Re: [glob2-devel] Minimap algorithm

[Kai Antweiler]

> I've been trying to recreate minimap code in its own class with sort of
> half/success. My algorithm bassically assigns a colorType to every point on
> the map, (colorType is just a smaller memory footprint used to reference the
> color itself), and then interpolates between these values. The minimap is
> only 100x100, but maps are 64x64 or 128x128 so they don't line up.

Good point.  Why don't we use a minimap that has a size which is a power of 2?
128x128 would make more sense than 100x100.  If it is too large 64x64
might be better.
Than you can cluster map squares easily to pixels.  Lets say, 4
squares become one
pixel, than you can mix the colors of those squares or determine the
most frequent or
most important color and assign it to the pixel.

> So I
> bassically I just interpolated between values, casting the 128 to 100. Some
> pixels will lie close to a point on the map, others will be half way
> inbetween two points, so the values are interpolated (in 2 dimensions) for
> every pixel.

> The algorithm works, and I did various optimizations on it, so it doesn't
> use very much CPU (less than the last one did). However, the algorithm
> doesn't produce as nice as a minimap as the old one. Specifically, it
> doesn't create the nice, curved terrain, instead, (thanks to the
> interpolated), the terrain edge seems like a sort of blurred conglomerate. I
> tried instead mapping a pixel to its closest map point (no interpolation),
> but that was far worse.
>
> As far as I can tell, the old minimap averaged between the colors of points
> of terrain, rather than interpolating. If anyone has the details on how this
> works, it would be greatly apprecciatted.

Maybe a look into a raster-to-vector converter can help:
http://autotrace.sourceforge.net/
But I doubt it.
Interpolation for tft-screens should fit our situation better.

If your code is fast, we can drop the line-by-line update of the
minimap and instead
update the whole map every 5 turns. The map is scanned anyway for
gradient construction.
--
Kai Antweiler