Previous | Next --- Slide 46 of 70
Back to Lecture Thumbnails
OtB_BlueBerry

Here it almost seems that each triangle has one single color. So in principle, instead of using texture mapping, one can just decide to tessellate into very fine triangles and assign each a single color, right? I guess when we use textures (a) it is like a coarse-grained approximation where we add details to each triangle and hence need less triangles, (b) it saves memory for recurring patterns, (c) we lose some 3D information about the geometry. Is that a reasonable understanding?

motoole2

@OtB_BlueBerry All three points sound reasonable! Yes, one could use very very fine triangles of different colors to reproduce a texture-mapped mesh, but this would be extremely inefficient in practice. For example, suppose you want to render out a textured wall. It is far more efficient to rasterize a simple quad (two triangles) and sample a NxN texture map, than to render out NxN quads of different colors.

Btw, it is also possible to produce the illusion of fine-grain geometry with special texture maps, known as normal maps or displacement maps. In the case of a normal map, this can be done without increasing the number of triangles at all. Here's a link to learn a bit more about this technique.