Airbrush/Brush Banding Effect At Low Pressure/Opacity

[Christopher Howard-3]

Hi. To try to learn GIMP development, I have been looking into old bug
#588681 (Airbrush tool produces banding with with soft brush). From the
initial problem summary:

"The Airbrush produces bandings when its used with a soft brush and low
pressure. The sharp edges in the bands is [sic] very visible when it
occurs in a otherwise soft image. The low pressure is very useful when I
want to have good control over the spray."

He uploaded an image as well of the effect, which I can easily
reproduce. When using the airbrush, the severity of the effect is
indirectly proportional to the pressure level selected. The exact same
problem can be reproduced with a normal fuzzy brush, simply by selecting
a low opacity, and repeatedly clicking on the canvas in the same location.

Initially I spent a lot of time looking at
app/core/gimpbrushgenerated.c, and especially gauss(),
gimp_brush_generated_calc_lut(), and gimp_brush_generated_calc(). This
is because I mistakenly supposed that the problem lay in the
construction of the lookup table used initially to create the brush
mask. However I discovered that this is completely unaffected by the
brush opacity or airbrush pressure.

Though having a far from sufficient understanding of how the GIMP brush
painting process works, it seems to me like this is the fundamental
problem: In making black-and-white brushes semi-transparent, GIMP
somehow "levels out" the grey-scale, so that certain ranges of darkness
become the same. On a fuzzy brush, the result is that the fine
graduation is changed to appear stepped or banded.

I would appreciate any insight into this matter. I have started to look
into how brushes are made (applied?) semi-transparent in the code, but
there are quite a few functions in the paint-funcs directory to look
through.

--
Christopher Howard
http://indicium.us
http://theologia.indicium.us



_______________________________________________
Gimp-developer mailing list
Gimp-developer@...
https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-developer

[Sven Neumann]

Hi,

On Tue, 2009-10-20 at 12:22 -0800, Christopher Howard wrote:

> Though having a far from sufficient understanding of how the GIMP brush
> painting process works, it seems to me like this is the fundamental
> problem: In making black-and-white brushes semi-transparent, GIMP
> somehow "levels out" the grey-scale, so that certain ranges of darkness
> become the same. On a fuzzy brush, the result is that the fine
> graduation is changed to appear stepped or banded.

The problem here is that the brush masks are 8bit only. With such a
limited set of values to start with, banding is basically unavoidable.


Sven


_______________________________________________
Gimp-developer mailing list
Gimp-developer@...
https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-developer

[Jay Cox]

On Oct 20, 2009, at 1:39 PM, Sven Neumann wrote:

> Hi,
>
> On Tue, 2009-10-20 at 12:22 -0800, Christopher Howard wrote:
>
>> Though having a far from sufficient understanding of how the GIMP  
>> brush
>> painting process works, it seems to me like this is the fundamental
>> problem: In making black-and-white brushes semi-transparent, GIMP
>> somehow "levels out" the grey-scale, so that certain ranges of  
>> darkness
>> become the same. On a fuzzy brush, the result is that the fine
>> graduation is changed to appear stepped or banded.
>
> The problem here is that the brush masks are 8bit only. With such a
> limited set of values to start with, banding is basically unavoidable.
>
>
> Sven

You should be able to get rid of most of the banding by performing  
error diffusion when multiplying the opacity into the brush mask.  I  
suspect that a simple 1d error diffusion algorithm will be sufficient  
and I wouldn't bother with a Floyd-Steinberg type algorithm unless you  
could come up with an example where the simpler algorithm showed  
visible artifacts.

I believe the function that introduces the banding is  
apply_mask_to_sub_region in paint-funcs.c or in the function  
apply_mask_to_alpha_channel depending on your point of view.  I have  
not worked with that code in a long time and I could be wrong about  
that.

You will need to make the starting error term  effectively random or  
you run the risk of letting the upper-left pixel in each tile get  
special treatment.

Good luck,
Jay Cox
jaycox@...
_______________________________________________
Gimp-developer mailing list
Gimp-developer@...
https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-developer