RE: [Algorithms] Mipmap generation

[Jon W. <hp...@mi...>]

> Is better to generate a mipmap from the previous-level mipmap or from =
the
> top-level mipmap? I suspect that the former is better for=20
> anti-aliasing and
> the latter is better for detail, so it depends on what I'm trying to
> achieve, or am I totally out of my depth here?

The top-level image has more detail, so if you have a filter=20
function which can take advantage of that (various wide windowed=20
kernels) then it's better to start where you have the detail.

If your filter function is just a box filter, there is no difference=20
between using the previous image, or using the top-level image;=20
just work out the math for each pixel and you'll see this is true.=20
(Well, if you get truncation, there is a little bit of difference,=20
and usually you do get a little bit truncation when stuffing the=20
average of four byte values into a byte...)

Cheers,

			/ h+