jffe |
I haven't tested every single module, but blur for sure, does not scale properly below a certain setting. Has anyone/everyone else had problems with range limiting controls and realizing that the scaling is completely out the window below 100ths of a unit ? Or, is this just a problem with the preview area, and if so, is that a known problem and on the fix-it-like-months-ago-already list ?
jffe Added : It occured to me that "scaling" is really only part of the issue, and might not be the best term to use when referring to my problem. I use a lot of range limiting in filters, and range limiting blur, then dropping the value below 1/10th of a blur unit, and the scaling then, is not correct to my eyes. Not correct meaning ? In the preview area, it literally seems to go back and forth, and sometimes just not change at all. You can duplicate the problem, by range-limiting a blur to 0.01, then setting it for 0.01 out in the presets area, then move it up to 0.1, I see no change usually, or try 0.05 it looks as blurry as if I had set it for 1.0 and that doesn't make any sense to me scale-wise. If it just has an absolute lower limit of 0.01, then why is it possible to set it lower in theory ? I dunno, try it out, and let me know if it's just me, or if you see the lack of change too. Filter Forger |
|||
Posted: September 16, 2007 3:50 pm | ||||
Vladimir Golovin
Administrator |
Radius is measured as a percentage of the global parameter Size. For example, if Size is set to 600 pixels, and Radius is set to 10, the actual blur radius will be 60 pixels.
However, due to technical reasons the actual blur raduis cannot be lower than one pixel. In such cases, FF produces the resulting image by interpolating between the original unblurred picture and 1-pixel blur. Maybe this is the cause of what you're seeing. |
|||
Posted: September 17, 2007 9:28 am | ||||
Crapadilla
![]() |
Seems logical. Maybe that should also go into the help file? --- Crapadilla says: "Damn you, stupid redundant feature requests!" ;) |
|||
Posted: September 17, 2007 9:43 am | ||||
Vladimir Golovin
Administrator |
Yes, good idea.
|
|||
Posted: September 17, 2007 10:05 am | ||||
jffe |
----Well, that's not the kind of scaling I meant, but it might explain what I was seeing. I never changed the Size, pixels, I just range limited the blur to the lowest possible setting (0.01) internally, then lowered it below 1.0 on the slider control, and it wasn't behaving as I would've expected to see. Setting both internal and external amounts of blur to 0.01 shows what looks like about (a non changable) 0.01 does. So I guess I'm wondering if that is the absolute lowest amount of blur you can give something (0.01), and if so, why it can be set to lower when it doesn't actually scale down any lower. Hopefully that makes my question clearer. Thanks. jffe Filter Forger |
|||
Posted: September 17, 2007 1:10 pm | ||||
Crapadilla
![]() |
If I understand the FF blur technicalities correctly, a blur radius of 0.01 would be equivalent to a 0,06 (!) pixel blur effect on a 600x600 px image, which would be way below the one-pixel threshold Vlad mentioned. On a 10000x10000 px image however, the same blur radius value (0.01) would result in an exact one-pixel blur. Consequently, you'd have to be working on an image larger than that to see blur effects larger than one pixel. And this is precisely why you won't see any noticable differences with very low blur radius values on low resolution images. On a 600x600 image, any blur radius values below 0,1666 (100/600) probably won't make any difference. Bottom line: Blur radius DOES scale down very low, but you need very large images to actually see it. Edit: One does wonder how PS does it, though. Their Gaussian Blur seems to go into sub-pixel ranges as low as 0,1... --- Crapadilla says: "Damn you, stupid redundant feature requests!" ;) |
|||
Posted: September 17, 2007 2:05 pm | ||||
jffe |
Thanks crapa, that could very well explain why I was unable to see it at 600 X 600. If FF can confirm that, then I'll consider this one answered, and poor crapa gets to write another wiki entry (he needs a raise soon ha-ha).
![]() jffe Filter Forger |
|||
Posted: September 17, 2007 2:37 pm | ||||
Vladimir Golovin
Administrator |
Their Radius is a mystery. It is perfectly possible that their radius of 0.1 translates to the actual coverage of 1.1 pixels or something like that. We spent some time trying to figure this out, but, if memory serves, we didn't find the results useful. It is simply impossible to do a blur whose kernel covers only one pixel because all samples fed into the kernel would be the same -- unless they are sampled with some kind of filtration, like bicubic, which will indirectly pull color data from the neighboring pixels into the calculation. |
|||
Posted: September 18, 2007 5:47 am | ||||
Vladimir Golovin
Administrator |
Update: Everything I wrote above on interpolation is incorrect.
I checked the actual code, and the situation is as follows: All blur-based components use actual sub-pixel blur radius. Since the kernel coverage area (radius*2) cannot be lower than 1 pixel, the Radius of 0 corresponds to the actual coverage (or diameter) of 1 pixel. |
|||
Posted: September 18, 2007 10:02 am |
Filter Forge has a thriving, vibrant, knowledgeable user community. Feel free to join us and have fun!
33,711 Registered Users
+18 new in 30 days!
153,531 Posts
+36 new in 30 days!
15,347 Topics
+72 new in year!
34 unregistered users.