MINI TUTORIAL: how to eliminate the PINCH effect at poles

[ElChristou]

Hello,

Those steps are a basis to give some ideas to beginners?€¦ as it, a 100% success is not guaranteed, some textures needs much more attention?€¦

In this example we are going to talk about north pole (top of image), the same procedures have to be done to south pole after a 180?° flip.



Step1: My texture as background:



The problem with it is that I don?€™t know how the pole will be once on a sphere?€¦


Step 2: Duplication of the background as a new layer (layer1);


Step 3: On layer 1: Filter distort > polar coordinates (rectangular to polar);

Here the result:



We can see the pinch effect?€¦




Step 4: We need to modify this zone as we want it to be:

Here the layer 1 after modification:




Step 5: Back to normal mapping: Filter distort > polar coordinates (polar to rectangular);

The result:



Well this map is too blurry, my background was really much sharpen?€¦ so:


Step 6: Duplicate the layer 1 to new layer (layer2);


Step 7: Deleting part of layer 1 as shown, then sharpen it and finally set it with 50% transparency;


...white is transparent...


Step 8: Deleting part of layer 2 as shown then sharpen it;


...white is transparent...


Step 9: Flatten Background + layer 1 + layer 2:




Ok here is a not so bad texture?€¦ unfortunately this method won?€™t give a high detailed pole, in some case a blurry pinch effect still present so you will have to make some others modifications?€¦ In this example, the layer 1 after modification of the pinch effect (step4) has been duplicate 1 time to create a gradient effect with the background?€¦ some textures need more layers and a finest work to have a good result.

Also, instead of sharpening in step 7 and 8, you can do one or two sharpen in step 4 after modifying the pole zone.

To get a higher detailed pole, you will need to work on large files?€¦
Personally, I work with 5K textures to reduce them to 4k then to 2k and 1k.

Don?€™t forget patience and perseverance are the keys for nice textures.



NOTE: There is a last problem, a bug, which not comes from the texture; it?€™s the inversion of the extreme point of the poles?€¦ Apparently for the moment we have to wait for the correction for this in a future version?€¦


Hope this will be utile for newcomers.

Bye.

[Brendan]

Thanks for writing this. It looks useful, so I'll try it in GIMP for my next system.

[t00fri]

ElChristou,

yes, thanks for your effort. That's essentially also how I usually fight the pinch effect.

Actually, after mapping /half/ a cylindrical planet map from the equator to one pole to polar coordinates, one may superimpose actual photos of the pole regions if they exist. After mapping them back to cylindrical shape, it is guaranteed that the imaging near the pole remains sharp, contains all the additional photographic information and is "pinch-free"...

Bye Fridger

[Rassilon]

NOTE: There is a last problem, a bug, which not comes from the texture; it?€™s the inversion of the extreme point of the poles?€¦ Apparently for the moment we have to wait for the correction for this in a future version?€¦

This inversion is actually an artifact that I see in OpenGL in general....It is the normal vectors being computed several times at the same point since each slice of the sphere meets at the poles and each vector is given a normal this produces the affect you see....Similar to the infamous seam in 3ds files...3ds files use two smoothing groups that overlap at the meridian...just as the slices do at the poles...One would need to weed out the extra normals computed and calculate it only one time....

Also as for polar pinch make sure you sharpen the texture before doing a polar distortion and not after...Adobe leaves nasty pixel artifacts when performing a polar pinch which are more pronounced when sharpening afterwards...

There is a great filter for removing artifacts however...It is either Image Doctor by the makers of Eye Candy Alien Skin and Xenoflex 2 or smart blur...

[essee]

So, is it possible to use the "actual photos of the pole regions" Fridger spoke of (or any such photographed regions) as part of the layering of images, steps 7 thru 9, from ElChristou? Im trying to create sort of an "illusion" of greater detail...

[ElChristou]

This inversion is actually an artifact that I see in OpenGL in general....It is the normal vectors being computed several times at the same point since each slice of the sphere meets at the poles and each vector is given a normal this produces the affect you see....Similar to the infamous seam in 3ds files...3ds files use two smoothing groups that overlap at the meridian...just as the slices do at the poles...One would need to weed out the extra normals computed and calculate it only one time....

Tx for the info... I hope something can be done to solve this.

Also as for polar pinch make sure you sharpen the texture before doing a polar distortion and not after...Adobe leaves nasty pixel artifacts when performing a polar pinch which are more pronounced when sharpening afterwards...

I haven't mentionned this because when you are working on a 5-6k file to have a final 2k, those problems generaly desepears with the reduction...

It's always better to work on a much bigger file in PS, then reduce it to the final format; in the process of reduction you can control some details level, personnaly I often do the reduction in 2 steps, to do a sharpen, or some little works of retouching, etc...

Bye

[ElChristou]

So, is it possible to use the "actual photos of the pole regions" Fridger spoke of (or any such photographed regions) as part of the layering of images, steps 7 thru 9, from ElChristou? Im trying to create sort of an "illusion" of greater detail...

Can you explain a little more about your "illusion"??

If you have a photo of a pole region you have to use it in step 4, before reverting the polar filter.

Bye

[essee]

Thanks for the quick response. It should have been obvious to me that an image would have to be included in step_4, oops... The "illusion" would be somewhat like Selden's A Map Plane for Celestia
http://www.lns.cornell.edu/~seb/celestia/map-plane.html
which is still pretty cool.

SteveC.

[scalbers]

This is hinted at earlier, but could this mostly be a Celestia or OpenGL issue somehow? I'm under the impression that using GIMP or other software to remap a cylindrical texture into a polar view would give much less pinching artifact compared with Celestia.

[Mr Noob]

One thing I would love to know is whether or not this "pinch" can be removed in a later Celestia version.. or will it always be there?

[chris]

The 'polar pinch' effect can be reduced significantly by enabling anisotropic texture filtering. Future versions of Celestia will turn on anisotropic filtering when it's supported by the systems's graphics card. In the mean time, you can usually turn on anisotropic filtering via the control panel (at least one Windows.) The exact steps required to enable it will vary based on the graphics card.

Anisotropic filtering does come with some performance cost, but so far in tests, I haven't seen it slow down Celestia much at all. I think Fridger had similar results: better texturing near the poles of planets, and not much impact on performance.

See this post for a comparison of normal trilinear filtering vs. 8x anisotropic filtering:
viewtopic.php?f=5&t=13776&p=115677&hilit=polar+pinch+anisotropic#p115679

--Chris

[t00fri]

The 'polar pinch' effect can be reduced significantly by enabling anisotropic texture filtering. Future versions of Celestia will turn on anisotropic filtering when it's supported by the systems's graphics card. In the mean time, you can usually turn on anisotropic filtering via the control panel (at least one Windows.) The exact steps required to enable it will vary based on the graphics card.

Anisotropic filtering does come with some performance cost, but so far in tests, I haven't seen it slow down Celestia much at all. I think Fridger had similar results: better texturing near the poles of planets, and not much impact on performance.

See this post for a comparison of normal trilinear filtering vs. 8x anisotropic filtering:
viewtopic.php?f=5&t=13776&p=115677&hilit=polar+pinch+anisotropic#p115679

--Chris

I completely agree. Even on my archaic FX 5900 Ultra card, there is only a mild reduction of performance with 8x anisotropic filtering. Because it is known to be a good and easy remedy against polar pinching, I have anisotropic filtering switched on continuously.

Let me add that under Windows, the NVIDIA GUI control interface allows to set OGL options selectively by program name! Perhaps not everyone is aware of this most convenient feature.

[Reiko]

Oh wow I didn't know anisotropic filtering worked with celestia.