Standard behaviour of image->texture functions like CreateGLTextureFrom(Multi)Image is: If graphic card supports non power of 2 textures and image is nonpow2 then texture is created directly from image.

PasteNonPow2ImagesIntoPow2: Boolean = False;

ImagingOpenGL (at line 174)

If graphic card does not support them input image is rescaled (bilinear) to power of 2 size. If you set PasteNonPow2ImagesIntoPow2 to True then instead of rescaling, a new pow2 texture is created and nonpow2 input image is pasted into it keeping its original size. This could be useful for some 2D stuff (and its faster than rescaling of course). Note that this is applied to all rescaling smaller->bigger operations that might ocurr during image->texture process (usually only pow2/nonpow2 stuff and when you set custom Width & Height in CreateGLTextureFrom(Multi)Image).