Gamma Node for Cycles

Gamma Node.

I tried to use powr instead of if (.. > 0) powf, but didn't work. No harm I guess.
For after bcon5.

This node is not so useful for 'materials' but I believe it comes handy for textures input. Although it's simple to do it with the math node.

*** committed on rev. 42671

[dfelinto, 2011-12-16 18:47:02]

No real change in the code, simply update  to help if someone wants to test.
trunk ready I think

[brechtvl, 2011-12-16 20:01:19]

LGTM.

[zanqdo_gmail.com, 2011-12-17 09:31:15]

I was testing the gamma node and noticed a discrepancy in the results from
bringing an image as linear vs linearizing with the node. However bringing
the image as sRGB vs bringing as linear and setting 2.2 gamma with the node
looks quite the same to me

http://www.pasteall.org/blend/10346

I used the simple formula 1/2.2 to linearize it. is that my mistake or
could it be precision issues, or a bug?

Daniel Salazar
3Developer.com


On Fri, Dec 16, 2011 at 2:01 PM, <brechtvanlommel@gmail.com> wrote:

> LGTM.
>
> http://codereview.appspot.com/5488046/
> _______________________________________________
> Bf-codereview mailing list
> Bf-codereview@blender.org
> http://lists.blender.org/mailman/listinfo/bf-codereview
>

[dfelinto, 2011-12-17 21:03:09]

The gamma node is doing a simple pow operation:
color_out = color_in ** gamma (as in 8 = 2 ** 3)

And looking at your sample file it seems consistent.
Linear ** 2.2 = sRGB
sRGB ** (1/2.2) = Linear

You don't use gamma 2.2 to linearize a sRGB image to Linear. You use 1/2.2 for
that.
Gamma 2.2 is the gamma already present in the sRGB colour space.

[zanqdo_gmail.com, 2011-12-18 08:02:02]

Yes yes, however they are NOT consistent, there are slight differences, so
I'm wondering what would be the explanation for the gamma process to not be
perfectly reversible?

cheers and thanks!

Daniel Salazar
3Developer.com


On Sat, Dec 17, 2011 at 3:03 PM, <dfelinto@gmail.com> wrote:

> The gamma node is doing a simple pow operation:
> color_out = color_in ** gamma (as in 8 = 2 ** 3)
>
> And looking at your sample file it seems consistent.
> Linear ** 2.2 = sRGB
> sRGB ** (1/2.2) = Linear
>
> You don't use gamma 2.2 to linearize a sRGB image to Linear. You use
> 1/2.2 for that.
> Gamma 2.2 is the gamma already present in the sRGB colour space.
>
>
http://codereview.appspot.com/**5488046/<http://codereview.appspot.com/5488046/>
>

[brechtvl, 2011-12-18 12:17:50]

Gamma 2.2 and 1.0/2.2 should cancel each other out. However, Blender sRGB <=>
linear conversion is not a simple gamma conversion, the formula for that is
quite different:

float srgb_to_linearrgb(float c)
{
    if (c < 0.04045f)
        return (c < 0.0f)? 0.0f: c * (1.0f/12.92f);
    else
        return powf((c + 0.055f)*(1.0f/1.055f), 2.4f);
}

float linearrgb_to_srgb(float c)
{
    if (c < 0.0031308f)
        return (c < 0.0f)? 0.0f: c * 12.92f;
    else
        return  1.055f * powf(c, 1.0f/2.4f) - 0.055f;
}

[dfelinto, 2011-12-18 20:00:59]

Brecht, is this to approximate the result from a real gamma curve?

On 2011/12/18 12:17:50, brechtvl wrote:
> Gamma 2.2 and 1.0/2.2 should cancel each other out. However, Blender sRGB <=>
> linear conversion is not a simple gamma conversion, the formula for that is
> quite different:
> 
> float srgb_to_linearrgb(float c)
> {
>     if (c < 0.04045f)
>         return (c < 0.0f)? 0.0f: c * (1.0f/12.92f);
>     else
>         return powf((c + 0.055f)*(1.0f/1.055f), 2.4f);
> }
> 
> float linearrgb_to_srgb(float c)
> {
>     if (c < 0.0031308f)
>         return (c < 0.0f)? 0.0f: c * 12.92f;
>     else
>         return  1.055f * powf(c, 1.0f/2.4f) - 0.055f;
> }

[brechtvl, 2011-12-18 20:24:34]

The purpose of this scene linear color space in Blender is to roughly match
physical units of light, while sRGB is supposed to be a color space to display
that on the screen. That's only an approximation in various ways, but it's more
like a gamma curve is an even simpler approximation to this type of transform.

On 2011/12/18 20:00:59, dfelinto wrote:
> Brecht, is this to approximate the result from a real gamma curve?

[zanqdo_gmail.com, 2011-12-19 12:20:27]

Ah! so they are indeed different algorithms. Humf I guess it's not a
problem atm

cheers

Daniel Salazar
3Developer.com


On Sun, Dec 18, 2011 at 2:24 PM, <brechtvanlommel@gmail.com> wrote:

> The purpose of this scene linear color space in Blender is to roughly
> match physical units of light, while sRGB is supposed to be a color
> space to display that on the screen. That's only an approximation in
> various ways, but it's more like a gamma curve is an even simpler
> approximation to this type of transform.
>
>
> On 2011/12/18 20:00:59, dfelinto wrote:
>
>> Brecht, is this to approximate the result from a real gamma curve?
>>
>
>
http://codereview.appspot.com/**5488046/<http://codereview.appspot.com/5488046/>
>