Page 1 of 1

Where does that blurred strip come from?

Posted: Mon Oct 03, 2005 1:03 pm
by Guest
Hi,

I was wondering if you also noticed those strange horizontal blurred static strips with examples demo.exe and collision.exe (especially on fullscreen mode). While camera is moving, we can clearly see them on screen.
I tested both programs on several different PC (windows XP) with different video cards (ATI and NVidia) and in all modes (DirectX and OGL), but they still are present, but they shouldn't...

Does it have something to do with the textures quality? Is it a bug, or is it something else ?

Jacopo

Posted: Mon Oct 03, 2005 4:14 pm
by Guest
are you talking about mip maps by any chance? the texture gets smaller as the distance gets further to try to minimize artifacts - on larger focal depths, or with poor textures and low quality mip maps you can see bands of "blurring".

try changing your gfx card settings to high quality / quality mip maps for example and you may notice a difference.

Posted: Mon Oct 03, 2005 5:04 pm
by Guest
Well, unfortunately, that's not a mip mapping problem. I already set it to get the best quality for textures. Of course, those strips are getting worst if you set mip mapping quality to a lower level.
But as I said, those bands almost are statics on screen (they represent areas in which pixels are affected) that's why I don't think it has something to do with the distance.

Just run the demo.exe example (make sure fullscreen check box is checked). In the very beginning, when camera is moving up and we discover the roof of the castle, those strange bands are clearly visible and they never disapear during the camera travelling.

Does someone know the reason?

Jacopo

Posted: Tue Oct 04, 2005 7:37 am
by Andi|xng
I can't see the stripes... I can only see the border where the mipmaps change and a black line in the skybox.
Perhaps you can show us a screenshot?

Posted: Tue Oct 04, 2005 2:22 pm
by pr3t3nd3r_guest
I think they are mip maps. try disabling the mip map from demo.
We need a better size reduction algorithm for textures if we want clearer textures.
Every scale of texture is adding an blurring factor. The more details the texture have the more blured will get ...
Or maybe are generated by stretching the texture ...
more test should be done/

Posted: Tue Oct 04, 2005 3:59 pm
by Guest
Ok, I see.
Well, IMHO Irrlicht really needs this kind of feature in priority.
It really is a great engine, but some incomplete features (such as missing texture filtering) could make us think it still has basic bugs.
Moreover, texture filtering is quite essential for rendering quality in real time.
Too bad that I'm not familiar enough with C++ :(

Let's just hope that someone will be interesting in implementing texture filtering in a near future.

Jacopo

Posted: Tue Oct 04, 2005 8:22 pm
by Guest
???

Texture filtering is already in there (bilinear by default and trilinear as an option). As for Ansiotropic there is a thread where it was implemented in OpenGL and I have some code I stuck in my custom build that enables it in D3D8.

However.. bear in mind irrlicht is only abstracting DirectX and OpenGL, it is piggybacking their graphics calls and what you see is mostly what the underlying API has to offer (except where features are missing or zbuffers are not fully implemented - ie D24X8 - )

I don't have any problems since scaling DOWN the size of meshes so that mip map problems were less obvious - too many people make huge meshes and not understand about finite resolution in the depth buffers and texture filters.

Posted: Sat Oct 08, 2005 10:00 am
by Pr3t3nd3r
say what ?

you say that if i scall down my models and i keep the same dimension from camera on the screen of the size of model i will get a bether texture quality ?? are you sure?

and yes ... zbuffers are not fully implemented ...