How does invert work in terms of luminance?

[Laurent Simon]

Dear all,

I wanted to invert a photo and apply a correction curve in order to create a digital film.

But, if I invert a test grey chart using an invert adjustment layer, photoshop actually applies some correction curve. So my L=0 becomes L=100, my L=100 becomes L=0 (so far, so good), but my L=53 becomes... L=69.

Can somebody explain me why and how I could invert so that my L= 53 becomes L=47 and so on? I did not see any option on the invert layer to choose this.

Thank you!

Laurent

[davescm]

Invert works in RGB and takes the RGB values from 255,255,255  so 180,170,0 becomes 75,85,255

Because of the gamma curve in RGB, the 128 does not translate to L 50

Dave

[Laurent Simon]

Thank you! Would you know a way to invert the luminance then?

[D Fosse]

I think you would need to work in linear gamma to get entirely predictable and mirrored numbers (gamma 1.0).

Aside from that, all of this is a bit of a murky area. Photoshop luminosity is a custom curve that is independent from the document's color space, and it doesn't match any of the standard tone response curves. It's likely based on Lab L, but it's not identical. I suspect its main purpose is to give a visually consistent result, and nevermind the numbers.

[Laurent Simon]

Thanks!

It seems GIMP offers several ways of inverting, so I will use that instead I guess. Thank you!

[davescm]

I've just tried, and it seems that if you change your Image mode to Lab Colour then Invert works the way you want - with 50 being the centre point

Dave

[D Fosse]

Interesting. I'm going to look more into this when I get time.

For a long time I thought Photoshop luminosity was Lab L. It would make perfect sense, since Lab is used as profile connection space in color management, and Lab numbers therefore are always easily available under the cover.

But then testing revealed that they wen't absolutely identical - there is a small difference in the tone curve. Since then, I've kept wondering what Luminosity really is, and never came up with a good answer.

[Laurent Simon]

Thank you i will try! This would make things easier for me!

Envoyé depuis Yahoo Mail pour Android

[elmimmo]

How can I apply this to channels? As a matter of fact, what value should I even read when sampling a channel to know what value that pixel has?

With sRGB as a working space, K=50% in the Color palette set to Grayscale returns HSB=0%,0%,63%, RGB=160,160,160 and Lab=66,0,0 in the color picker. Conversely, B=50% in the color picker becomes RGB=128,128,128 and Lab=54,0,0 in it too and 64% in the Color palette.

If instead of standard sRGB I use a customized sRGB with gamma set to 1.0 as a working space, things are not any more clear. K=50% becomes HSB=0%,0%,35%, RGB=89,89,89 and Lab=66,0,0. Conversely, B=50% becomes RGB=128,128,128, Lab=76,0,0 and K=37%.

Isn't there a mode where B=50% = RGB=127,127,127 = Lab=50,0,0 so that the inverse is exactly the same as the input?

[davescm]

If you set your greyscale working space to sGray (which you normally should). Then K 50% on the colour pallete becomes RGB 128,128,128 or HSB0,0,50%. But that is Lab 53,0,0

To get the 50% K to give HSB 0,0,50 RGB 127,127,127 and Lab 50,0,0 then you would need to set the RGB profile to Adobe RGB but with a gamma of 2.41  and the working grey also to gamma 2.41. Not sure why you want to do that though 🙂

Dave