Milky Way

[abramson]

Thanks, Cham. I will take all this into consideration for an improved version.
G

[abramson]

duplicate the RGB image and convert the copy to grayscale. After some optional enhancements of brightness+contrast, it will serve as a perfect alpha mask.

The method suggested by Fridger works very, very well, either in Gimp or Photoshop, either with the current png or with the original ESO panorama. Actually, I don't know why I didn't do it right away, since this self-masking technique is something I routinely do with my real astrophotos...

I will eventually update my texture. But this add-on is something reeeeally simple. Anybody should be able to do it themselves to their own satisfaction, tweaking vertex number and normals in the model, model size in the .dsc, orientation if you dare, texture and you'll learn something in the process...

Guille

[t00fri]

duplicate the RGB image and convert the copy to grayscale. After some optional enhancements of brightness+contrast, it will serve as a perfect alpha mask.

The method suggested by Fridger works very, very well, either in Gimp or Photoshop, either with the current png or with the original ESO panorama. Actually, I don't know why I didn't do it right away, since this self-masking technique is something I routinely do with my real astrophotos...

I will eventually update my texture. But this add-on is something reeeeally simple. Anybody should be able to do it themselves to their own satisfaction, tweaking vertex number and normals in the model, model size in the .dsc, orientation if you dare, texture and you'll learn something in the process...

Guille

Guillermo,

good to read that this worked well for you, too. This technique is indeed indispensable if fine details are to be rendered semi-transparent. And it's really easy in addition.

Unfortunately, in the original ESO image (that I did first actually), the many misaligned (faint) image stars along with their "real" counterparts are highly disturbing.

Fridger

[bh]

Can anyone send me an improved mac?

[t00fri]

Guillermo,

I like Cosmographia a lot, and when I saw it I immediately wanted to have a photoreallistic rendering of the Milky Way in Celestia.

Since 40 minutes Cosmographia now has YOUR MilkyWay

Cheers,
Fridger

[abramson]

Since 40 minutes Cosmographia now has YOUR MilkyWay

Chris's quick. He asked first, of course.

Humbly say I that it looks even better than the other one! I'm glad.

[t00fri]

Since 40 minutes Cosmographia now has YOUR MilkyWay

Chris's quick. He asked first, of course.

Humbly say I that it looks even beter than the other one! I'm glad.

Definitely so! It's also the ONLY add-on that I have installed on my system

Cheers,
F.

[abramson]

If anyone is interested (before I repack the add-on), I made a better version of the Milky way texture. It's based on Fridger suggestion of using a self mask for the transparency. Find it here. It's a zipped file with two images (6kx3k): a desaturated one (similar to my previous one) and a "punch" one, with the full saturation and brightness of the long exposure photograph. Just use the one you like best, by renaming it to mw.png, resizing, etc.

There is also a README file with details of the license and the processing steps if you want to try your hand at it. If you use any in public, I'd just love to know.

Guillermo

EDIT (sunday 20 november): For unknown reasons, the link above still doesn't work. Or perhaps the server indeed turned into a pumpkin at midnight. I hope to fix it eventually. In the meantime, find the textures as a google doc here.

EDIT (7 december): The problem with the link above has been solved. I am removing the texture from Google Docs.

[t00fri]

Guillermo,

If anyone is interested (before I repack the add-on), I made a better version of the Milky way texture. It's based on Fridger suggestion of using a self mask for the transparency. Find it here.
...

NOTE: it will be available after midnight today, due to sheer weirdeness in my lab's server, that turns into a pumpkin at 00:00. Sorry Europeans, that's 03:00 UT. (It could be worse, we are at longitude 71W, it should be UT-4...)

Something must be wrong with the link under 'here'. I am getting an error although it's way past 3:00 UT.

Incidentally, some older graphics cards (including mine) require texture widths < 4096
for add-on textures (where a partition is not automatically performed, if I correctly remember). I cannot display the 6000x3000 format, for example.

Fridger

[t00fri]

Guillermo,

after further looking into the alignment issue, I am now convinced that the problem is due to a mismatch between the projection used in the ESO image and the geometry of the Celestia model (spherical). By using the far away reference points of the 47 Tuc and M4 globulars along with some bright stars in the galactic plane, this projection mismatch becomes obvious.

In your add-on the alignment of the 47 Tuc globular is way...way off. And so is M4.

After searching carefully, yet in vain, for a statement on the ESO page about the (vertical) projection used for the image, I suggest to contact directly

St?phane Guisard, sguisard@eso.org,

http://www.gigagalaxyzoom.org/photographers.html

He is the head Optics Engineer for ESO’s Very Large Telescope (VLT). To create this stunning, true-colour MW mosaic of the Galactic Centre region, Guisard assembled about 1200 individual images, totalling more than 200 hours of exposure time, collected over 29 nights, during his free time, while working during the day at Paranal.

If you prefer me to contact him, let me know.

It is specified on the ESO page that the horizontal axis of the MW image runs along the galactic plane, but this info does not fully specify the vertical nature of the projection. Perhaps you simply have to replace the model sphere by a cylinder?

Once the projection used is known, it is easy for me to reproject with ISIS3 the image to exactly the projection required for Celestia. An alternative possibility is that they used galactic coordinates.

Cheers,
Fridger

[abramson]

Something must be wrong with the link under 'here'. I am getting an error although it's way past 3:00 UT.

Fridger's right, the file is still not accessible. I just checked, and I put it in the right place, with the right permissions. I ignore what happened, and I have no control over it. It will smooth out eventually, so be patient.

Incidentally, some older graphics cards (including mine) require texture widths < 4096
for add-on textures (where a partition is not automatically performed, if I correctly remember). I cannot display the 6000x3000 format, for example.

I know. That's why the add-on texture is 4k. The improved one is just the texture, without the dsc, the model, and the file structure. You should resize it to your conveninece.

[t00fri]

Something must be wrong with the link under 'here'. I am getting an error although it's way past 3:00 UT.

Fridger's right, the file is still not accessible. I just checked, and I put it in the right place, with the right permissions. I ignore what happened, and I have no control over it. It will smooth out eventually, so be patient.

Incidentally, some older graphics cards (including mine) require texture widths < 4096
for add-on textures (where a partition is not automatically performed, if I correctly remember). I cannot display the 6000x3000 format, for example.

I know. That's why the add-on texture is 4k. The improved one is just the texture, without the dsc, the model, and the file structure. You should resize it to your conveninece.

So I'll just wait.

Incidentally, I noted that you have (accidentally?) eliminated the beautiful (colored) nebulosity around Antares and M4. ESO even shows an animation from this region.

(EDIT: Just noted that on the 6000x3000 original not too much of the nebulosity is visible though)

Here is a small hires image of these famous colored nebulosities:

M4_region_ESO.jpg

It would make a phantastic, realistic background for my new globular rendering, notably M4 and Antares...

Fridger

PS: I just wrote an email to St?phane Guisard, asking him for the 'monster' resolution 24k image(!) of the MilkyWay. At this occasion, I also asked him about the projection issue. Let's see what he will answer.

[abramson]

Thanks. I'll be a little busy in coming days, with end-of-the-year academic chores taking over everything else...

EDIT (NEW IMAGE): The nebulosity around Antares in the original image, in the texture, and in Celestia. The conspicuous orange nebulosity, as well as the less obvious Pipe Nebula and the little cloud around rho Oph can be seen. As well as the misaligment of the texture with respect to stars and clusters in Celestia. Still rather pleasing, though.

antares.jpg

(the Celestia rendering was done with the non-desaturated version of the texture).

Guillermo

[t00fri]

Thanks. I'll be a little busy in coming days, with end-of-the-year academic chores taking over everything else...

The nebulosity around Antares is there in the original image:

antares.jpg

Guillermo

Ah OK. I havn't seen your newer versions yet (see above). The nebulosity was not there, I think, in your original 4k add-on image which is all I have besides the originals, of course.

Fridger

[abramson]

Something must be wrong with the link under 'here'. I am getting an error although it's way past 3:00 UT.

Fridger, Chris, everybody interested: please note the last edit of a previous message. I uploaded the 6k textures to
Google Docs while the problem persists at cnea.gov.ar.
G

[t00fri]

Something must be wrong with the link under 'here'. I am getting an error although it's way past 3:00 UT.

Fridger, Chris, everybody interested: please note the last edit of a previous message. I uploaded the 6k textures to
Google Docs while the problem persists at cnea.gov.ar.
G

Guillermo,

many thanks for your efforts to make your v2 MW available. Probably, your institute switched their server off over the weekend? Quite a common practice at times to save resources...


I am currently examining your v2 MW images. So far I looked at mw_v2_lvl.png (non-desaturated). The RGB image is really perfect, very clean and devoid of stars.
However to my surprise, your alpha map is again VERY close to your original one and to my taste creates a too 'solid' impression of the MW. So I made again one following my own recipe, which I think gives a much better appearance. One characteristic difference is that with your alpha mask, some horizontal galactic grid lines are totally covered by the MW, while with my alpha map they are always shining though.

In the meantime, I also made a MW version from scratch myself, starting from the original 6k image.

I used GIMP and there started with Despeckle (r=0.5), which is very effective in eliminating all those thousands of tiny stars in the bright MW regions. Then I used some "special" commands like Selective Gaussian Blur, which is a VERY powerful command for cleaning backgrounds. Moreover some of my usual "weapons" like wavelet denoise and wavelet sharpen. Wavelets are also VERY effective for such kind of problems.

I'll come back here to report, when I am done.

Thanks again,
Fridger

[fungun]

EDIT (sunday 20 november): For unknown reasons, the link above still doesn't work. Or perhaps the server indeed turned into a pumpkin at midnight. I hope to fix it eventually. In the meantime, find the textures as a google doc here.

It seems I cannot download your file with this link either. It says "unavaliable" or that I don't have permission to view.

Thanks,
Tim

[abramson]

It seems I cannot download your file with this link either. It says "unavaliable" or that I don't have permission to view.

The Google Doc is available to anyone with the link, no sign-in required. So says Google.

[fungun]

This must have been the umpteenth time I'ved tried and it worked this time.
Thank you.

Tim

[t00fri]

I have finished re-doing the MilkyWay, starting from scratch with ESO's 6k panorama.

Since I have plenty of practice with image manipulation techniques, I just wanted to go through the various steps myself, to experiment with different kinds of tools and parameter settings during the star-cleaning operation. One task was to keep the resolution and definition of the many dark filiaments as high as possible.

I am quite content with the result. If there is interest, I'll be happy to share it. But I don't want to interfere too much with Guillermo's nice work...

I found out that for the first step of removing the many faint stars, the Photoshop routine Dust & Scratches is clearly superior to the GIMP analog (Despeckle).
Judging from Guillermo's clean result, I guess he also used PS here at least.

D & S has 2 parameters: filter radius and threshold.

The task is to get rid of the faint stars for a minimal radius and a maximal threshold. Best is to start with r=2, threshold=0 (watching the image at high zoom!) and then to move the threshold slider slowly up. The point where to stop is quite well defined. I used r=2, threshold=22 for the stars near the galactic plane.

For the stars further away from the galactic plane, I used r=5, threshold=0, which did a clean job. Of course one needs to select these two star regions separately e.g. with the lasso.

After that (important) first part, I switched over to GIMP/Linux.

What remains is the careful elimination of a number of brighter left-over stars, globular clusters and planets. For that purpose the Clone tool is most handy. Each elimination involves a double action: first select the right color a little away from the star, click CTRL+ mouse_Left then click with the tool onto the star and it is perfectly replaced with the appropriate color. It is essential to chose the diameter of the clone tool not too small (color averaging across the tool's circular shape!)

Finally, a touch with the excellent wavelet denoise plugin ...and that was it for the RGB image. The making of the alpha mask only takes a few minutes and so does the assembly into a RGBA file that can be used in Guillermo's add-on. We discussed that previously at length.

Here are two consecutive views of the resulting MW (unfortunately reduced quite a lot in size)
[ Click on images by all means]

mw1.jpg

mw2.jpg

Fridger