Do you mean MIPMAPPING?

First I would use several models with different quantity of polygons for the different distances (view <-> object)
Most performance drop comes from the poly count.
But if you reduce the amount of polygons you have to hide the loss of details with a different texture applied.

If you are very close to a statue with a crack in its leg, you need to see the crack as geometry.
If you go away a bit, you don't need to model the crack - it is sufficient to alter the texture
(with a dark painted shadow at the crack position ) and use a more simple geometry.

MIP Maps are reduced in size for each LOD step and avoid the need to scale the texture
at runtime from 1024x1024 down to an object which is only 75 pixel heigh at the time of rendering.

So, what do you need?

-- slacer