The Guys at PANCAM know what they are doing!
-------------------------------------------------------

Last winter I thought about super-res (SR) and read many articles on-line.
The technique rests on mapping several images onto a finer background to allow sub-pixel registration.
I hit upon an idea to enlarge an image to a finer grid and and build a statistical method to reduce registration error (similar to root mean square score).

I chose Opportunities backshell and the results were very good.
I moved onto Spirit's heatshield and gained an appreciation for the Professionals! (the Pancam Team)

My SR heatshield is good -- as good as my SR backshell -- Team Pancam's SR heatshield is MUCH BETTER
( This indicates to me that for whatever reason, they did not publish their best SR opportunuity backshell image, but rather, some over processed version. )

Spirit's heatshield by Team Pancam is very large ( zoomed 10x the original) mine is 3x

Team Pancam's image is very crisp (sharp, focused) Mine is softer and cant be blown up to 10x and stay crisp

Registration is one step in super-res and there are many ways to get there; my technique is very fast but relies only on luminance. Other methods are more complex - and the results show.
Post processing after registration (or even during) is the what makes the final image crisp. Exactly knowing the Pancam's PSF numbers is important. (I dont do PSF correction). Deblur is very complex and unintuitive.

Now I know where my my leisure-time research will concentrate on over the next few months!
(one of the nice things about people working on a method for over 15 years is that they leave a lot of articles and papers on the internet!)

Here are the images:

* Team pancam's is first and very crisp for such a large image (mine gets softer as I zoom it out - This is what separates The Men from the Boys)
* Mine is second ( I made it lighter than Team Pancam's to bring out some contrast)

(pancams original image is here -

http://marswatch.astro.cornell.edu/pancam_...uperres.IMG.tif

I converted it to a png below)


Cheers !


Click to view attachmentClick to view attachment

Good work!

There was a thread about this long, long ago when several of us worked with those images. Here are my results. The first is a comparison with the Cornell version (left) and mine (right). The second is my color version.