QUOTE (ngunn @ Mar 3 2010, 01:57 PM)
Fascinating interpretation there, thanks for sharing. There is one note of caution I'd like to sound in relation to small details of these images - it may be wrong and if so I hope Nirgal will correct me. He himself pointed out that albedo differences could affect his results here.
Correct. that's why I usually add my "for visualization purposes only" disclaimer
without knowledge of the true albedo variations, it's difficult to interpret the apparent height variations.
QUOTE
in relation to small details of these images.
actually it's quite the opposite: the smaller the detail, the better the accuracy of the 3D reconstruction (look at how the small craters and mini dune ripples got reproduced) because albedo variations occur more at larger spatial scales.
That's why the photoclinometric method can nicely complement the mult-image stereo-based methods that are less accurate at the very finest scale (you need features spanning areas of multiple pixels for the point matching algorithm) but are not so easily tricked by albedo variations.
So as mentinoed before: the real potential of the method for scientific usage could only be exploited by combing it with stereo/altimeter based data.
That being said: it is surprising how well the single-image reconstruction often works even in presence of albedo differences and cast shadows (see for example the dark dust steaks over the fine dune ripples in the image posted at the other thread) However, while often working pretty well, there is unfortunately no guarantee so (I have also encountered examples where the albedo problem messed up the 3D reconstruction completely )
I'm sure we will have an official stereo-based HiRISE DEM at 1m/pixel reslution of this area soon.
Then I can try to combine the 1m/pixel stereo-DEM it with my 25cm/pixel photoclinometric DEM to possibly get the best out of both worlds ...
But until then I reiterate my note of caution with respect to scientific/geologic interpretations based on the single-image DEMs that I want to be understood mainly as a means of visualization rtaher than a measurement tool ...
However, it's totally ok and a pleasure for me to see that the visualizations help to spark new and crystalize previously existing ideas... But for such ideas to be finally proven, I think we will need the real ground truth data in the end.