32bit textures look like they're being displayed in 16bit colors

Hi all!

Sorry for asking so many questions in quick succession - I’ve been storing them up before I found out about this board

My problem is that I am loading 24 bit textures (straight from a .raw file into glTexImage2d, although I add an alpha channel), but when they are displayed there is obvious banding in what should be smooth gradients, etc, with the general effect looking like the textures are being displayed in 12 (!) bit or similar colour resolution

This happens on the 3 graphics cards I’ve tested, in both 16 and 32 bit colour modes…
I load my textures in GL_RGBA format, with the buffer in unsigned bytes…

Any ideas?

Thanks in advance,

Nick

I once had the same problem, and the reason is probably the same for you as it was for me. I guess you load your image like this:

glTexImage2D(GL_TEXTURE_2D, 0, 4, SizeX, SizeY, 0, GL_RGBA, GL_UNSIGNED_BYTE, &Data);

Do it like this instead:

glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA8, SizeX, SizeY, 0, GL_RGBA, GL_UNSIGNED_BYTE, &Data);

This third parameter sets the internal format and is not the number of components as said in both the MSVC docs and the OpenGL SuperBible.

[This message has been edited by Humus (edited 05-29-2000).]

Thanks - they work a lot better now!

[This message has been edited by nickL (edited 05-30-2000).]

I have read about the third parameter in MSVC documentation, but never thought more about it, until now. Since you CAN pass a 4 as parameter and it works (more or less), OpenGL must assume it means four elements (red, green, blue and alpha). Is this correct?

I guess that if you pass a incorrect number you just get some sort of default internal format which depends on your drivers. At my G400 i got 16bit textures with a 4, but i think most nVidia cards gives you 32bit textures with a 4 (maybe to handle the common mistakes done by the misleading info in the MSVC docs and the SuperBible)

[This message has been edited by Humus (edited 05-30-2000).]

4 is accepted because of compatibility issues with OpenGL 1.0. It means GL_RGBA.
1 - 3 respecitvely mean GL_LUMINANCE, GL_LUMINANCE_ALPHA, and GL_RGB.