[EMAIL PROTECTED] (2002-01-15 at 1512.52 +0100):
> > I've been following some instructions oriented toward photoshop that
> > talk of unsharpen mask settings including a radius of 0.3 or 0.8.
> > I notice that Gimp's implementation has a minimum radius of one.
> I'm quite sure that it's just the UI which is limiting the value.
In my travels around the code (!= I understand it) I saw some cases in
which the functions accept float values, but some widgets do not like
it. The one with unit selection in guassian blurs is the example I
remember, and Sven told me it could be fixed for 1.3.
One note about Photoshop: sometimes the values used in it are not the
same than in GIMP, if you want the same results. IIRC the histograms
are different.
> Would you try if it works correct with smaller values and report
> back so we can remove the silly limitation (if it's silly :) ?
In the case of gaussian blurs, half pixels show differences. I did the
test, I just had to change the units to one that allowed to input
float values (pt vs px), and I got blur (zoom and color picker
recommended if the case is not a good one and your eyes are not good
either).
<nonsense=maybe>
It does not work with 0.9 and 1.0, but there is difference with 1.5
and 2.0. Dunno why it takes values less than 1 as 0. Did people
checked it or just think it will be wrong? If you think in analogic
mode or using pixel subsampling, it makes sense, IMO... that or I
remember wrongly how the function is, the radius is the distance from
the center of the "bell" (integral is 99/2% of total) or from one side
to the other (so you have 99% inside)?
-+----|----+----|----+----|----+- | means limit, + center of pixel
|---------| Radius?
_
\
' - _ _
|---------| Radius? (diameter is more correct, no?)
_ (at least from an UI POV)
/ \
__-' '-__
As you can see, if the top case is the right one, it would be made
sense to support up to a bit over .5 (check "> 0.5") cos you still get
a bit of info from the neighbor pixels. I understand that pixels have
a radius of .5 pix to the side (sqrt of .5 to the corner, but the
function does vertical and horizontal passes only).
What it wrong? The terms used (in UI, code or by me)? Test in code? My
brain? Explanations highly appreciated (and why blur is done in two
perpendicular passes instead of a matrix, too :] ).
</nonsense>
GSR
_______________________________________________
Gimp-developer mailing list
[EMAIL PROTECTED]
http://lists.xcf.berkeley.edu/mailman/listinfo/gimp-developer
