|
Sphinx.
Filter Optimizer
|
Filter:
Organic/Organics
Bug:
black edge artifacts appears during temporary preview rendering.
Reproduction:
Open FF3 Beta 1, click category "Organic", click filter "Organics". Pay attention to temporary preview rendering in right and bottom edge areas. Notice how the artifacts are reduced for each pass, and disappears completely in the final pass.
 Njyldgarkn sample cache! |
| Posted: May 2, 2011 5:10 am |
Details
E-Mail
|
|
Sphinx.
Filter Optimizer
|
Here is another view of the bug with the kingfisher preview image. The filter was set to seamless tiling. Again the edges dissappear in the final pass.
 Njyldgarkn sample cache! |
| Posted: May 2, 2011 5:42 am |
Details
E-Mail
|
GMM
Moderator
Filter Forge, Inc
Posts: 1759
|
This is by design. Progressive preview is meant to provide a rough version of the image on the first passes. |
| Posted: May 3, 2011 4:37 am |
Details
E-Mail
|
|
Sphinx.
Filter Optimizer
|
Hmm ok, then I think there is a bug in the design  I can't imagine proper edge handling will kill performance... at least wrap around or repeat edge values instead of interpolating towards black...Njyldgarkn sample cache! |
| Posted: May 3, 2011 5:30 am |
Details
E-Mail
|
|
Dmitry Sapelnikov
Filter Forge, Inc. AKA Egret
|
The situation with the kingfisher example is clear - 'border' is created by blending with pixels outside the canvas.
This pixels are transparent if the kingfisher image is loaded into color control and tiling mode of the control is set to "None". You can check it out by plugging the color control into the Scale component.
But the organics example looks more complicated. Frankly speaking I haven't fully understood the problem yet. |
| Posted: May 4, 2011 8:04 am |
Details
E-Mail
|
Vladimir Golovin
Administrator
Filter Forge, Inc.
|
The kingfisher example is easy to fix by setting Tiling to Repeat or Mirror in the image placement window. The Organics example is a mystery. Most likely, this is related to Ambient Occlusion. We'll investigate this. |
| Posted: May 4, 2011 8:46 am |
Details
E-Mail
|
|
Sphinx.
Filter Optimizer
|
Ok, yes, the two cases are behaving differently. I thought perhaps they were related due to the visual resemblance.
About organics - if you enable seamless tiling it goes away. And yes it seems related to AO (since the artifacts are not there when AO is off)Njyldgarkn sample cache! |
| Posted: May 4, 2011 8:55 am |
Details
E-Mail
|
|
Sphinx.
Filter Optimizer
|
Second thoughts about the kingfisher example - I think it is a different bug.
If the region outside is transparent, then how can it contribute to the visual appearance?  The interpolation scheme seem to make the common mistake of not taking alpha into consideration (i.e. the contribution of samples are not scaled by alpha)Njyldgarkn sample cache! |
| Posted: May 4, 2011 9:03 am |
Details
E-Mail
|
|
Dmitry Sapelnikov
Filter Forge, Inc. AKA Egret
|
| Quote |
|---|
Sphinx. wrote:
If the region outside is transparent, then how can it contribute to the visual appearance? |
It's quite simple. Progressive preview requires squares of samples, which are interpolated to approximate pixels not computed on the current pass.
So we need some extra pixels outside the canvas for correct interpolation of the squares intersecting the canvas. I hope the attached sketch will clarify the concept.
 |
| Posted: May 4, 2011 9:26 am |
Details
E-Mail
|
|
Sphinx.
Filter Optimizer
|
Ah, I was confused by my own set up. The edge handling with the kingfisher IS correct. I didn't realize the main canvas color was the one I saw "bleeding through". That is correct. Sorry guys  Njyldgarkn sample cache! |
| Posted: May 4, 2011 10:38 am |
Details
E-Mail
|
