Lunar Landscape Photography Gallery, Using very high resolution DEMs derived from LROC images Options

[Nirgal]

This is my first attempt at "Lunar Landscape Photography" by creating very high resolution DEMs and renderings from LRO imagery. (Click on image to view the Gallery)



The DEMs used for the renderings consist of about 50-100 million polygons each, modeling the lunar surface in 3D at full LRO resolution of 50 cm per pixel.

Enjoy

[dougkeenan]

Thank you very much! Beautiful work.

[ElkGroveDan]

This is how smart people back in the Middle Ages got reputations as wizards.

Ingenuitive application of existing technology with a creative flair in ways never tried before. My hat's off to you Bernhard!

[djellison]

wizards.

Personally - I think he should be burnt for witchcraft.

[SandiBandi]

Bernhard of Oz - A- ma-zing! I was on terminator with long shadows wow

[nprev]

Agreed; this is revolutionary work, and beautiful to boot.

[Stu]

It's not often we actually get to see Things Change, progress creeps up on us usually, but I really think this is one of those moments when a new process or technique appears which allows us to see things differently. Amazing work.

[Hungry4info]

Personally - I think he should be burnt for witchcraft.

Haha!

This is amazing work, seeing this, you can just imagine actually being there. It's incredible. Has this been done for some of the later Apollo landing sites? (so we can compare images from LRO and from the actual surface).

[Nirgal]

Thank you all very much for the kind words

I would like to add that much of the credit must really go to those true wizards at NASA/JPL for not only making and bringing to orbit these almost unearthly powerful camers like LROC and HiRISE ... but also for sharing the whole image catalog via the internet with everyone in the world !

It is this unprecedented quality together with the unprecedented availability of the raw data that opens the door for everyone to explore new ideas and processing techniques (such as I do with my "planetary landscape photography" projects)

Actually this has been my very first attempt with lunar imagery after my previous work has almost been exclusively devoted to Mars (and yet barely scratching the surface of HiRISE's Terapixel-Universe)
And until recently I was not even aware of just how amazing and extensive an image catalog already exists from the LRO mission

A challenging problem has been the sheer size of the data sets yielding to DEMs with hundreds of millions or even billions of polygons.
So I had to re-write major parts of my software as a true 64 bit application in order to run the core algorithm (that has a peak memory consumption of many times the size of the input data) on such large data sets.
For example, the peak main memory usage during creation of the DEM used for the last one of the images in the gallery has been about 11 Gigabytes.

Has this been done for some of the later Apollo landing sites? (so we can compare images from LRO and from the actual surface).

Yes that's what I'm going to try next ... however the results probably won't be perfectly comparable because of the different lighting conditions (sun elevation and azimuth) of the LRO images compared with those from the ground. For the moon DEMs its also difficult to simulate other than the original sun azimut angle because of the large cast shadows that I can, of course, only reconstruct as smooth featureless shadow volumes in absence of any image information in the totally black (atmosphere-less) shadow areas.

P.S.:
in the original post I forgot to mention the LRO observation ID of the LROC raw images associated with each renedering (they should have been within the image filename but my web gallery script just renamed them as image-000 ... image-001 )
So here they are:

Gallery-Image# : LRO image ID
----------------------
#1 M102230053
#2 M103668324
#3 M103668324
#4 M109305653
#5 M109305653
#6 M109305653
#7 nacl00000e09
#8 nacl00000e09
#9 nacl00000e09
#10 M102230053

[Zack Moratto]

Wow! That's some gorgeous work you have there.

[Bjorn Jonsson]

This is absolutely amazing and what's especially impressive is the almost total absence of visually obvious artifacts, for example 'stripes' (I noticed possible hints of very subdued stripe-like artifacts in only two or three images). I assume the illumination is different from the illumination of the source images in some/all of the DEM renders?

Compared to this, when I tested the programs from the survey paper mentioned by Nirgal in the first message of the High-Res DEMs from single HiRISE images thread the results were bad in all cases. Either the DEM was simply unrecognizable when compared to the original image, the DEM was far too noisy, far too smooth or had other major problems (and I should note that to me inaccurate low frequency features are a minor problem in most cases).

For example, the peak main memory usage during creation of the DEM used for the last one of the images in the gallery has been about 11 Gigabytes.

As a matter of curiosity, how much RAM do you have? I'm curious because I'm about to get myself a 64 bit system; the 2 GB/process limitation of 32 bit Windows has been a major source of frustration for me for some time.

[dilo]

Bernhard, I discovered this thread only now and I'm amazed!
Besides the odd shadow shading in some of these renderings (vaguely recalling Enceladus gaysers ) pictures seems to come from Apollo archive...
I think now you are on top of your personal moon rover!

[Nirgal]

I assume the illumination is different from the illumination of the source images in some/all of the DEM renders?

Simulating other than the original illumination is more difficult with the Moon DEMs compared to e.g. Mars DEMs simply because of the totally black shadows in the absence of any ambient light without an atmosphere.
So within the featureless shadow areas, there is of course absolutely no way to infere anything about the 3D structure of the invisible surface.
But it is always possible to simulate lower sun angles than the original illumination because in this case the shadows just get longer and, fortunately, it is those lower illumination that is usually more visually interesting anyway. To my own surprise it is also possible to vary not only the incidence angle but also the simulated sun azimuth angle quite a bit until artifacts become visible ... So it's a lot of fun to play around with both camera and sun positions until an interesting landscape shot is found.

As a matter of curiosity, how much RAM do you have? I'm curious because I'm about to get myself a 64 bit system; the 2 GB/process limitation of 32 bit Windows has been a major source of frustration for me for some time.

Yes, I feel your pain ! I too came to the conclusion that with the recent explosion in ultra-high-volume planetary data sets (HiRISE etc.) a complete 64-bit enabled tool chain is absolutely necessary for any serious image processing work.
So I bit the bullet and upgraded to Vista 64 Bit plus 24 Gigabytes of RAM and adapted/re-wrote most of my software plus many of the third-party libraries as 64 Bit applications.
Here are the, IMO two most useful links for setting up a complete free 64 bit develpoment environment under cygwin (gnu/unix tool chain under windows)
http://mingw-w64.sourceforge.net/
http://www.horizonchess.com/wpg64/index.php?page=index

BTW: Looking forward to see the 64 bit version of img2png

[Nirgal]

Bernhard, I discovered this thread only now and I'm amazed!

Thanks, Marco ! Glad you like the images
The odd shadows are actually the illuminated "solid volumes" of the originally black shadow areas that sometimes become visible if the simulated illumination differs too much from the original sun position.

[Bjorn Jonsson]

BTW: Looking forward to see the 64 bit version of img2png

Shouldn't be too long until that happens - I got an "Out of memory" error yesterday when working with LRO data