How does mipmap overblur? Ben mentioned that in a previous slide, there were shapes with sides of different lengths represented by squares instead, and that it was "averaging too much of the underlying texture." But, that didn't really make any sense to me lol. Could someone explain more simply what he means?
If you think about a texture tilted aggressively away from the camera, the tilted plane has the shape of a trapezoid from screen space. But the algorithm matches it with a texture square. If you think about a trapezoid with height Z versus the square of height Z, the trapezoid is smaller in area than the square. Smaller areas (texture footprints) need more detail (because they are closer). But in this case, the smaller area is being fitted with a texture space that is designed for something bigger than it. So you lose detail, hence blur. Hope that helps, -Kevin