Examining the pano, it appears that hugin began flipping the photos on photo img1566.jpg. This is about the sixth image, and that image is blurred s small bit. From this image on, all the images were flipped. For some reason the blurred photo was flipped, and then hugin flipped the rest of the images.
I pulled the blurred image out of the set, and loaded hugin again. This time hugin did not flip any of the images, but the pano is "longer" than the view. Essentially it's not guessing the FOV correctly on the "Auto Assistant" tab. I assume it's guessing and not pulling the info from the EXIF data. I am using ImageExif Tool. I did zoom in to look at the control points on the image, and the control points are dead on. I assume at this point that hugin does not look at just the control points but considers image rotation in relation to it's stitching? Hugin knows how to find control points on the blurry photo, but not what rotation it is? It seems the blurry photo confused hugin on it's rotation. I've attached a link with the ziped files and the *.pto http://www.tatteredmoons.org/20090902-003_pano.zip From: [email protected] To: [email protected] Subject: [hugin-ptx] Re: Flipping Photos ! Date: Tue, 1 Sep 2009 22:42:55 -0500 I do not believe one can look make the observation that an image is "improperly flipped" by examining the control points between only one pair of images. There could easily be other control points between the flipped image and other images that is driving the optimized orientation to be flipped. Consider that the control points shown are tightly clustered. The error distance between these points in the flipped position is small compared the distance between other control points when the image is not flipped if those points are indeed wider apart and such that the rotated image matches them up. You just might have some zinger control points. They can be hard to spot. Judging from that tight cluster of points, which looks like a typical autopano-sift-c matching, I would not be surprised if you relied entirely upon automatic matching, which can and will come up with some surprisingly exact matches that are total zingers. You have to check every point combination. Use the point table or the green bar matching indicator to spot which images have control points between them that you know are impossible. I believe you when you report progressive strangeness. I have seen similar but I have also seen Hugin suddenly return to rock solid at times when I was sure it was spiraling out of control. Nevertheless, I would start this pano over from scratch and use manual match point, saving the project at every step. BTW, I prefer pan-o-matic over autopano-sift-c because it results in wider control point spreads. Allan On Sep 1, 2009, at 1:42 PM, Dale Beams wrote:Canon camera, no change of FOV. The problem is that it worked on the first pano, errors on the second and then progressively got worse for each succeeding pano. Only upon closing hugin and then restarting it was I able to correct some of the pano's. However, when this happens, it flips the photos every which way, some 90, some 180, some 2**, etc. It's a bizarre thing. I'm on an AMD processor with a nvidia chipset. > Date: Tue, 1 Sep 2009 08:02:04 -0700 > Subject: [hugin-ptx] Re: Flipping Photos ! > From: [email protected] > To: [email protected] > > > Could this be a problem with a 360 degree pano where the initial FOV > guess is completely wrong? If so, you might circumvent this issue by > resetting all parameters and entering a more appropriate FOV guess > manually. > > I've also had this problem on occasion, usually with photos lacking > EXIF or having "incorrect" EXIF (I've experimented with holding a > separate lens in front of my el cheapo digital compact camera, which > of course invalidates the focal information of the EXIF). > > The existence of control point between photos which shouldn't be > connected can cause similar issues, I've seen this as well. > > -- > Bart > > On 1 sep, 13:26, RizThon <[email protected]> wrote: > > Hi, > > I've also had weird behaviors like this, where one or more of my pictures > > would be rotated by 90, 180 or 270°, even if the control points are all > > correct. Looking at Dale's Control Points picture, which reminds me of what > > I've experienced a few times, I definitely think there's a bug somewhere > > because how could hugin possibly think that the image 6 should be rotated > > 180° while all the control points are correct? > > This happened to me while trying to stitch panoramas not of landscapes, > > where even if you don't have a pano head it's not really an issue, but of > > closer things, which is just what Dale is trying to do...but still, why > > would hugin rotate that image... > _________________________________________________________________ With Windows Live, you can organize, edit, and share your photos. http://www.windowslive.com/Desktop/PhotoGallery --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "hugin and other free panoramic software" group. A list of frequently asked questions is available at: http://wiki.panotools.org/Hugin_FAQ To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] For more options, visit this group at http://groups.google.com/group/hugin-ptx -~----------~----~----~----~------~----~------~--~---
