polar bleed-over from one pole into the other

Report bugs, bug fixes and workarounds here.
Topic author
Malcolm

polar bleed-over from one pole into the other

Post #1by Malcolm » 27.07.2002, 22:35

Whilst trying to make smooth natural looking polar regions
(without the radial striations)
I find that it can only be done with the same colour in both polar regions.

If I use say a pale white or off-white in the north pole and a darkish colour in the south pole then I get a dark spot from the south showing up in a small spot in the exact center of the north polar cap.

Topic author
Malcolm

polar bleed-over from one pole into the other

Post #2by Malcolm » 31.07.2002, 10:09

Hi Malcom,

this is IMHO unfortunately unavoaidable.

When you paint a map for Celestia (and almost any other space sim) you have to create a socalled simple cylindric or 'Mercator' projection.

That is a rectangular map whose left and right ends are put together and then it's equator is placed on the sphere's equator.

If you visualize the process, you'll see that this is fairly exact for the regions around the equator. The further away you get, the worse it gets.

At the poles you have the whole with of the map (at it's top) mapped to one point of the sphere, the pole. So if you have the slightest change in color above the polar circle on your map, you'll recoginze those radial artifacts.

The second prob is the smoothing method modern 3D cards use to remove pixel artifacts from textures. They smooth the whole original texture by generating arithmetic middles for each and every pixel. So you get a nice texture even if res is pretty low.

The bad thing is that you can't do this for the edges of the texture, since there's no 2cnd pixel left. This would disrupt the whole texture, so the cards, drivers or OpenGL is just taking the next pixel in line and that's the one on the oposite end of the texture.

For all pixels on the vertical edges this works very well, you don't see a vertical see on many textures and if you do it's damaged.

However, this only works for the horizontal edges (aka the poles) if booth are of the same color.

Just stick with uniform pole areas in close approximation to the poles. This is easy for planets like Earth, but it's almost impossible to get a decent cratered moon to look perfect around there with simple cylidric projection.

Take care, 8)
Axel




Malcolm wrote:Whilst trying to make smooth natural looking polar regions
(without the radial striations)
I find that it can only be done with the same colour in both polar regions.

If I use say a pale white or off-white in the north pole and a darkish colour in the south pole then I get a dark spot from the south showing up in a small spot in the exact center of the north polar cap.

Axel

polar bleed-over from one pole into the other

Post #3by Axel » 31.07.2002, 10:09

Hi Malcom,

this is IMHO unfortunately unavoaidable.

When you paint a map for Celestia (and almost any other space sim) you have to create a socalled simple cylindric or 'Mercator' projection.

That is a rectangular map whose left and right ends are put together and then it's equator is placed on the sphere's equator.

If you visualize the process, you'll see that this is fairly exact for the regions around the equator. The further away you get, the worse it gets.

At the poles you have the whole with of the map (at it's top) mapped to one point of the sphere, the pole. So if you have the slightest change in color above the polar circle on your map, you'll recoginze those radial artifacts.

The second prob is the smoothing method modern 3D cards use to remove pixel artifacts from textures. They smooth the whole original texture by generating arithmetic middles for each and every pixel. So you get a nice texture even if res is pretty low.

The bad thing is that you can't do this for the edges of the texture, since there's no 2cnd pixel left. This would disrupt the whole texture, so the cards, drivers or OpenGL is just taking the next pixel in line and that's the one on the oposite end of the texture.

For all pixels on the vertical edges this works very well, you don't see a vertical see on many textures and if you do it's damaged.

However, this only works for the horizontal edges (aka the poles) if booth are of the same color.

Just stick with uniform pole areas in close approximation to the poles. This is easy for planets like Earth, but it's almost impossible to get a decent cratered moon to look perfect around there with simple cylidric projection.

Take care, 8)
Axel




Malcolm wrote:Whilst trying to make smooth natural looking polar regions
(without the radial striations)
I find that it can only be done with the same colour in both polar regions.

If I use say a pale white or off-white in the north pole and a darkish colour in the south pole then I get a dark spot from the south showing up in a small spot in the exact center of the north polar cap.

Topic author
Malcolm

Post #4by Malcolm » 31.07.2002, 18:25

Hello Axel, very many thanks for your excellent explanation and thank you for the time you put into it.
Yes, I see it all now ! Thank you !
I was previously thinking that the averaging would be going on only in the northern rectangle(Mercator) region for the north pole
and similarly, averaging from the south rectangle region to go into the south pole only
Thus _my_mistake_ was not to think of the whole map being wraped round top - to- bottom, as well as east-to-west.

> almost impossible to get a decent cratered moon to look perfect around there <
Yes. I had played with a construct in which this agrevating spot became the central peak in a classic crater (it made quite a nice mountain with a bit of radial pixelation) and where the crater wall was formed by a few very long (east to west mercator) rows of blured pixels, but always it looked too artificial :( I could hear other Celestials flying round my Triton saying - " ha ha, I know who's been here and it was not Kermit " !

I had found a Voyager (partial)Triton medium/highres - on a nasa site - and thought it would be a good exersize.
( http://maps.jpl.nasa.gov/neptune.html )
I perhaps should not have typed about the radial striations, because although that was _why_ I had got involved in smoothing the poles, I was in fact getting on quite well with them by bluring and stretching the shading side to side over several pixels.
Bluring wider horizontally (over more pixels) the closer that strip of pixels was to the pole (top edge) With the polar few strips being, finally, all the same colour.
My problem arose when I tried to do the same thing at the other pole. If I used a copy of what I had done in the 'north' then all was OK.
But because the 'south' regions near their pole were much darker than those in the 'north' the copy pole was much too light and did not match well.
So, I just stretched and blured the darker regions into their own pole (getting a bit lighter as I went ) but then found that a ghost of this darker stuff arrived in the _north_ the closer I was getting to the _south_ pole.
Wierd, wormhole perhaps, :) I thought.

Best wishes,
Malcolm.


Return to “Bugs”

Who is online