Good Mona,

you are simply charming! We should go out these days again and have a glass of red wine in a cosy little bar, ahem...

I am bad...right?

F.

t00fri wrote:

I am bad...right?

I guess so. I do hope your wife is not lurking around here from time to time!
That said, let's stay on topic!

cartrite wrote:

...I was looking at an ico sphere. There you have 5 triangles and no pole pinch. Like a bee's honey cone. But can that be mapped or run in Celestia or graphics hardware?

I guess yes in theory but that imply to not use anymore cylindrical projection, both for mapping or mesh generation (like Phobos in this thread). As far as I can tell, that would be a change way too huge in the overall habits within the Celestia project. Not only the code itself should be rewritten but all tools and all maps redone...
As it seems we are close to see these polar pinch disappearing once for all* still using cylindrical projections, I'm not sure digging for new solutions is worthwhile...

*(via anisotropic filtering for textures and perhaps now some solution for geometry too)

I am working on some anti-pinch ideas of mine...

Fridger

I tried it last night. No code change is needed. Just a lot of work. The results were not good. But I didn't spend too much time mapping the UV's as you can see here.

Here is the result. There isn't any pole pinching but there are other artifacts. It may have to do with the mapping because I didn't see any artifacts near the equator. But I didn't look to hard at the equator either. It may require a special texture. I'm not sure. Here is an image of the south pole of Mars with an icosphere.


I think the best answer for "Polar Models" where a polar model would be a model that you want a lot of detail in the polar region would be a cube map. There was some talk about cube mapping a while back. Maybe there will be more talk about it now.
cartrite

cartrite wrote:

I tried it last night. No code change is needed...

I don't follow you well; by code change, I mean the spheres used in the Celestia code are not adapted (as far as I know) to something else than cylindrical projections...
Now if you want to deal with models, then yes whatever projection will work!

ElChristou wrote:

cartrite wrote: I tried it last night. No code change is needed... ElChristou wrote: I don't follow you well; by code change, I mean the spheres used in the Celestia code are not adapted (as far as I know) to something else than cylindrical projections...

I think I misspoke. Yes I agree that a cylindrical projection is not well suited for an icosphere. I thought that Celestia would not render a texture on a icosphere. It can. But it doesn't do a good job on cylindrical textures in the polar regions. It may be possible to map a polar stereographic projection to the poles and cylindrical maps around the equator. But it would have to be done in a modeling program.

ElChristou wrote:

Now if you want to deal with models, then yes whatever projection will work!

I thought that this conversation was all about models. Pole pinching can be handled on Celestia's internal spheres with anisotropic filtering. Models actually create pole pinching that anisotropic filtering can not handle.
cartrite

cartrite wrote:

ElChristou wrote: Now if you want to deal with models, then yes whatever projection will work! I thought that this conversation was all about models. Pole pinching can be handled on Celestia's internal spheres with anisotropic filtering. Models actually create pole pinching that anisotropic filtering can not handle.

My bad! Your previous mention about Mars pole region makes me think you wanted a better pole geometry system for any bodies including planets! Sorry didn't realize you were referring to a model!

Guys I just found a nasty seam near Flimnap; someone see such thing?

ElChristou wrote:

Guys I just found a nasty seam near Flimnap; someone see such thing?

Yes, I am still investigating. It's NOT on my new 4k Phobos map, but rather related to the shape model. It extends from the North pole quite a bit, passing closely by the crater Flimnap. Depending on ilumination, it can be well or barely visible. Did you play with the cmodfix parameters or did you use directly the ascii-CMOD model?

F.

Some additional info about this seam along the positive x-axis:

with the old phobos.cmod from the distribution it's much weaker, BUT it's still there.
So this might be some CMOD model delicacy, which perhaps ChrisL has an idea about.

Fridger

t00fri wrote:

Yes, I am still investigating. It's NOT on my new 4k Phobos map, but rather related to the shape model. It extends from the North pole quite a bit, passing closely by the crater Flimnap. Depending on ilumination, it can be well or barely visible. Did you play with the cmodfix parameters or did you use directly the ascii-CMOD model?

Yes definitively a geometry problem. And yes my model went through cmodfix...

t00fri wrote:

Yes, I am still investigating. It's NOT on my new 4k Phobos map, but rather related to the shape model. It extends from the North pole quite a bit, passing closely by the crater Flimnap. Depending on ilumination, it can be well or barely visible. Did you play with the cmodfix parameters or did you use directly the ascii-CMOD model?

It's a problem with the m1phobos.tab file. I was suspicious when I noticed that the crater Flimnap was centered at longitude 350, i.e. very near the 0/360 degrees where the geometry joins. It turns out that the radii at 0 and 360 degrees are identical over most of the most of the model, but *not* around 50-70 degrees north. Have a look:

Code:

48.0000    0.0000   10.1403    48.0000  360.0000   10.1403    50.0000    0.0000   10.0356    50.0000  360.0000   10.0373    52.0000    0.0000   10.0051    52.0000  360.0000    9.9945    54.0000    0.0000    9.9701    54.0000  360.0000    9.9474    56.0000    0.0000    9.8698    56.0000  360.0000    9.8557    58.0000    0.0000    9.7666    58.0000  360.0000    9.7633    60.0000    0.0000    9.7445    60.0000  360.0000    9.7431    62.0000    0.0000    9.7540    62.0000  360.0000    9.7546    64.0000    0.0000    9.7382    64.0000  360.0000    9.7398    66.0000    0.0000    9.6929    66.0000  360.0000    9.6941    68.0000    0.0000    9.6340    68.0000  360.0000    9.6344

--Chris

Very well observed, Chris!

how about simply averaging the given radii at 0 and 360 degs within the problematic interval of 50 and 70 degs? That should do the trick and probably makes the model look smoother there under illumination.

I'll just do it and check.

Fridger

t00fri wrote:

Very well observed, Chris!

Thanks! A little emacs keyboard macro did most of the work...

Quote:

how about simply averaging the given radii at 0 and 360 degs within the problematic interval of 50 and 70 degs? That should do the trick and probably makes the model look smoother there under illumination. I'll just do it and check.

Maybe just always average the radii at 0 and 360 degrees? It'd be good to emit warning when the radii differ.

--Chris