Celestia Forums

Hi, I just came across some pretty detailed shots of the 6.5 km deep Valles Marineris Canyons on Mars. Thats about what I want to get at eventually with 16bit/channel accuracy;-)

Enjoy!

Bye Fridger

woooooooow......

Looks like a photo from an electron microscope!

Hello guy.

Please, tell me, where you found these (excelents) bumpmaps? Like the eath bump 16 bits?

So much thanks.

Hey, Fridger-

What resolution is that image? It must be huge. Much bigger than the Blue Marble Earth textures. There's way more than six pixels down the edge of the canyon! How much of Mars' surface is mapped to this detail?

I seriously doubt we'll get these kinds of results with just normal maps. We need some creative way to use very hi-res textures.

- a b i o g e n e s i s -

There is no useful information in your post, only scream-?I am so coooool!? Please mark some points A,B,C from the image on the map and show the corresponding cut of working texture(in actual pixels).After that we?ll discuss resolution, light directon, view direction and shadering.
GARR

I this Fridger was just showing a striking picture of Mars--and it certainly is an amazing image. He didn't claim that it was a Celestia screenshot; it's meant to show what may be possible with Celestia in the future.

--Chris

Chris,

When you say "future", do you mean "this is a feature that may be implemented by my offspring in a few generations" or do you have some neat tricks for using big textures in the works?

- a b i o g e n e s i s -

abiogenesis wrote:Chris,

When you say "future", do you mean "this is a feature that may be implemented by my offspring in a few generations" or do you have some neat tricks for using big textures in the works?

- a b i o g e n e s i s -

I meant that I'm working on huge texture support right now, but it's not close to done yet. A lot of the framework exists already, in the form of a recursive culling system I implemented it to limit the number of triangles draw when the camera is very close to the planet. No promises on how well this is going to work though

--Chris

GARR wrote:There is no useful information in your post, only scream-?I am so coooool!? ...
GARR

Thanks for your "charming" remarks. Please note that I
tend to form a pretty low opinion about "fast-shooters" like you, who criticize others before attempting to understand what's actually going on...

How about reading my above post carefully again?

You are a member of this forum since May 3 2003 and since then have come up with exactly one further "brilliant" reply to alan_federman:

GARR wrote:Thanks! You rules!

Truly remarkable, but I think in proper "cracker English" this should be rather "You rulez"? Indeed, the forum lives through such powerful
contributions...

Bye Fridger

I suppose, most of you have certainly realized that the above amazing view of the Valles Marineres on Mars is actually a 'collage' of hires photographs rather than a Celestia texture...

Yet, let me illustrate below, where I stand with the MOLA 16bit/channel 16k sized normalmap attempts. These are indeed Celestia screendumps. The main Mars texture is a 8k FLAT one from Mario/Space-Graphics.com. So any 3d detail you see is actually from the normal map!

With my somewhat oldish equipment, it was not entirely trivial to handle such big textures. Remember, I have 512MB of RAM and an old GeForce2 GTS/32MB card.

Perhaps some of you are interested, how one may arrive nevertheless at a 16k normalmap that is of very high quality. Below I summarize the commands, I used under Linux. An essential role was played by the ImageMagick command line utilities (5.4.7), that exist also for Windows! The great thing is that -- unlike Photoshop-7 or the standard GIMP -- they work perfectly for 16bit/channel and also, under Linux, collaborate with a STDIN/STDOUT pipeline. Hence these utilities are very practical together with Chris' recent 'nm16' 16bit/channel *.img -> 8bit normalmap converter code!

1) The highest resolution 64pix/degree MOLA Mars elevation data come in form of 4 tiles (11520x5760) in 16bit grayscale *.raw/*.img format:

megt00n000gb.img, megt00n180gb.img, megt90n000gb.img, megt90n180gb.img.

from

http://wufs.wustl.edu/missions/mgs/mola/megdr.html

2) after downloading, I seamlessly reassembled these 4 tiles into a huge 23k 16bit/channel elevation texture in *.img format by means of the very flexible ImageMagick utility 'convert', after renaming *.img -> *.gray. The trick was to do this in console batch mode, such that no memory is vasted on the X-server etc.

2a) reassemble megt90n000gb.gray, megt90n180gb.gray left-to-right into a 23kx5.76k *.gray texture at 16bit/channel depth via

convert -depth 16 -size 11520x5760 megt90n000gb.gray megt90n180gb.gray +append upper.gray

2b) reassemble megt00n000gb.gray, megt00n180gb.gray left-to-right into a 23kx5.76k *.gray texture at 16bit/channel depth via

convert -depth 16 -size 11520x5760 megt00n000gb.gray megt00n180gb.gray +append lower.gray

3) reassemble upper.gray and lower.gray into a 23kx11k elevation texture top-to-bottom:

convert -depth 16 -size 23040x5760 upper.gray lower.gray -append Mola23k-16bit.gray

4) In principle one should now reduce the size to 16kx8k before converting the elevation map into a normal map. However, strong artefacts seem unavoidable this way. Hence I converted first to a 23k normal map with Chris' tool and thereafter size reduced the result with the fancy Lanczos algorithm to 16kx8k!
This requires a renormalization step at the end to retain a proper normalmap after resizing!. Before that, we must do a warped offset by width/2 in order to get the prime meridian into the center.

Here are the steps:

nm16 23040 11520 100.0 0 < Mola23k-16bit.gray | convert -resize 16384x8192 - Mola16k-8bit-nm16.bmp

Chris' conversion routine is applied and STDOUT piped again into ImageMagick's convert routine to do the resizing. The '-' sign in 'convert' indicates as usual that the input to 'convert' comes from STDIN.

Finally, the resulting normalmap is loaded into GIMP, the width/2 offset is done and at the end the GIMP-normalmap plugin does the renormalization of the map!

That's it! Now have a look at the result. You may compare with the above hires photo and judge yourself, how far we are still away from that quality with Celestia;-)

Since these steps take several hours on my 1GHz PIII machine, I wrote all above commands into a shell script and went to a restaurant with my wife while things were working;-). Also note that the 23k *.bmp files are about 531MB and the 16K *.bmp is ~ 400MB!
The *.bmp files may then be converted with Pixel's 'bmp2dxt.exe' tool under Windows to 24bit DXT1c without using much RAM, or alternatively under Linux, with my own DevIL based tools ('texconvert')...

Bye Fridger

Actually fridger your coming pretty damn close to government work there...I cannot see why you couldnt make something like the above screenie...now...

sadly without the vram to spare I suppose would hinder the process...

I did a quick forum search but I could not find anything about this new converter tool nm16. Where can I find this?
Also, well, I know that fridger offered to make his devil-based dds converter available a long time ago and there was not a lot of response, but is there a chance to get this tool too? Building from source under linux should be no problem...

thanks, steffens

steffens wrote:I did a quick forum search but I could not find anything about this new converter tool nm16. Where can I find this?
Also, well, I know that fridger offered to make his devil-based dds converter available a long time ago and there was not a lot of response, but is there a chance to get this tool too? Building from source under linux should be no problem...

thanks, steffens

Correct, nm16 has not yet been distributed. I might even be the only one besides Chris who presently has the code...?

I am sure Chris will not hesitate to make nm16.cpp public. For Linux users its compilation and use is absolutely trivial. Since it works with STDIN and STDOUT as (pipe) input/output, it will need some further (trivial?) coding for Windows users, I guess.

Bye Fridger

Since some people have asked where the above hires Valles Marineris photo was from, here is the URL:

http://photojournal.jpl.nasa.gov/catalog/?IDNumber=pia00006

Another very nice photo is this one:

http://photojournal.jpl.nasa.gov/catalog/?IDNumber=pia00005

Bye Fridger

Sorry for late reply, but I was fishing.
Ooooh... You are so volcanic ( like the well-known volcano on Sicily).
OK! In some things you are right.
I ask you to write all posts in the future in such style ( more science, less
blah-blah).
This pictures are real, and so far from dream...I cannot see critical
difference between old pics and the latest ones...
Couse I don'y know english well enough, I'll write in
telegraphic-shizophrenic style (Kurt Vonnegut, you know?).
Q1 : used texture? ( file, url, resolution )
Q2 : show 6 pictures -
Use CONSTANT location!
1. Basic, early morning, volks Normal Map
2. Basic, late evening, volks Normal Map
3. OpenGL/NV comb, early morning, volks Normal Map
4. OpenGL/NV comb, late evening, volks Normal Map
5. OpenGL/NV comb, early morning, new Normal Map
6. OpenGL/NV comb, late evening, new Normal Map
Please work in actual pixels, and don't use hard jpg-compression, because
your pictures are too blurry (excluding far-away structures), and there are
some jpg artefacts on them. And 800-600 is enough.

GARR wrote:Ooooh... You are so volcanic ( like the well-known volcano on Sicily).
OK! In some things you are right.
I ask you to write all posts in the future in such style ( more science, less
blah-blah).

And we ask you to be polite.

GARR wrote:This pictures are real, and so far from dream...I cannot see critical
difference between old pics and the latest ones...

What did you expect? A real live experience, all of sudden? Did they take your money and deceive you? Remember, these Normalmaps are a WIP, unfinished.

GARR wrote:Please work in actual pixels, and don't use hard jpg-compression, because
your pictures are too blurry (excluding far-away structures), and there are
some jpg artefacts on them. And 800-600 is enough.

And what if you STOP demanding things, and start contributing (or at least, asking better)? Fridger makes a really good and hard work here, and we appreciate it very much.
If you don't like his work, you can start making your own and showing us what your improvements are.

great work ,you but my maps to shame . I still have a LOT more to learn

GARR wrote:Ooooh... You are so volcanic ( like the well-known volcano on Sicily).

Not at all, but I am simply not used that people talk to me like this...

I ask you to write all posts in the future in such style ( more science, less
blah-blah).

Many people actually prefer to read about these things in a rather less technical style. The main idea here is to produce some sort of qualitative status report. Many people are mainly interested in the "end product" and how this might look like rather than in the technical details.

And before I do something specially for you, I suggest you first let us know whether you have more to offer than "You rulez!"...

Sorry for late reply, but I was fishing.

Q1 : used texture? ( file, url, resolution )

Are you "fishing" again? At least you should try to read my previous posts before writing in this style!

Go back one page, I gave the URL's of the MOLA elevation data, the resolution and every single command I used to produce the shown results...

Q2 : show 6 pictures -
Use CONSTANT location!
1. Basic, early morning, volks Normal Map
2. Basic, late evening, volks Normal Map
3. OpenGL/NV comb, early morning, volks Normal Map
4. OpenGL/NV comb, late evening, volks Normal Map
5. OpenGL/NV comb, early morning, new Normal Map
6. OpenGL/NV comb, late evening, new Normal Map
Please work in actual pixels, and don't use hard jpg-compression, because
your pictures are too blurry (excluding far-away structures), and there are
some jpg artefacts on them. And 800-600 is enough.

Is this all or do you have further
work I can do for you?;-)

Bye Fridger

john Van Vliet wrote:great work ,you but my maps to shame . I still have a LOT more to learn

Thanks John, clearly the main point is that we make progess together in this important aspect of Celestia. What is really needed is a Windows version of Chris' nm16 tool. Unfortunately, I cannot produce that.

Using bad textures in Celestia is like

a powerful hifi set with bad loudspeakers;-)


Bye Fridger