I'm currently exploring the application of my single-image photoclinometric 3D reconstruction algorithm to generate DEMs for various areas of interest, in particular where we have no other (i.e stereo-imaging based) detailed 3D reconstructions yet.
One of the first areas I have always wanted to visit "from the ground" is the famous Dust Avalanche caught live in action by HiRISE.

Follow this link for a small web gallery with color 3D renderings of a DEM derived from the originally published HiRISE image, rendered under various viewing positions and light source directions.



The original 2D image can be found at the HiRISE site under the observation ID PSP_007338_2640

Disclaimer: The single-image derived 3D DEMs are mainly intended for visualization purposes. Although they are quite exact at reproducing relative heights at small scales, they are less accurate with respect to absolute heights at large scales.

Awesome!!!

That is just... well I can't use the word I want to use, so I'll just go with "nuts"...!!!!

Seriously, fantastic work. I get the very real feeling that you've opened the door on a whole new way of looking at, and visiting, Mars with these images of yours.

Although it is well known that I studiously abstain from profanity of any sort, your images did cause me to utter a few choice expletives, Nirgal. Marvelous!!!

Wait Where did my jaw go?!

Oh, there it is ..... ON THE FLOOR!!!!!!

Incredible work Nirgal.

On Mars, On Mars, On Mars!

This is awesome. Are you taking into account the fact that apparantly there are large albedo variations in the source image? The terrain at left in the source image looks brighter than the terrain in other parts of the image.

Man...just want to punch you!

Incredible... like we're soaring right by it!

Makes me wonder if the avalanche made a detectable noise if anyone had been close enough to hear it when it happened.

Very good question, Björn: I for myself was quite surprised on how well the large albedo differences got handled by the reconstruction algorithm that in theory should have more trouble with stark albedo variations.
I did notice remarkable "albedo-robustness" before, in particular when I tried reconstructions of sand dune surfaces that contain large dust devil streak patterns whose extreme albedo differences should in theory disturb the reconstruction process, which, fortunately does not seem to be always the case and the streaked dunes DEMs got reproduced quite well:
Click to view attachment
I suspect that it is the various adaptive regularization terms I put into the optimization process that alleviate the albedo problem to some extend.
Of course the perfect solution would be some sort of true albedo-correction based on stereo or altimeter-based DEMs but then again such data is not available for many regions ...

Wow! Speechless. So beautiful!

awesome and quite unique!!!

Wow, exceptional! I love having the opportunity to see something like this from the ground that I would never actually want to see from the ground (because landslides hurt).

> with this HiRISE data I was able to do some full detail 3D renderings posted at the other thread where it is more on topic

That's correct. Moving posts...done!

As is often the case, I am a little late getting to this party. I am at a loss for words, but "Holy Moses!" comes to mind.

Just wanted to point out that congratulations are in order for Nirgal who's the subject of the first article of next month french astronomy magazine "Ciel et Espace", following a little article published a couple months ago in which this very forum was also referenced.

http://www.cieletespace.fr/image-du-jour/4...martienne-en-3d