<snip>  how can I generate more color
>>or luminosity steps when height mapping, as 256 isn't nearly enough for a
>>smooth mesh?
>
>Average and multiply by 3, giving you 0-765. Using the original RGB
>value, if the "R" component is larger than the G & B, leave your result
>alone. If the "G" component is larger than the R & B, add 1. And if the
>"B" component is largest, add 2. Your final range is 0-767 - many more
>"steps", and consistent with (I think) the way the eye focuses different
>colors in different planes.

>If you _really_ want to get inventive, you can multiply by larger
>numbers, and make the add-on value determined by the size of the
>difference between channels. You don't want to get to where _too_ much
>difference is made because of the actual colors, though, you still want
>your "color-dependent" result to be very close to what you'd get from an
>original in grey scale, just with more resolution. The only way to know
>when you've got this method "right" is when the output looks visually
>consistent with the input.
>
>Bill


Thanks for the idea Bill.

I will try it out. I bet that using the difference between channels based
on the standard RGB multipliers for correct grayscale conversion would give
a representation that is yet more consistent with the input.

If all this fails, I will sample at a higher resolution (bilinear or
bicubic), then interpolate the vertex points, and resample back down,
discarding vertex points with some yet to be written decision algorithm.

Doubling the resolution will help at least, between smoothing the normals
and voodoo, I should be able to come up with a nice final mesh.

Thanks

Robert