Just a test - 64K Moon

[Chuft-Captain]

Does nips/vips do jp2

oops no .It dose not support jp2 but it has 0 problems handling32k(1.5gig) 64k(6gig) and 128k(24gig) maps

Thanks anyway for the reference... you never know...nips/vips may come in useful.
I haven't looked at it in any detail, but i assume that it operates with a similar command-line paradigm to ImageMagick ??... or does it use a GUI ?

[cartrite]

The gimp has a plugin that will open jp2's. But my system was never able to open a big jp2 file. It may have to do with ram. The whole image may have to fit in memory?
cartrite

[t00fri]

The gimp has a plugin that will open jp2's. But my system was never able to open a big jp2 file. It may have to do with ram. The whole image may have to fit in memory?
cartrite

We are talking here about a 992 MB sized JP2 file! This needs special provisions in the code. You bet that the GIMP plugin cannot cope with it... If you are lucky then most of the file will end up in swap space on the HD, but then operation will be dead slow. GIMP doesn't have the appropriate memory management for "monster" files.

Fridger

[Chuft-Captain]

gdal.jpg

65 mins using the 2.0 version of GDAL, and not much else running.

Thanks for the advice... should have a working 64K Moon VT in the next 1/2 hour.

CC

[t00fri]

gdal.jpg

65 mins using the 2.0 version of GDAL, and not much else running.

Thanks for the advice... should have a working 64K Moon VT in the next 1/2 hour.

CC

Excellent! So perhaps it should be a bit more emphasized that the latest version of GDAL is buggy!

Actually my Clementine64k.bin.zip is almost uploaded at CM... . It took 2.5 hours! So I guess you don't need it anymore, but others, perhaps. The rest goes much faster locally than another GB download.

Fridger

[Chuft-Captain]

Excellent! So perhaps it should be a bit more emphasized that the latest version of GDAL is buggy!

Well, not nescessarily...I had also closed down other apps the second time thru, and the fact that it managed to run to 80% performing the same repetitive tasks before failing suggests it's ok (unless it's some obscure memory management bug). If it was really buggy, I'd expect it to fail a lot sooner.
All in all, I'm inclined to conclude that it was something happening on my machine which caused the crash. (Although, I'm not too keen on trying the later version again anytime soon )

Actually my Clementine64k.bin.zip is almost uploaded at CM... . It took 2.5 hours! So I guess you don't need it anymore, but others, perhaps. The rest goes much faster locally than another GB download.

Thanks for that. I still might be interested in it... I understand that there are some black holes/pixels in the version I've just generated which you have fixed in your uploaded version???

Looking forward to a colour version too!

CC

[t00fri]

Please note:
=============

For others who want to do their own 64k Moon VT with my F-TexTools, following praesepe's commands, I have uploaded the binary 64k input file Clementine64k.bin.zip to Celestial Matters. This saves a lot of time, since the 1hour lasting conversion of the JP2 format into PNG format with gdal_translate is becoming superfluous with this file. The latest version of gdal_translate for Windows might be buggy as it seems from CC's run.

Also I have already reduced the 90k texture width to 64k as is needed for the VTs. No other changes were introduced relative to the scientific original! Notably the artefacts have not been eliminated! This is NOT possible with this size manually. We didn't find yet an automatic "cleaning" algorithm within ISIS3...

++++++++++++++++++++ Download ++++++++++++++
http://www.celestialmatters.org/users/t ... 4k.bin.zip
+++++++++++++++++++++++++++++++++++++++++++++

The zip-compressed file is 992 MB long, while after uncompression it's 1.77 GB long!
Hence check your HD space beforehand!

@Daniel: the mere upload of this file took me 2 1/2 hours!!

Enjoy,
Fridger

Please use the latest 2.0pre1 version of the F-TexTools as pointed at by praesepe.
Use the installer for Windows in order to have the registry entry correctly done.
This latest version also offers directly DXT format. I would advise to activate Fast DXT Compression mode, the quality of which is good enough for most applications.
You should select DXT1 format for the Moon VTs. By default the VT tiles are 1k x 1k in size.

[Chuft-Captain]

Hmm,

Nice detail... but I can see lots of obvious joins ( between tiles? ). (click on the pic)

moon64k.jpg

Note the vertical joins running away from the eye in this pic.

Anyone know why this is? Are these artifacts because it's from the original raw scientific data?
This is a PNG version.

CC

EDIT: There are some even worse artifacts at the North Pole:

northpole.jpg

Call me "Mr Fussy" if you like!

[t00fri]

Hmm,

Nice detail... but I can see lots of obvious joins ( between tiles? ). (click on the pic)

moon64k.jpg

Note the vertical joins running away from the eye in this pic.

Anyone know why this is? Are these artifacts because it's from the original raw scientific data?
This is a PNG version.

CC

EDIT: There are some even worse artifacts at the North Pole:

northpole.jpg

Call me "Mr Fussy" if you like!

CC,

indeed, it is unfortunately true that the Celementine-based imaging is of a quite low quality.

I am fully aware of the examples of artefacts that you showed and many more. The sharp mosaic boundaries of different brightness are unfortunately a frequent issue of scientific maps. As a reminder, here is a little piece from the latest Titan map of Feb 2009. Here that issue is even worse:



In case of Titan, I spent many hours to remove all these unpleasant sharp transitions by hand. Clearly, this can be done with patience with 4k or 8k textures. But 64k textures are beyond any possibilities for manual cleaning. They mostly don't even fit into normal image manipulation programs (GIMP, PS,...) . Or every operation is becoming dead slow, since storage proceeds mostly via the HD swap space (rather than the fast RAM).

The polar artefacts that you showed are also very badly visible in the smaller, published color textures of the Moon.

So altogether, I am afraid, there is not much what we can do here at the level of 64k. Presumably the best is to take a smaller sized map ( 8k or at best 16k) as a basis and do tedious repair work by hand...

Fridger

[t00fri]

CC,

it would be important for me to know whether the F-TexTools 2.0pre1 worked fine after installation via the installer? Did you use DXT1 format or PNG for the tiles? No problems encountered? Were the tools known to the system in any directory by just typing their names at the console prompt?

Do you have a MS compiler installed on your machine or at least some runtime package for VS2005 or VS2008?

Thanks for letting me know,
Fridger

[cartrite]

Perhaps, in a few generations, we will have enough spacecraft orbiting these objects to get good quality images. The main problem is changing conditions. One image is taken in March and the image next to it is taken in October. Cloudy moons or planets will never fit no matter what you do. With Mars, it may be possible to remove the clouds and dust but each image will still have different lighting conditions. ISIS3 can use incidence and emission angles to mask certain areas out of an image but I don't think there is any program that will change an image to match the incidence and emission angles to a certain value. I thought the best hope for doing that was the program photomet in ISIS. They have Topo and TopoAtm algorithms but I'm not sure what those models really do. I thought it may have averaged out the shadows but this was probably a wrong assumption.

But any cleaning of images that is done automatically must be started from the lowest level processing. At best, the image of Titan posted above can be used as an underlying mosaic that images can be added to either above or below the base texture you showed. Trying to clean up an image that was already taken to the highest level of processing will probably be futile.

PS. I couldn't build in JP2 support to my Linux version of GDAL so I downloaded FWTOOLS. I decided to try the win32, 2.3 version of FWTOOLS and run it with WINE on Linux. It took 1 hour and 18 minutes but it did work. The Linux version of FWTOOLS also worked but I didn't let it finish. That's as far as I'm going to take this for now because I'm running 1 GB of RAM. Still using this sick computer. I'm also running low on disk space.
cartrite

[t00fri]

Morning, Steve

I did the JP2 conversion also under Linux with FWtools 2.0.6. No problems encountered, except it took far too long (60').

Still trying various destriping techniques for the Moon texture. With the GIMP Filters->Enhance->Destripe plugin and working on an 8k texture, works VERY well! So there must be an equivalent lowpass-highpass mix with suitable boxcars in isis3 that does that stripe cleaning automatically.

Incidentally, in GIMP I need the same filter width = boxcar width = 2 * stripewidth as is recommended for best stripe elimination in isis3! The GIMP plugin has a real time preview, which makes it very illuminating.

Fridger

[John Van Vliet]

--- edit ---

[t00fri]

John,

in your polar view, are those craters very close to the pole fact or fiction?

Fridger

[John Van Vliet]

--- edit ---

[Reiko]

Is it possible to make VT textures with your tools using one of my own textures? Like if i made 16k planet texture or something?

Reiko,

don't you think us Andorians should help each other?

Oh yes!

It is really simple to convert any (big) texture of yours into VTs: Just take an arbitrary tool of your preference to convert your texture into PNG (e.g. GIMP). Let me know your texture's name and width. Note, the ratio width / height should be 2. Then i'll write out the commands for you.

Meanwhile, install the F-TexTools 2.0 and check that they are correctly installed by typing
e.g. 'txtiles' at the Windows command prompt (Windows-Start->Run->cmd.exe). If you see a help text displayed, everything is ready to rumble

If not, let me know what you get...

Fridger

I installed F-TexTools 2.0 but when I type txtiles in the command prompt all I get is this.

Code: Select all

Microsoft Windows [Version 6.0.6001]
Copyright (c) 2006 Microsoft Corporation.  All rights reserved.

C:\Users\rei>txtiles
'txtiles' is not recognized as an internal or external command,
operable program or batch file.

C:\Users\rei>

[John Van Vliet]

--- edit ---

[t00fri]

Reiko,

indeed, as John was writing already, you need to use the Windows installer
(double click on it)

F-TexTools-2.0pre1.exe

in order to have the PATH entry automatically done. After the installation is over, do a reboot to be sure. Then make the check for the help text display upon typing the program name once more. Once you get that help test, you are on the "winning track"

It is precisely the same procedure than wiith any Windows programme.

Good luck,
Fridger

[Reiko]

Aha I didn't restart. That's why it didn't work.

[cartrite]

indeed, it is unfortunately true that the Celementine-based imaging is of a quite low quality.

I am fully aware of the examples of artefacts that you showed and many more. The sharp mosaic boundaries of different brightness are unfortunately a frequent issue of scientific maps.

LRO is due to liftoff next month. ISIS has a few programs to work with that data already. Maybe this will produce better mosaics.
http://lunar.gsfc.nasa.gov/mission.html
cartrite