USM with radiuses smaller than 1 pixel

[lutz]

I have the impression that unsharpmasking does not work as it is supposed to for radiuses smaller than 1 pixel.
Sometimes PL32 seems to interpret a "0.5" even as a "5" . Am I doing something wrong? Do I have to use a period or a comma?

[Martin Huber]

I have the impression that unsharpmasking does not work as it is supposed to for radiuses smaller than 1 pixel.
Sometimes PL32 seems to interpret a "0.5" even as a "5" .

I couldn't reproduce this error. Further informations would be helpful.

Am I doing something wrong? Do I have to use a period or a comma?

You can use both: a period or a comma. But in the current version the smallest possible filter radius is 1.0.
We will try to reduce the minimum filter radius to 0.6 in one of the next versions. Values smaller than 0.6 would have no effect, because with those values the filter would cover only a single pixel.

Martin

[lutz]

Hello Martin,

thanks for the reply. Maybe I got it wrong. I have tried it again at extreme settings and the result for 0.5 and 1 pixel radius seem indeed to be the same.
It would be great if you could get smaller radiuses than 1 going as announced.
Even radiuses smaller than 0.5 can produce very usable results. Please see the attached USM example with a radius of 0.4 . On the left the original, on the right the sharpened version.

[Martin Huber]

Hello Martin,

thanks for the reply. Maybe I got it wrong. I have tried it again at extreme settings and the result for 0.5 and 1 pixel radius seem indeed to be the same.
It would be great if you could get smaller radiuses than 1 going as announced.
Even radiuses smaller than 0.5 can produce very usable results. Please see the attached USM example with a radius of 0.4 . On the left the original, on the right the sharpened version.

In PL32 radii smaller than or equal to 0.5 pixel can't produce any effect, because the resulting filter matrix will cover only one pixel and interpolating a single pixel always results in the given pixel.
If a software allows radii smaller than 0.5 pixel, it probably uses a different definition of radius. Perhaps the 0.5 pixel, that do not have a effect, are excluded from the radius. This way a radius of 0.5 pixel in that software would correspond to 1.0 pixel in PL32.

Martin

[lutz]

Thanks!

all this math !!

[Peter Heckert]

Try this:

Custom filter (Userdefined filter) with this matrix:

0 -1 0
-1 5 -1
0 -1 0

You can weeken the effect, when you reduce Amount.

Peter

[lutz]

Hello Peter,

I am really not sure if I do see a sharpening effect using your settings or not? I have applied them to 3 by 3 matrix.
Do I have to set a divisor?


Yes, I guess. I just tried it (set divisor to 1). Now I do see the effect.
That seems very useful.

thanks a lot!

Try this:

Custom filter (Userdefined filter) with this matrix:

0 -1 0
-1 5 -1
0 -1 0

You can weeken the effect, when you reduce Amount.

Peter

[Peter Heckert]

Hello Lutz,

yes, I have also checked the "clip" box to avoid color artifacts.

This mask:

-1 -2 -1
-2 13 -2
-1 -2 -1

makes stronger sharpening, possibly too strong.
It can be used for problem images, but then a mask needs to be used, so that edges
only are sharpened but not the noise.

greetings,

Peter