Post by andrehitsoy on Dec 23, 2009 7:14:40 GMT 1
Hi Petter.
Thanks for your effort in making these applications and especially for making them public. I'm hugely impressed with your skills.
However, I was just reading your project page on CherryBrush and I just wanted to give you a heads up on some inaccuracies in your description.
I like what you have done with the color blending mode ( allthough it seems to be a visually pleasing hack and not trying to be scientifically accurate).
When you refer to linear blending in RGB space you're actually talking about linearly blending on an image that allready has an srgb curve applied to it. Hence, the image is allready in a nonlinear (monitor gamma of 2.2) colorspace so the blending isn't really happening in a linear fashion.
"The color channels use a non-linear mapping internally, which means that linear blending between two colors takes a curved path in normal RGB space."
true, but false... but true.
What your S-curve is doing is sort of emulating something closer to blending in linear color space. And with a slight film look even...
A photo like your example has a gamma curve applied to it, so to blend it linearly you could (try in PS) change the photo to 16 bit depth, apply a levels adjustment layer with a 0,4545 gamma, blend the layer with brush strokes in normal mode and apply a new adjustment layer with gamma 2.2 to bring it back to original.
So true blending would happen with an inverse 2.2 gamma curve applied. You're using an S-curve which more resembles how film behaves.. sort of.
What really promted me to write this post though, was the text on your example images. Instead of saying "Linear (traditional RGB blending)" it should say "Traditional blending in sRGB space" and on the right it should say "Nonlinear blending to emulate Linear Color blending in gamma 1"... -or something like that.
Do you agree?
Thanks for your effort in making these applications and especially for making them public. I'm hugely impressed with your skills.
However, I was just reading your project page on CherryBrush and I just wanted to give you a heads up on some inaccuracies in your description.
I like what you have done with the color blending mode ( allthough it seems to be a visually pleasing hack and not trying to be scientifically accurate).
When you refer to linear blending in RGB space you're actually talking about linearly blending on an image that allready has an srgb curve applied to it. Hence, the image is allready in a nonlinear (monitor gamma of 2.2) colorspace so the blending isn't really happening in a linear fashion.
"The color channels use a non-linear mapping internally, which means that linear blending between two colors takes a curved path in normal RGB space."
true, but false... but true.
What your S-curve is doing is sort of emulating something closer to blending in linear color space. And with a slight film look even...
A photo like your example has a gamma curve applied to it, so to blend it linearly you could (try in PS) change the photo to 16 bit depth, apply a levels adjustment layer with a 0,4545 gamma, blend the layer with brush strokes in normal mode and apply a new adjustment layer with gamma 2.2 to bring it back to original.
So true blending would happen with an inverse 2.2 gamma curve applied. You're using an S-curve which more resembles how film behaves.. sort of.
What really promted me to write this post though, was the text on your example images. Instead of saying "Linear (traditional RGB blending)" it should say "Traditional blending in sRGB space" and on the right it should say "Nonlinear blending to emulate Linear Color blending in gamma 1"... -or something like that.
Do you agree?