New spiral galaxy types rendering

[t00fri]

I've missed for a few days and I'm glad to see that the things are going on quite good.

Fridger and Toti congratulations! The screenshots are really spectacular.

To enhance a bit the visual appearance and simulate the dark clouds I suggest to set some blobs with a dark color and a low transparency.

Many thanks guys. I think that when Fridger will exctract the 14000 galaxy database this will be the greatest cosmological enhancement of Celestia so far!

Paulo,

I think next we (I) might study coloration and luminosity distribution in a more systematic fashion in the code. I am almost certain that it is much better to start from a (H,S,L~V [alue]) =(Hue, Saturation, Luminosity) model rather than using (R,G,B) color parametrizations!

The changes in color could then simply be controlled with one parameter (Hue) and similarly for L (Luminosity). Tonight I shall try a Gaussian decrease in L (exp(-a*(255-L)^2) to make the luminosity concentrated more in the galaxy center.

The idea is to arrive at a semi-realistic coloration, just from the typical catalog input parameters on luminosity/color:

Vmag (visual), Bmag (blue) + Surface brightness.

We can build a semi-empirical (Hue) color lookup table based on photographic color images and the associated values of (Vmag, Bmag, Sb).

The (H,S,V) parametrization uses essentially "polar coordinates in color space" with a direct correspondence with the complex numbers. In the HSV color scheme, the parameter Hue(H) is an angle measuring position around the color wheel. Red has a Hue of 0 or 360 degrees; green has Hue 120 degrees, and blue has Hue 240 degrees.


Also I want to gain some experience with that approximate formula (that I quoted further up), expressing the galaxy inclination 'i' in terms of the ratio d/D of its min/max extension parameters, as tabulated in catalogs. The underlying idea is that most galaxies should have a /spherical/ shape (d=D) when looked at 'face-on'. Then, for d<>D in observations, one may easily calculate the corresponding inclination...(a trivial piece of homework, really).

Bye Fridger

[PlutonianEmpire]

Wow! Those screenshots look absolutely terrific!

Can't wait to see those in live 3d....

[Toti]

Hi,

Since blobs now have different colors and transparencies, I deliberately changed the openGL blending mode to avoid having to sort all blobs for each galaxy from back to front prior to render them (otherwise it would look ugly). This saves a lot of CPU, but also introduces the artifact of overbrightness when galaxies are watched edge-on. As I said, this is easy to fix.

I think the galactic centers should be brighter, or more densely packed.

Perhaps enlarging each blob so that they overlap more would make things brighter?

The center bulge issue is 50% of the whole problem. Enlarging blobs makes the galaxy look more "gaseous" and realistic, and fills the center with bright, dense matter. It is much better, but inherently involves more frame buffer overdraw, implying a strong performance hit in low end systems (specially when using small FOVs) Also, enlarging only the centermost blobs conflicts with the LOD: as we move closer, the larger billboards are rejected first, so a huge center "hole" shows up.
Perhaps a better solution would be to use a quality parameter in the .cfg file, with a "common denominator" default, so users of advanced hardware could enjoy a better galaxy portrayal.

These are with default colours and light gain at ~50%. Some color fringing
appears around galaxy edges, and distant galaxies flicker due to aliasing but
overall the appearance and performance seem excellent on my PowerBook.

Indeed there are some incorrectly colored blobs at the edges. I didn't have time to address this problem. There is also the remaining issue of rendering the missing opaque clouds.

Bye

[Paolo]

I think next we (I) might study coloration and luminosity distribution in a more systematic fashion in the code. I am almost certain that it is much better to start from a (H,S,L~V [alue]) =(Hue, Saturation, Luminosity) model rather than using (R,G,B) color parametrizations!

The changes in color could then simply be controlled with one parameter (Hue) and similarly for L (Luminosity). Tonight I shall try a Gaussian decrease in L (exp(-a*(255-L)^2) to make the luminosity concentrated more in the galaxy center.

I agree with this decision. This approach should even lead to a semi-realistic wavelenght filter feature simply changing the colortable.

Bud about dark clouds I was referring to something like your image of the sombrero galaxy as seen in visible spectra.

Some kind of dust must be considered in order to create umbra and penumbra zones.
Indeed this requires that Z-sorting of blobs must be enabled and Toti has said that now is disabled.

[NoXion]

That's truly brilliant! I can't wait to see these in action.

My gripe with the old galaxies is that they were far too dark; I know that in real life they aren't that bright, but Celestia is a display program is it not?

[ElChristou]

Now I dreaming of a tool for "painting" galaxies with blobs... setting size, intensity, transparency and color... would be cool, no?

[t00fri]

Now I dreaming of a tool for "painting"
galaxies with blobs... setting size, intensity, transparency and
color... would be cool, no?

I am not so sure...at least from my point of view.

Since years we had those efforts related to "hand-making"
of individual deep-sky objects. The results were often
beautiful and unbeatably detailed => cf. Motherlode!

But in view of our typical catalogs with 14000 such
objects, we must get hold of methods that
eventually will be suitable for "mass-production".
Otherwise, we are not simulating the universe, but rather are
preparing a "museum" with a few individual hand-selected
"exhibition pieces" (good to show your girl|boy - friends on Sunday afternoon )

I see the real challenge now ( repeating myself from further
above), to arrive at semi-realistic and pleasing galaxy layouts
(including Hubble shapes, colors, luminosity distributions,
inclination and correct extensions), just from a few input
values that are extracted from the professional catalogs!

Bye Fridger

[NoXion]

To be perfectly honest, I think this new galaxy code is good enough as it is. But if you can make it even better without sacrificing CPU time, I will be amazed and worship you as programming gods

I'd be willing to download Celestia again even if this was the only improvement.

[lukr]

I've been closely following this thread over the last week. I'm so impressed with the developments. Fantastic work!

When will it be available as an add-on, do you think? Will it run under Celestia 1.3.2 ?

Cheers, Luk

[t00fri]

I've been closely following this thread over the last week. I'm so impressed with the developments. Fantastic work!

When will it be available as an add-on, do you think? Will it run under Celestia 1.3.2 ?

Cheers, Luk

It will go into the main Celestia distribution. But it still needs some further development work.

Bye Fridger

[Toti]

Hi,

Some kind of dust must be considered in order to create umbra and penumbra zones.
Indeed this requires that Z-sorting of blobs must be enabled and Toti has said that now is disabled.

Indeed. I just enabled depth sorting of blobs, so opaque cloud rendering should be possible now. Overall, it's quite fast (the code sorts only those blobs that will be sent to the openGL pipeline: it could be event faster, I think)



Bye

[Paolo]

Very beautiful Toti! It is almost exactly what I was meaning.

Going beyond this, in order to increase realism it should be possible to
use perlin noise textured blobs. The noise3d fractal function should be
used to create the squared texture instead than filling it with white color.

After that the gaussian transparency function should be applied to the
texture to create the halo effect on the blob. Using this texture the
amount of detail and randomness will improve a lot.

I have some other ideas about this issue but it will be better to proceed step by step.

Waiting for Fridger's work on HSL (Hue, Saturation, Luminosity) model
and colortables experiments I think that the current results obtained by
Toti and Fridger are quite remarkable, even considering that Chris was
not yet involved at all in this OpenGL stuff.

[t00fri]

Very beautiful Toti! It is almost exactly what I was meaning.

Going beyond this, in order to increase realism it should be possible to
use perlin noise textured blobs. The noise3d fractal function should be
used to create the squared texture instead than filling it with white color.

After that the gaussian transparency function should be applied to the
texture to create the halo effect on the blob. Using this texture the
amount of detail and randomness will improve a lot.

I have some other ideas about this issue but it will be better to proceed step by step.

Waiting for Fridger's work on HSL (Hue, Saturation, Luminosity) model
and colortables experiments I think that the current results obtained by
Toti and Fridger are quite remarkable, even considering that Chris was
not yet involved at all in this OpenGL stuff.

Paulo & Toti,

you seem to find the above image a big improvement. Unfortunately, I don't. It doesn't really look much like a galaxy anymore. The boundaries are way to sharp and the "fluffyness" is entirely gone. I agree that one might use /elements/ of this for inside areas of galaxies.


I will only be able to do some real coding (HSV etc) over the weekend. My job tends to leave me in a rather "soft" state in the evening during working days . Beer and a few emails is all that I can absorb||produce under such conditions ...

Bye Fridger

[Cham]

I agree with t00fri, the sharpness is too strong on the last picture. The galactic edges are looking weird.

[Paolo]

The boundaries are way to sharp and the "fluffyness" is entirely gone.
I agree that one might use /elements/ of this for inside areas of galaxies.

I think that Toti has implemented the dark clouds suggestion as an
example. IMHO the side view is not so bad, it approximates the sombrero
galaxy view. The things has to be improved a bit more, but we are not so
far from having a good result.
I think that currently we are in an experimentation phase. Some
tentatives will be discarded, so new paths has to be suggested, discussed,
implemented in order to refine the results.

I will only be able to do some real coding (HSV etc) over the weekend.
My job tends to leave me in a rather "soft" state in the evening during
working days. Beer and a few emails is all that I can absorb||produce
under such conditions ...

Relax is important. I remember that when I've joined these forums
somebody replied me, "Hey we are doing this for fun!".
We /users/ have large expectations, but you /developers/ are not in a
hurry in order to produce this patch. The really important thing is that it
must be FUN for you to do it!

Fridger and Toti
It should be kind if you will publish the updated diffs of your tests. So if
someone else like Dirkpitt will be interested will be able to give a hand or
do some more experimentation.

[t00fri]

I tried to get into some real comparisons...

So I installed Seldens Messier dsc file...and, knowing a lot
about pitfalls in this game, picked typical bugs at the first
attempt :

I looked at M65 and M66 in the DSS Real Sky
photograph and compared this to our new galaxy
rendering. The result is depicted here:



What is good and what is bad??

a) Since Selden's DSC file had the galaxy type correctly
marked (Sa and SB, respectively), we see that as to the
shape of M65 and M66 Celestia does quite a decent job
already!

b) Since Selden's list (and Celestia) do not support yet the
galaxy position angle, the orientation is wrong by about 90
degrees. So we need to support that
urgently.

c) The coordinates of Selden's M65 and M66 are not very
accurate. XEphem uses Steinicke's accurate coordinates
instead that accurately match the direct DSS images (see
green circles versus images in the XEphem screenshot.

So we (I ) got to extract the real stuff from Steinicke's
catalog via Perl...

Altogether, I nevertheless find this comparison encouraging.
We see what is still missing, but the rendering of the
galaxies is already quite pleasing.

Bye Fridger

[t00fri]

Here is another nice comparison with the DSS Real Sky
photo that is far less critical:

M101, a famous, almost face-on spiral galaxy that is also
close to circular. Hence the lacking position angle info is not so
apparent.



Enjoy,

Bye Fridger

[t00fri]

... and one more : M74

NOT bad, I think...



Bye Fridger

[Cham]

I'm getting very impatient, with all that great stuff. I'm a galaxies lover.

I can't wait to have a full working version on my HD.



T00fri, what about the andromeda galaxy, and its two satellites ?

[t00fri]

I'm getting very impatient, with all that great stuff. I'm a galaxies lover.

I can't wait to have a full working version on my HD.



T00fri, what about the andromeda galaxy, and its two satellites ?

And what if I compile you a Windows version close to the weekend? It's your fault to run OS X though .

Bye Fridger