Lennart Mollerstrom
Posts: 97
Location: Stockholm, Sweden,
Registered: 17 Dec 2006
|
Match-n-shift
Posted: 30 May 2008 at 19:30 GMT
|
|
Has anyone else tried Match-n-shift? I searched for it here in the forum but didn't find a thing about it. I've been using it now since almost two months to generate control points between fish-eye images. I think it works great. I have hardly picked a single point manually since I started to use it. Plain Autopano otherwise only picks points close to the equator and I used to have to add lots of points closer to zenith and nadir to get a good stich. I use Pentax 10-17 mm fish-eye @ 10 mm.
Lennart
|
|
alert moderator
|
|
John Houghton
Posts: 2317
Location: Hitchin, United Kingdom
Registered: 17 Jan 2005
|
Re: Match-n-shift
Posted: 31 May 2008 at 6:51 GMT
|
|
Lennart, Perhaps others might be tempted to try it if you could point them in the right direction - e.g. download link and documentation on what it is and how it is used in conjunction with stitchers on different platforms. I found tracking it down somewhat frustrating. I located a recent Hugin development Windows binary that might contain it but that refused to install.
John
|
|
alert moderator
|
|
Smooth
Posts: 1472
Location: Australia
Registered: 21 Jul 2004
|
|
John Houghton
Posts: 2317
Location: Hitchin, United Kingdom
Registered: 17 Jan 2005
|
Re: Match-n-shift
Posted: 31 May 2008 at 10:00 GMT
|
|
Smooth, Thanks.
John
|
|
alert moderator
|
|
Lennart Mollerstrom
Posts: 97
Location: Stockholm, Sweden,
Registered: 17 Dec 2006
|
Re: Match-n-shift
Posted: 31 May 2008 at 18:00 GMT
updated: 31 May 2008 at 18:01 GMT
|
|
Thanks Smooth for posting the link. I know it wasn't easy to find the exe-version though, so your help is appreciated.
I learnt about Match-n-shift on this link where it also said there was an exe version.
panospace.wordpress.com/
You would otherwise have to install Perl and a bunch of libraries. A bit too much even for me.
Match-n-shift is a utility that re-maps fish-eye images in a way that the Autopano algorithm can find matches between fish-eye images. Once it has found the control points it "re-maps" the control points back to fish-eye projection and outputs a project file. I wasn't able to integrate it to run from within Hugin so I made a batch-file instead that I'm using. I prefer not to post the batch-file since it has a major problem and that is that if I forget to edit the batch-file it will overwrite the previous project file. It is a much better idea to put some work in making it work from within Hugin. I suppose it could be called from within PTGui as well since it is a command line utility like Autopano. One thing that I found was that I had to set the Windows decimal separator in regional settings to point instead of comma to make it work.
Lennart
|
|
alert moderator
|
|
John Houghton
Posts: 2317
Location: Hitchin, United Kingdom
Registered: 17 Jan 2005
|
Re: Match-n-shift
Posted: 31 May 2008 at 20:59 GMT
|
|
It took me a couple of hours to finally get a project file out of match-n-shift. It generates a .pto file, which PTGui turns its nose up at, but I was able to copy/paste the control points from the .pto file into a PTGui .pts project file. I thought there might be problems with the wonderously precise coordinates (specified to no less than 12 decimal places), but PTGui coped and simply rounded them to the nearest integer. A real problem is that PTGui uses a dummy image for image 0. Consequently, the real images are numbered from image 1. However, the control point numbering in the .pto file reflects Hugin's numbering starting from image 0. I anticipated this and added a dummy image to the set of 8mm fisheye test images used as input to match-n-shift, so the control points then came out numbered correctly for PTGui.
In this very limited test, the control points were certainly more widely spread than is normally the case for the usual generators. The central area seemed comparatively neglected though, with bunching at the top and bottom. Having generated 15 points per seam, optimization revealed only one really bad point, and I deleted 3 or 4 others that were slightly substandard to get a really good optimization. It looks promising.
John
|
|
alert moderator
|
|
