Hi all,

since both current surface textures of the Moon, Chang'e 2 and LRO-WAC, have their PROs and CONs,

http://forum.celestialmatters.org/viewtopic.php?t=426

I decided to go the hard way, hoping to finally arrive at a texture that satisfies my own ambitious criteria....

1) Perfect alignment with the official 32k LRO-LOLA normalmap
2) A minimum of fixed shading for good cooperation with normalmap (dynamical shading)!
3) Scientific coloration (UVVIS, 5 band colors, Clementine mission)
4) Excellent contrast, not worse than original imaging

After a lot of work, I am now really pleased with the result. Note that unlike previous moon textures, my new texture has almost no left-over artefacts. In this thread, I'll sketch the procedure applied and show some nice illustrations from my final texture.

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++
If you like the texture too and want me to wrap it up for download, let me know.
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++

So, how did my hard way look like?

Requirement (2) only leaves 2 alternatives: Either I base my work on the new Chang'e 2 texture with it's very high sun angle or I attempt to eliminate the strong shading of the craters of the LRO-WAC map by means of standard image manipulation tricks.
The latter alternative will always deteriorate contrast and image quality in general.

Hence, because of my request (4), I went for the Chang'e 2 texture!

The real hard part of this strategy was that I had to systematically remap the Chang'e 2 imaging data to match the LRO-WAC normalmap. Again two alternatives:

Either one uses the qnet application of the ISIS3 package for remapping.

http://isis.astrogeology.usgs.gov/Appli ... /qnet.html

Yet this requires unprojected images etc. => Kind of messy.

The other alternative is to do the remapping by hand. Given my longstanding expertise with GIMP, I went for the latter "handicraft" solution. Since I need the GIMP also for coloration (item 3), I restricted the size of my moon texture to 16k. This can still be handled fairly fast in GIMP and seems a VERY good compromise between lots of moon details and required hard disk space...If one uses high-quality DXT1 (dds) format, no VT tiles are needed for 16k texture sizes!

I'll now rush through the less trivial steps with GIMP, but I need to assume a basic familiarity with GIMP usage. However, if sufficiently many people would care for a more tutorial type of description, let me know. In case of real interest, my time is always well spent

Remapping by hand (GIMP)
====================

Basically the required steps are obvious:

I first scaled (cubically!) both the LRO-WAC and the Chang'e 2 original surface textures to 16k size with 2:1 aspect ratio. As to the extra complication with the Chang'e 2 non-2:1 ratio, see my previous thread:

http://forum.celestialmatters.org/viewtopic.php?t=426
(c.f. Chang'e 2 map CONs)

Since the LRO-WAC map matches the LRO-LOLA normalmap perfectly, I use the LRO-WAC map as an alignment reference and place it as lowest (background) layer. The Chang'e 2 texture is subsequently loaded (using "load as layer") on top of the LRO-WAC map. The blend mode remains at normal and the opacity slider in the layer dialog is set to 50 - 60 %. This way you can immediately inspect (at high zoom) the various places where the Chang'e 2 map is misaligned, since the reference texture shines through! Unfortunately, the misalignment is NOT global, and that's why the tedious local alignment procedure is needed.

Here is how I did it: I organized the 16384 pixel wide Chang'e texture into 11 (1600x1200) sized screens of 1200 (pixel) lines, with 7 lines of 11 screens to cover the full 16k texture. So I had 77 screens to realign by hand .

After inspection of misalignment areas in each screen, I selected these either with the lasso or the rectangular selection tool. It is crucial that beforehand, you need to activate "feather edges" with a big radius of ~ 50 pixels. This will smoothly re-integrate the marked misalignment region back into the texture after alignment. Crucial: activate the top layer (i.e. the Chang'e 2 map) by clicking on it in the layer dialog. Otherwise nothing will react.

By far your quickest option is by means of a key shortcut:

Just push SHIFT + ALT, keep it pushed throughout the operation and carefully move the selected texture region via mouse dragging, such as to match the LRO-WAC reference perfectly! Use high zoom and tiny craters as marks. Move the mouse very gently...Moving the selection creates a "floated layer" or "floating selection" in the layer dialog. If you found the best position, click mouse-right -> Layers -> Anchor layer. That's it for ONE screen .

Sometimes, however, the misalignment region is smaller than 1 screen size, whence several independent selections/re-alignments per screen are required

After several hours of high-concentration work and 77 screens, one can look proudly at a perfectly re-aligned high-quality Chang'e 2 moon texture!

Coloration with UVVIS 5 band template (Clementine)
======================================

Here another ingenious GIMP function is used: Colors->Map->Sample Colorize.

Its basic function is to map the gray tones in the target image (Chang'e 2) to the shades of a color, given a colored image or better (and faster), a colored gradient image as template. If the template color image contains different base colors (like in our case), one first color-selects one type of color. Later one inverts the selection to map the remaining color.

We got mainly two basic colors in the 5 band colored UVVIS reference image.: a pale greenish one and beige-orange shades. First select the greenish color (with color select, 10% threshold, and "add to current selection" activated): Make a gradient image from those. Here it is:



Then invert the selection and make another gradient image thereof. Here it is:


The GIMP manual describes in detail how to prepare color gradient images!

Next, open the Colors->Map->Sample Colorize function and start with the greenish gradient template on the right. On the left the 16k grayscale Chang'e 2 texture should show up with selections activated and correponding to the greenish regions after the coloration. It is crucial to switch the color mode of the grayscale Chang'e 2 texture to RGB before calling the Sample Colorize function (Image->Mode->RGB)! Otherwise, it cannot be colored .

The GIMP manual describes, how to transfer the color selections in the 1k UUVIS reference image to our grayscale 16k Chang'e 2 texture!

Activate "use subcolors" near the bottom of the Sample Colorize dialog. Then hit "Get Sample Colors". If this is finished, hit "Apply". Finally, close the Sample Colors dialog, invert the color selection in the Chang'e 2 texture, change the gradient template, reopen Sample Colors and repeat the same actions there.

Your texture should now look VERY close to the original UVVIS texture!

For a deeper understanding you best have a look in the GIMP manual under "Sample Colorize". Once understood, it's really easy and most powerful!

Finally, I added in some amount of "wavelet denoise" . That's it!

Now it's time for some illustrations. Since the textures are very big, I rather show you some zoomed-in partial views:

Here is a tiny sized (1024x512) copy of my finalized, colored and remapped global Chang'e texture:


So you can judge the faint shades of color altogether. The coloration is virtually indistinguishable from the UVVIS 5 band original!

Here are some zoom-ins. So you can judge the delicate shades of color:

[Click on images by all means!]











Enjoy,
Fridger

Excellent job.
Really like the colors.

Tim

Here is a representative selection of both high-detail zoom-ins and global views of my finalized Moon texture, as displayed in Celestia(.Sci).

Note the delicate color shades in the zoom-ins! Also, note the excellent contrast of the texture images...

Enjoy,
Fridger

[Click on images by all means!]

Looks very, very good! 77 screens! Is the normal map also 16K, or did you use a higher resolution VT together with the 16k surface?
Guillermo

_________________
Guillermo Abramson
Bariloche, Argentina

Thanks Guillermo,

the official LRO-LOLA normalmap submitted by emazarico@LOLA is a 32k VT:

http://www.celestiamotherlode.net/catal ... on_id=1549

It's not bad, if the normalmap resolution is higher than that of the surface map. The converse should of course always be avoided. While I also made a 32k Chang'e 2 VT with my tools, it can hardly be remapped and colored by hand . But from a comparison with that 32k surface texture, I concluded that a perfectly matching and colored 16k texture is superior.

Regards,
Fridger

Dear Fridger,

While I cannot professionally judge about the quality, the texture at least looks perfect to me! So don't hesitate to offer it for download if you are ready to do it!

Thanks for the efforts.

Kind regards

Andy

Hi Andy,

thanks! Your request has been recorded...

Cheers,
Fridger

Hi Fridger,

I like the texture and a download would be very much appreciated by me too. (And most probable by a lot of others.)

Thanks for your efforts.

Best regards,

Guckytos

Hello Fridger, I think your one the best lunar map possible, given the actual data availability, so thanks a lot for all the effort needed to obtain it.
I would be very pleased to use it in my students courses, if possible, so please add me to the list.
Well done, appreciated.
Bye

Goofy

_________________
ASUS G750JH- Intel i7 2.5 GHz-32 GB RAM-SSD 128 GB+SSD 256 GB-GeForce GTX 780M 2 GB

OK guys,

request noted. As soon as I find a little extra time, I'll wrap it all up (with README etc) and offer it here for download. Since it is CRUCIAL to install my remapped 16k Chang' e 2 base map together with the official 32k LRO-LOLA normalmap, I am not sure yet what is the "best" (i.e. most fail-safe) setting for both packages.

It would be trivial, of course if I'd set up the twin-package already beforehand, but offering the LOLA package here would not be good style...

Fridger

Thank you, Fridger, take your time.
BTW, is there any possibility for us to have the 32k version of the "Chang" texture, so matching the LRO-LOLA normalmap resolution?
I hope your reply will be yes... please!
Bye and thank you once again.

Goofy

_________________
ASUS G750JH- Intel i7 2.5 GHz-32 GB RAM-SSD 128 GB+SSD 256 GB-GeForce GTX 780M 2 GB

a question
you are using the "LRO-LOLA " data and not the
LRO-WAC DEM

the stereo imaging dem is much better , except for the poles

Goofy,

yes, it will take a little while, since we (the Celestia devs) are currently busy reactivating the development of Celestia-1.7.x based on the cross-platform Qt4 widget library.

As to a possible 32k VT version of the Chang'e 2 texture: I initially prepared a 32k VT version using my texture tools, but it turns out to be badly misaligned with the LRO-LOLA normalmap data. Hence I was forced to perform a most tediuos manual remapping of Chang'e 2!

See my detailed description above:

http://forum.celestialmatters.org/viewtopic.php?t=428
http://forum.celestialmatters.org/viewtopic.php?t=426

Already for my 16k size restriction this amounted to realigning 77 screens (of 1600x1200 pixel). For a 32k size this would have amounted to 308 (!) screens. This is beyond my "patience". Coloration was another size-limiting factor.

My colored 16k ~flat surface texture is just perfect in quality, size and resolution to go with the 32k LRO-LOLA normalmap VTs.

Fridger

Before starting off, I did examine both DEM sources, including your 64k VT set . During this early exploratory stage I wanted to refer to unmodified official published DEMs in order to evaluate their unprocessed alignment, contrast, smoothness etc. Near the poles the LRO-WAC DEMs didn't excite me. Since other constraints led me to restrict myself to a 16k colored surface texture (Chang'e 2), the 32k LRO-LOLA normalmap from the Motherlode was just fine for the time being. At highest zoom the LRO-LOLA VTs are getting 'blocky', however.

Before releasing my Moon results, I'll certainly return to the LRO-WAC DEMs from a somewhat different point of view. We'll see...

Thanks for reminding me about the LRO-WAC DEMs.

Fridger

Thank you Fridger, I understand the problem of misalignment, so I'll exclude the possibility to use the 32k matched with LRO-LOLA.
But, when we knew of the Change 2 monstrous texture, I downloaded it and tried to convert it to a 32k map using the command-line tool gdal_translate of the FWTools, but it looks like this doesn't work under windows 7, at least for what I know, (but may be I'm wrong, anyway).
So if possible I would like to have, for testing it (as is and without normalmap), the 32k Chang 2 you was able to obtain with FWTools (with Linux, I suppose, correct?).
Could you be so kind to share it .... please?

Anyway I'll wait patiently for the full 16k, when it will be available.
Thank you for the clear explanation and for all the stuff.
Bye

Goofy

_________________
ASUS G750JH- Intel i7 2.5 GHz-32 GB RAM-SSD 128 GB+SSD 256 GB-GeForce GTX 780M 2 GB