the blur/smudge was better in 2.8

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

the blur/smudge was better in 2.8

Developers mailing list
I've redownloaded version 2.8 because i've found no way to emulate the
mathematical precision that the blur/smudge tools give me. The gradient
tool with perceptual RGB works almost as expected, somehow the blue value
goes up even though i'm trying to gradient between pure red and pure green.
Making a transition with blur/smudge from 255-0-0 to 0-255-0 (red to green)
i do not expect to get any colours where red+green > 255. I expect to find
254-1-0 followed by 253-2-0 followed by 252-3-0 etc. This feature is
extremely important to me because of a shader i wrote that reads colour
values mathematically. So, if i get a vote, please reintroduce the old way
to blur. Here is an image with blur and smudge examples. The top half is
what i desire, it's using blur/smudge from 2.8 resulting in colour values
where R+G+B is very close to 255. The bottom half is far too yellow, those
colour values are close to 200-200 which is far greater than the 255 i
require.
https://gyazo.com/4bf04c204f6abe42524d1dabe37b0a22
_______________________________________________
gimp-developer-list mailing list
List address:    [hidden email]
List membership: https://mail.gnome.org/mailman/listinfo/gimp-developer-list
List archives:   https://mail.gnome.org/archives/gimp-developer-list
Reply | Threaded
Open this post in threaded view
|

Re: the blur/smudge was better in 2.8

Developers mailing list
On June 20, 2019 8:12:32 AM GMT+02:00, Natalie Baldwin via gimp-developer-list <[hidden email]> wrote:

>I've redownloaded version 2.8 because i've found no way to emulate the
>mathematical precision that the blur/smudge tools give me.
>
> [...]
>
>Making a transition with blur/smudge from 255-0-0 to 0-255-0 (red to
>green)
>i do not expect to get any colours where red+green > 255. I expect to
>find
>254-1-0 followed by 253-2-0 followed by 252-3-0 etc. This feature is
>extremely important to me because of a shader i wrote that reads colour
>values mathematically. So, if i get a vote, please reintroduce the old
>way
>to blur.

In 2.10, many operations, especially ones involving color blending, transitioned from working in perceptual RGB, to linear RGB. This is the physically-correct way to blend colors, as, in perceptual RGB, the component intensities are compressed using a nonlinear gamma function. In other words, this change is intentional, and is generally the Right Thing to do.

That being said, there are still valid reasons to want to keep working in perceptual RGB, as in your case. Unfortunately, while some parts of GIMP provide this choice, others (like the smudge and blur tools) don't: it's a balancing act between allowing enough control, while not cluttering the UI with too many options, which we haven't fully nailed down yet. That's not to say we don't want to get there eventually, just that we don't have a general solution that feels quite *right* yet.

While there are workarounds for working perceptually in 2.10, even when GIMP doesn't provide this option otherwise, they're so terrible that I'd rather not mention them :)

Lastly, just to reiterate, the smudge and blur tools *do* still have the properties you're looking for, they just apply to the *linear* component values, rather than to the perceptual ones (which can be checked by converting the image precision to 32-bit linear float, and using the color picker tool in "pixel" mode). So, another option is to switch your shaders from working in perceptual RGB to linear RGB, if that's applicable.

>The gradient
>tool with perceptual RGB works almost as expected, somehow the blue
>value
>goes up even though i'm trying to gradient between pure red and pure
>green.

Oof, good catch! That's a bug. You can take care of that by disabling dithering in the gradient-tool options, but it shouldn't happen with dithering enabled either.

--
Ell
--
Sent from my Android device with K-9 Mail. Please excuse my brevity.
_______________________________________________
gimp-developer-list mailing list
List address:    [hidden email]
List membership: https://mail.gnome.org/mailman/listinfo/gimp-developer-list
List archives:   https://mail.gnome.org/archives/gimp-developer-list