Vivian,
Your description makes sense in conjunction with a mipmapping issue.
Think about how mipmapping works. The default mipmap generation algorithm
creates the 1/2 size image by averaging each 2x2 block of pixels in the
higher level down to 1 pixel in the next level. Then repeat, generating
the 1/4 size image from the 1/2 size image. You can see that at smaller
image sizes, each pixel will draw information from a wide block of pixels
out of the original so there can be a lot of bleeding across. This also
impacts alpha levels so transparency is another one that doesn't usually
mipmap as you'd expect/wish.
Note to people who hadn't noticed this before -- the requirement that
texture sizes be powers of two allows the mipmapping level creating scheme
to be well defined, and if your texture dimensions aren't an even power of
two, they are probably getting scaled to the next size down under the hood
when they are loaded by OSG.
I'm not sure what to do about the tree problem though -- it might help to
divide up the internal image into power of 2 chunks, go 4 trees across the
texture rather than 5 (just for example.) That might split down further
before producing artifacts if we aren't already doing that. (?)
I've heard graphics guru's speak of using other algorithms to generate the
mip map levels, or even manually creating them. In some cases I think you
need to consider this sort of thing even though it's a pain and blows up
the size of the distributed texture if we need to also include the
pregenerated mipmap levels.
Curt.
On Thu, Apr 25, 2013 at 5:50 PM, Vivian Meazza <vivian.mea...@lineone.net>wrote:
> Stuart
>
> > On Tue, Apr 23, 2013 at 7:22 AM, Renk Thorsten wrote:
> > >> Definitely looks like it. Could you provide some further details on
> > >> this please:
> > >> a) Where are you seeing this ?
> > >> b) which materials file (dds ? regions? )
> > >> c) Have you deleted the Textures.high file to use lower resolution
> > >> textures? The trees in the screenshot look even more blocky than
> > >> normal.
> > >
> > > After fresh pull yesterday, I can confirm the issue.
> > >
> > > a) Caribbean and French Alps so far
> > > b) using regional definitions
> > > c) no - just tried out of the box
> >
> > I spent some time last night trying to repro this without much success.
> There
> > is an issue with the very nice Caribbean texture
> > (Textures.high/Trees/tropical-alt.png) which I've got a fix for, but
> other
> than
> > that The only time I saw anything like what Vivian and you have reported
> is at
> > very very long range where I can just make out a "hat" if I look
> carefully
> > enough.
> >
> > This makes me think that this might be something to do with the way that
> > our graphics cards are generating the mipmaps. Do either of you see the
> > same issue with dds textures?
> >
> > I also went through all the tree textures to check that there weren't
> overlaps
> > on the boundaries, and in all cases except as noted above, there's at
> least a 1
> > pixel gap above the top of the UV map. I could push a speculative fix
> setting
> > the UV map for trees to a maximum height of 0.24 rather than 0.25 and see
> if
> > that makes a difference, but it feels very much like a workaround.
> >
> > Any other ideas would be most welcome, as at the moment I'm a bit
> > stumped as to how to fix this.
> >
>
> I'm using a very recent pull of fgdata with no local mods. The "hat effect"
> shows up from low angles in all material modes - regional/global/dds. It's
> most apparent at EGMH, but can also be seen at KSFO. At higher angles or if
> I zoom in it disappears from the closer trees - but, like you, I can still
> see it at longer ranges. The angle/range effect would suggest that it's a
> mipmap thing - perhaps try a bit more space around the trees in the
> texture?
>
>
> I would give you some screenshots - but that's broken here.
>
> Vivian
>
>
>
>
>
>
> ------------------------------------------------------------------------------
> Try New Relic Now & We'll Send You this Cool Shirt
> New Relic is the only SaaS-based application performance monitoring service
> that delivers powerful full stack analytics. Optimize and monitor your
> browser, app, & servers with just a few lines of code. Try New Relic
> and get this awesome Nerd Life shirt! http://p.sf.net/sfu/newrelic_d2d_apr
> _______________________________________________
> Flightgear-devel mailing list
> Flightgear-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/flightgear-devel
>
--
Curtis Olson:
http://www.atiak.com - http://aem.umn.edu/~uav/
http://www.flightgear.org - http://gallinazo.flightgear.org
------------------------------------------------------------------------------
Try New Relic Now & We'll Send You this Cool Shirt
New Relic is the only SaaS-based application performance monitoring service
that delivers powerful full stack analytics. Optimize and monitor your
browser, app, & servers with just a few lines of code. Try New Relic
and get this awesome Nerd Life shirt! http://p.sf.net/sfu/newrelic_d2d_apr
_______________________________________________
Flightgear-devel mailing list
Flightgear-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/flightgear-devel
