[X] My Assistant

[Phil Stooke]

I have posted several polar projections of MER panoramas (and Apollo and Surveyor pans in other places). This is how I do it...

I use Photoshop... any recent version is OK.

The tool to use is Filter>Distort>polar coordinates.

polar coordinates can work both ways - rectangular coordinates to circular or back. In map terms it lets you go from a cylindrical projection to an azimuthal (polar) one... in fact I have frequently used this with maps (for instance, take the cylindrical map of Titan from the JPL Cassini website, crop the southern half, and turn it into a pole-centered view of the southern hemisphere...)

For these pans you do the following.

1. 'polar coordinates' turns a rectangle into an ellipse. If you want a circle you have to make the rectangle a square. So I begin by resampling the pan to a square shape.. a 5000 by 1000 pixel pan becomes a 3000 by 3000 square. I cut off any unwanted sky first.

2. 'polar coordinates' puts the top of the rectangle at the center of the circle. So rotate the pan 180 degrees.

3. Then press the button and see what happens! You get a pan with the same horizon foreshortening as the original but wrapped into a circle.

4. I prefer a more maplike geometry, so the foreshortened distant areas need to be expanded relative to the foreground. I find it easier to do it this way:
- in step 1, I begin by making the 5000 by 1000 pan tall and narrow, maybe 3000 (wide) by 8000 (high). Then I select everythig except the near-horizon area and use edit>transform>scale (in my version, layer>transform>scale in others) to shrink the height of the selection about 10% or 15%. I keep repeating that with progressively smaller selections until at the end I'm only selecting a narrow strip of foreground and shrinking that. The result: the pan is now stretched higher near the horizon, and compressed near the foreground. Try it - trial and error will get decent results quite soon. NOTE: you could do this with some calculated figures to get more accurate geometry, but I just guess.

- then that modified pan is adjusted to be 3000 by 3000 pixels... usually with a bit of white space included at the bottom.

- then on to step 3 and see how it looks.

Let's keep them coming!

Phil

[Phil Stooke]

Dilo, very nice polar projection. I have never tried using PovRay, but this is very interesting. I look forward to seeing more of your work with this.

I have a daydream of making polar projections from several nearby locations and fitting them together to make a larger map. The only place I have done it so far (using JPL versions) is the heatshield, where it worked very well. I will post it later. But it would be interesting to try it for the inside of Endurance or Methuselah where we have a lot of pans all close together. Maybe we could work together on this.

Phil

[dilo]

Hi Phil, I'm happy you like it!
I have to make an important correction about the image geometry: this isn't a polar projection (where elevation angle should increase radially from center); this is a "vertical projection", where I used original image to texture a flat terrain using "spherical" map tool...
Herebelow you find a slightly improved version (on the right, with center detail on the left/top) and also a low-elevation angle version (left/bottom):

I was thinking to use PovRay (or other 3D rendering program) for a while because is a very powerful instrument... do you remember Erwann which posted some time ago some nice mosaics stitched with this SW? I'm convined that using this approach (project images as textures on a 3D model) can have really great potential, but implementation is not straightforward
... in theory, this should allow to:
1) enhanced stitching, without parallax effect due to small motion of Nav/Pan camera betrween two adjacent frames (some parallax remain on the closest objects, without detailed 3D model).
2) exact re-projection over soil, but only if one knows the exact 3D structure of this soil or if terrain is quite flat (like in my Opportunity image).

In both applications, it would be very useful to know exact pointing orientation of each image... I know this is possible, at least from more tecnical sources (like MER analyst book) but I do not know exactly how...

There are also other issues, related to non rigorous stitch obtained through AutoStitch (this is the reason for the imperfections visible in my projection, especially if compared to the official one...). With almost-standars settings, this sowtware makes also undesired rescaling of some images in the mosaic, so projection geometry is not very reliable .
Probably, I need to manually set some options inside this program (I didn't find a true tutorial, however); best thing, anyway, should be to directly stitch/project with PovRay (knowing mentioned pointing values)!

[djellison]

Hi Phil, I'm happy you like it!
I have to make an important correction about the image geometry: this isn't a polar projection (where elevation angle should increase radially from center); this is a "vertical projection", where I used original image to texture a flat terrain using "spherical" map tool...

I've not been able to figure that out myself - but I need a favour - could you take this mosaic - http://marswatch.astro.cornell.edu/pancam_...b_al_khali.html - and project it onto a flat surface like you have with other imager, and render it out as quite a large jpg?

I used 3ds max myself - but can not figure out how to do that

Doug

[dilo]

I've not been able to figure that out myself - but I need a favour - could you take this mosaic - http://marswatch.astro.cornell.edu/pancam_...b_al_khali.html  - and project it onto a flat surface like you have with other imager, and render it out as quite a large jpg?

I used 3ds max myself - but can not figure out how to do that

Doug

Doug, I'm very sorry but only now I discovered this post!
I prepared vertical projection you requested (hoping you're still interested to have it!) and results appear nice to me...

Some importand notes are need:
1) due to huge original panorama dimension, after download I had to reduce it to 60% of original size ("only" 13668 pixel wide); otherwise, PovRay is not able to handle it (at least, with 512 MB ram of my PC).
2) projection is not perfectly horizontal but is slightly inclined in order to flatten as much as possible the horizon (correction is less than 1 degree); in fact, if you look to original mosaic, horizon line follow a almost-sinusoidal profile.
3) In order to have more realistic result, I projected on a sphere with the Mars radius instead of an infinite plane.
4) No contrast/sharpness correction was made, even if images should greatly improve with these cosmetic operations... is up to you!

I didn't know the scale you are interested, so I made four versions at different scales (1,3,10,25 cm/pixel, distance ranging from 10m to 500m) plus a pseudo-polar projection.
Due to images size, I needed to attach results in two consecutive posts (I used low-compression jpeg format, png original versions are more than 2MB each).
Enjoy!

Attached thumbnail(s)

 

[dilo]

remaining attachments:

Attached thumbnail(s)