vimeo.

 

Musicvid wrote on 10/21/2017, 1:48 PM

I'm going with this - thanks for the help!

Hmm...

I'm reminded of a popular T. S. Elliot quote here, something to do with "exploration."

NickHope wrote on 10/21/2017, 1:52 PM
Nick, how can you tell what is noise and what isn't?

From experience of the shape of histograms, but mainly in this case by cross referencing to the waveform scope. All those random pixels below the 0 line can be safely thrown out. There is more going on above 100 where that hospital sign is, including a bit of a highlight on the right edge of the sign, but really it's mostly noise. Some or all of it might just be halos from your camera's sharpening.

(don't forget you need "Studio RGB (16 to 235)" set in your video scopes settings to see your waveform with this scale)

Adis-a wrote on 10/21/2017, 2:08 PM

nvidia control panel the right place to be looking? Here's some color settings, dont' see anything out of the ordinary here?

For Vimeo: here You go, Mike! 😎

Wow, lots of replies overnight, I'll try to reply to everyone. Adis-a This is what my nvidia settings look like on that tab. Should I change it to match your screenshot?

For Vimeo? Sure. There's a bug in Vimeo player (that's what I've been told, at least) which prevents it from displaying video as it should, making proper 16-235 material look "flat", and this is the fix. It's not hurting YouTube either...😉

Adis-a wrote on 10/21/2017, 2:47 PM

Adis-a,

Well not quite. Zero in this case is an integer scale value that is a reference, not a null value.

It gets excluded of multiplication by gamma exponent? Only that, one, value?

Same as 0 volts. Just a reference point.

 

Alessandro (Volta) is overrated, as Anders (Celsius) once was, because he too is showing negativities (-V). 😁

I predict excommunication from Metric System and replacement with an absolute metric - like Kelvin, Newton, Watt, Joule and such. 😇

Musicvid wrote on 10/21/2017, 7:42 PM

Although posted before, these linear grayscales may assist someone.

They've been optimized and tested for Vegas' scopes.

Musicvid wrote on 10/21/2017, 10:30 PM

It gets excluded of multiplication by gamma exponent? Only that, one, value?

And with that trivialized remark, you've unconsciously answered your own question.

Yes, zero is a VALUE, and along with the rest of the RANGE OF VALUES, [0, 255], exists as an inclusive endpoint, and thus a stable solution to the gamma log scale.

Refer to middle school math if that is not resonating with you.

You've got some potential here, but flippant remarks and sarcasm arent helping your credibility index. Just a heads up, adis-a.

Adis-a wrote on 10/22/2017, 9:41 AM

It gets excluded of multiplication by gamma exponent? Only that, one, value?

And with that trivialized remark, you've unconsciously answered your own question.

Yes, zero is a VALUE, and along with the rest of the RANGE OF VALUES, [0, 255], exists as an inclusive endpoint, and thus a stable solution to the gamma log scale.

Refer to middle school math if that is not resonating with you.

You've got some potential here, but flippant remarks and sarcasm arent helping your credibility index. Just a heads up, adis-a.

Well, since this is about me now, and before I go any further replying to the above, may I ask how old You are? I'm 45, and if You're older than me, then I'll just say: "You're right about everything. Apologies!", and call it a day. But if You're not...well, I really might show some of that potential You've been kind enough of crediting me with.

Musicvid wrote on 10/22/2017, 10:03 AM

You baited a moderator with a ridiculous statement; that's when you made it about you.

And now you complain when someone dishes your own words back to you?

Sorry, we don't get to have it both ways here.

But if you'd like someone to demonstrate by example why your "multiply by zero" assertion is flawed, I'd be happy to.

Oh, I was out of college and teaching at 23 when you were born, since age is important to you.

Welcome to the forums.

 

Adis-a wrote on 10/23/2017, 5:07 AM

You baited a moderator with a ridiculous statement; that's when you made it about you.

And now you complain when someone dishes your own words back to you?

Sorry, we don't get to have it both ways here.

But if you'd like someone to demonstrate by example why your "multiply by zero" assertion is flawed, I'd be happy to.

Oh, I was out of college and teaching at 23 when you were born, since age is important to you.

Welcome to the forums.

 

You're right about everything. Apologies!

Musicvid wrote on 10/23/2017, 12:18 PM

 

One popular teaching technique is to plug in arbitrary numbers to test an assumption.

We can do this easily because the output scale 0-255, although convenient, is just a bunch of labels. We can name the labels any way that we want, as long as it is consistent. Other scales used to represent the SAME DATA are Decimal, Percentages and IRE (which can be negative), and Hexadecimal for the web, which includes letters in the label as well.

So let's create a case to test the contradiction between Conventional Zero ("no apples") and Scale Zero, an assignable reference point, and thus a value in any energy equation. If we apply a simple linear shift to the scale values, leaving all the math intact, we get ([0, 255]+1), simply stated, [1, 256] as relabled.

With that pesky zero now outside the range of values, conventional math operations now work seamlessly on all all values including the endpoints, or anchors. Whatever gamma we apply theoretically stretches the rubber band tied to the anchors without affecting them. Remember, I said theoretically.

And of course, no one gets to be right about everything here. As you read this, there are lurkers about who can put my math to shame (and probably will)!

Starting your own threads here with observations, testing, and research you've done on your own is a great way to win the attention of the oldtimers and tragically curious who stop by here.

Musicvid wrote on 10/23/2017, 12:56 PM

And the no-brainer answer is that we have 256 values in the 2^8 range, not 255.

https://superuser.com/questions/578292/why-do-computers-count-from-zero

karma17 wrote on 10/24/2017, 7:19 PM

@NickHope Thanks for the reply. I totally forgot about how great Glen Chan's channel was. Also, I have that book and it is pretty good. One thing he says in there is that getting flesh tones right is half of color grading, and I think that's a fair statement. I'm really hoping Magix upgrades the color correction tools. If they can do that, then I think Vegas will be back in the game, so to speak.

Musicvid wrote on 10/25/2017, 5:29 PM

The problem with Caucasian skin tones is that an excess of either yellow or green can make the skin look jaundiced. It's sometimes needed to test both in order to discover which, or in what combination they are present.