Panorama Stitching, Rick Szeliski's Work Is the Best Options

[Guest_DonPMitchell_*]

I noticed the thread on Photosynth and thought I would add a little more about that research group. I know Rick Szeliski and Matt Uyttendaele, very smart people.

I've tried pretty much all of the panorama stitchers out there. Adobe's is by far the worst, and Microsoft's is by far the best. Szeliski's algorithms are currently found in Microsoft Image Suite 2006. Unlike most stitchers, it is based on a robust treatment of the general camera projective transformations, so it can handle rotations, panning, zooming and even limited camera-position changes (modulo big visiblity changes of course).

[attachment=7042:attachment]

Here's an example that first impressed me. In a Soviet documentary film I own, they discuss this diagram of the Venera soil-drilling apparatus, but they never show the whole diagram. They pan around and zoom in and out, and I didn't expect to be able to resolve all that into one image. I was quite surprised when Rick's program did it.

An updated version of the stitcher is in a new product, Microsoft Graphic Designer, which will give users more control over a big class of transformations (affine, homography, 3D rotation, etc). It handles lens distortion and a variety of projection surfaces. I haven't tried it myself yet. link

I have Panorama Factory and Autostitcher of course, but I never use them now, they are nowhere near as powerful as DIS2006.

Uyttendaele also recommended a program to me called ptlens for some non-projective transformations like vignetting.

[edstrick]

The very simple fact is that the Cassini Imaging team is NOT doing it's job getting the public what it's paying for. We are not getting good periapsis-pass previews (That's the Project's PIO fault, mostly), we're getting dribs-and-drabs of 1 megapixel small press release images.

Then we get the dog-in-a-manger growls from Carolin being porky about us raiding her (I've more then once gotten the impression that it's grudgingly) posted "raw" data. The latest annoyance is the block on ftp style batch downloads from the cassini raw pages that started last summer. They may carp all they want about mass downloaders hogging the bandwidth, but we don't hear that whining from the MER people or the Exploratorium. I'm afraid it's simply an attempt to make it much more work for amateurs to do something substantially better than the team is doing but with substantially worse starting data.

What the stupididity is, is that this hampers all of the many un-paid promoters of planetary exploration out here, giving talks or one-on-one pushing space at public events like last months World Science Fiction Convention in LA (I was on 7 panels, and though none were slide-show or video presentations, I came loaded with piles of pictures, including Enceladus mosaics better than anything the team has doled out to the underserving public.

I'd better stop ranting before I start calling people things.....