Thomas de Groot <tho### [at] degrootorg> wrote:
> Op 5-10-2023 om 22:37 schreef Samuel B.:
> > Thomas de Groot <tho### [at] degrootorg> wrote:
> >> A first wip...
> >>
> >> --
> >> Thomas
> >
> > It's looking promising, Thomas.
> >
> > [...]
> > Gimp now supports 16 bits per pixel, probably even more.
> >
> Ah! Thanks for reminding me, Sam. I was (un-)aware of this in my daily
> use of Gimp as 8 bits are enough for most of my purposes, but here a
> higher resolution might be asked for indeed. So, I first converted my
> base (layered) sediment .xcf image to 32 bits, and exported to 16 bits
> .png. The difference (see attached image) is not spectacular but still
> noticeable in the small details.

It looks a bit improved (or maybe just different), but I still see artifacts
characteristic of an 8-bit height map. The problem is that you can't add any
in-between elevations you didn't originally create... so, while the new image
might be a higher bit-depth, it will still contain the same stair-stepping
produced by the original 8-bits-per-channel image. (It might look slightly
different due to changes in the gamma level, or any up/down-scaling you might
have done, or something else, but the issue will remain.)

I can see two main options at this moment: 1) create a new height map starting
with an 16 bits per channel format, and painting everything again; or 2) convert
the map to 16 bit grayscale, upscale x2, blur slightly, and then downscale back
to the original x & y resolution using a good interpolation method. You will
lose some details with the second option. A third option might be to convert to
16-bit gray scale and then play with the blur and smudge tools, which would let
you put in some new details you might have wanted to add.

> I need to give an explanation (with code) about what I have been doing
> up till now. I shall do that over the week-end.
>
> In any case, this project has its serious limitations indeed as it
> relies entirely on provided image maps and their (limited) manipulation
> within POV-Ray by different (matrix) transformations while building the
> height_field(s). I shall explain later...
>
> --
> Thomas

I don't know exactly what your goals are, but can some things be done using
pigment maps and/or functions?

Sam

Post a reply to this message