[hugin-ptx] Re: hugin creating random panoramas
On Sat 08-Nov-2008 at 18:21 -0500, Robert Krawitz wrote: > >My experience has been that using automatic control point generation >for typical outdoor panoramas is all but useless -- it picks a fair >number of seemingly random control points in the sky and clouds that >at best aren't very well aligned and at worst are completely incorrect >(from the wrong images). The 'celeste' functionality in the current hugin trunk should be very good at removing these 'sky' control points. Another technique that works with bracketed images is to only set control points on the most exposed set, where the sky is blown-out. This is what the match-n-shift --stacks option does when it finds a bracketed series. >I'd rather see an assistant (or at least an option) with an interface >more like Fotoxx -- drag and drop the images into rough alignment. >This might speed up control point generation, since it would have to >search much smaller regions of the images. This is really a manual workaround for problems in current image matching systems, the best solution is for control point generation to become much faster and much more accurate. -- Bruno --~--~-~--~~~---~--~~ 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 hugin-ptx@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/hugin-ptx -~--~~~~--~~--~--~---
[hugin-ptx] Re: hugin creating random panoramas
On Sun 09-Nov-2008 at 12:59 +0100, Erik Krause wrote: > >This is not very feasible if there are lots of images and it doesn't >provide a convenient way to identify unrelated images either. The >thumbnails would be far too small. The idea of using an undirected graph to navigate the project is attractive, but with a spherical panorama somewhere the graph has to turn inside out, so it isn't exactly clear: http://www.flickr.com/photos/[EMAIL PROTECTED]/232199/ >I'd prefer the global CP's table would highlight (put a border around) >the respective images in pano preview instead of loading those images in >CP pane (which takes ages on my computer, same as loading and scaling >images for pano preview BTW). Hugin loads and caches images when first used, so if you open a project and only use tabs that don't display images then it is very fast. Clicking on the Crop tab will load and cache a single image to display, which will take a noticeable amount of time. If you know you are going to need to see all the images in a project, you may as well just open the preview and wait for every image to get cached, after that the rest of hugin is quite fast. >A feature to pre-arrange images in a regular pattern would be nice as >well (not to speak of a possibility to arrange them manually WYSIWYG)... Dragging photos around manually is already possible in the OpenGL preview, it also has a feature where you can click on the overlap area between a pair of images and this pair is brought up in the Control Points tab. -- Bruno --~--~-~--~~~---~--~~ 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 hugin-ptx@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/hugin-ptx -~--~~~~--~~--~--~---
[hugin-ptx] Re: hugin creating random panoramas
Bart.van.Andel wrote: > > > I'd prefer the global CP's table would highlight (put a border around) > > > the respective images in pano preview instead of loading those images in > > > CP pane (which takes ages on my computer, same as loading and scaling > > > images for pano preview BTW). > > Not agreed. Even though I'd really like to see image highlighting in > the pano preview, the CP table now provides easy access to CPs with a > large distance, so these can directly be edited in the CP pane. Or am > I misunderstanding you? Providing it as an option (with a checkbox in > the CP window for instance) would be quite nice though. If the pano preview is open it could be used, else not. If the CP editor is visible it could be used, else not. A checkbox would be good, too: Show in CP editor and Show in pano preview. > >> > > This is only true of course if the images are taken in order, and it > >> > > doesn't apply to multi-row panoramas either. > > > > > > This should be a rather unusual case. You easily miss an image if you > > > shoot out of order. However, I often shoot a panorama in order and then > > > additional images. But those images must be treated specially anyway. > > How would I miss an image? The danger is big you forget to shoot one. Been there, done that ;-) > I don't see > why additional images would need different treatment, in general. If I shoot a regular panorama and then some images at arbitrary angles I usually do that because there are certain features, people or other moving things. Most of the time I need to set manual CPs for that images, since no control point generator will know what parts of the image are moving and what are static... -- Erik Krause http://www.erik-krause.de --~--~-~--~~~---~--~~ 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 hugin-ptx@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/hugin-ptx -~--~~~~--~~--~--~---
[hugin-ptx] Re: hugin creating random panoramas
Erik Krause wrote: > Bart.van.Andel wrote: > > Especially with a large > > number of input images. It could be as simple as found in M. Brown and > > Lowe's paper on "Automatic Panoramic Image Stitching using Invariant > > Features" (http://www.cs.ubc.ca/~mbrown/papers/ijcv2007.pdf, top of > > page 5). > > This is not very feasible if there are lots of images and it doesn't > provide a convenient way to identify unrelated images either. The > thumbnails would be far too small. Of course the thumbs are too small in this example, but since our screen space may be expected to be somewhat bigger than in the paper, larger thumbs can be used. Or simply a list of images on a canvas, with matched images shown to the right of each list item, maybe also showing the number of matched CPs and their average distance, for instance. Like this (img# represents a visual image): img0: img1, img5 img1: img0, img2 img2: img1, img3, img5 img3: img2, img4 img4: img3, img5 img5: img0, img2, img4 Such a list (or the graph view I mentioned earlier) could also be used to trigger the control point to find correspondences between images not linked together yet (for instance when an image is added later). Not sure if this will be used a lot though... > I'd prefer the global CP's table would highlight (put a border around) > the respective images in pano preview instead of loading those images in > CP pane (which takes ages on my computer, same as loading and scaling > images for pano preview BTW). Not agreed. Even though I'd really like to see image highlighting in the pano preview, the CP table now provides easy access to CPs with a large distance, so these can directly be edited in the CP pane. Or am I misunderstanding you? Providing it as an option (with a checkbox in the CP window for instance) would be quite nice though. > > This is only true of course if the images are taken in order, and it > > doesn't apply to multi-row panoramas either. > > This should be a rather unusual case. You easily miss an image if you > shoot out of order. However, I often shoot a panorama in order and then > additional images. But those images must be treated specially anyway. How would I miss an image? I'm adding them to Hugin which finds the CPs very nicely, no matter what order the images are fed. I don't see why additional images would need different treatment, in general. Again, maybe I misunderstood your point? Best, Bart --~--~-~--~~~---~--~~ 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 hugin-ptx@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/hugin-ptx -~--~~~~--~~--~--~---
[hugin-ptx] Re: hugin creating random panoramas
Bart.van.Andel wrote: > Especially with a large > number of input images. It could be as simple as found in M. Brown and > Lowe's paper on "Automatic Panoramic Image Stitching using Invariant > Features" (http://www.cs.ubc.ca/~mbrown/papers/ijcv2007.pdf, top of > page 5). This is not very feasible if there are lots of images and it doesn't provide a convenient way to identify unrelated images either. The thumbnails would be far too small. I'd prefer the global CP's table would highlight (put a border around) the respective images in pano preview instead of loading those images in CP pane (which takes ages on my computer, same as loading and scaling images for pano preview BTW). A feature to pre-arrange images in a regular pattern would be nice as well (not to speak of a possibility to arrange them manually WYSIWYG)... > This is only true of course if the images are taken in order, and it > doesn't apply to multi-row panoramas either. This should be a rather unusual case. You easily miss an image if you shoot out of order. However, I often shoot a panorama in order and then additional images. But those images must be treated specially anyway. -- Erik Krause http://www.erik-krause.de --~--~-~--~~~---~--~~ 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 hugin-ptx@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/hugin-ptx -~--~~~~--~~--~--~---
[hugin-ptx] Re: hugin creating random panoramas
On Nov 9, 12:00 am, "Erik Krause" <[EMAIL PROTECTED]> wrote: > Am Saturday, November 08, 2008 um 14:10 schrieb Bart.van.Andel: > > > After removing these wrong matches, the problem was > > solved. It took quite a lot of effort to remove all these CPs by hand > > though. > > The place to do that is the global control point table (F3). First > sort by right image, then sort by left image. Now scroll down and > compare image numbers. That is one possible way to deal with it, although maybe not the most intuitive one, having to sort twice first. Using a graph displaying thumbnails of the images and connections between them if Hugin found at least one matching CP, it should be far more easy to see the correspondences found by Hugin very quickly. Especially with a large number of input images. It could be as simple as found in M. Brown and Lowe's paper on "Automatic Panoramic Image Stitching using Invariant Features" (http://www.cs.ubc.ca/~mbrown/papers/ijcv2007.pdf, top of page 5). > compare image numbers. If it is a one row panorama any image n should > have control points to n-1 and n+1 only. If it is a 360° one the last > image should have points with the first one of course. This is only true of course if the images are taken in order, and it doesn't apply to multi-row panoramas either. > Ususally there are few wrong points only, hence you might spot them > in the "P CP" (private control point number) column, too. Ah, that's what "P CP" stands for. I assume that "G CP" then stands for global control point? Guess this is one of these examples where tool tips come in handy :) --~--~-~--~~~---~--~~ 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 hugin-ptx@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/hugin-ptx -~--~~~~--~~--~--~---
[hugin-ptx] Re: hugin creating random panoramas
Date: Sat, 08 Nov 2008 23:56:49 +0100 From: "J. Schneider" <[EMAIL PROTECTED]> Bart.van.Andel schrieb: >> most of the time hugin forms the panoramas >> totally wrong and on the preview i get all kinds of shapes and circles >> and diagonals > ... > I've had a problem recently which kind of sounds like what you're > trying to explain, I think. In my case, there were very nice control > points between the right images, but also some between images which > aren't adjacent at all. This caused Hugin to screw up the alignment > completely. After removing these wrong matches, the problem was > solved. It took quite a lot of effort to remove all these CPs by hand > though. This is one possibility. Another one is that the hfov is not exact enough. If you entered 38mm this might be just rounded. So let hugin optimize for hfov, maybe it works then. I've had this case several times. My experience has been that using automatic control point generation for typical outdoor panoramas is all but useless -- it picks a fair number of seemingly random control points in the sky and clouds that at best aren't very well aligned and at worst are completely incorrect (from the wrong images). So I wind up having to go through the table and remove a significant fraction of the control points. I don't know what features it's seeing -- maybe sensor noise or something that just happens to match up in a few places? I'd rather see an assistant (or at least an option) with an interface more like Fotoxx -- drag and drop the images into rough alignment. This might speed up control point generation, since it would have to search much smaller regions of the images. Fotoxx itself has a lot of limitations; you have to place the images left to right (so it can really only do strip panoramas), and it doesn't offer nearly as much control as Hugin. But the interface makes it a lot easier to get started. -- Robert Krawitz <[EMAIL PROTECTED]> Tall Clubs International -- http://www.tall.org/ or 1-888-IM-TALL-2 Member of the League for Programming Freedom -- mail [EMAIL PROTECTED] Project lead for Gutenprint --http://gimp-print.sourceforge.net "Linux doesn't dictate how I work, I dictate how Linux works." --Eric Crampton --~--~-~--~~~---~--~~ 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 hugin-ptx@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/hugin-ptx -~--~~~~--~~--~--~---
[hugin-ptx] Re: hugin creating random panoramas
Am Saturday, November 08, 2008 um 14:10 schrieb Bart.van.Andel: > After removing these wrong matches, the problem was > solved. It took quite a lot of effort to remove all these CPs by hand > though. The place to do that is the global control point table (F3). First sort by right image, then sort by left image. Now scroll down and compare image numbers. If it is a one row panorama any image n should have control points to n-1 and n+1 only. If it is a 360° one the last image should have points with the first one of course. Ususally there are few wrong points only, hence you might spot them in the "P CP" (private control point number) column, too. best regards -- Erik Krause Offenburger Str. 33 79108 Freiburg --~--~-~--~~~---~--~~ 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 hugin-ptx@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/hugin-ptx -~--~~~~--~~--~--~---
[hugin-ptx] Re: hugin creating random panoramas
Bart.van.Andel schrieb: >> most of the time hugin forms the panoramas >> totally wrong and on the preview i get all kinds of shapes and circles >> and diagonals > ... > I've had a problem recently which kind of sounds like what you're > trying to explain, I think. In my case, there were very nice control > points between the right images, but also some between images which > aren't adjacent at all. This caused Hugin to screw up the alignment > completely. After removing these wrong matches, the problem was > solved. It took quite a lot of effort to remove all these CPs by hand > though. This is one possibility. Another one is that the hfov is not exact enough. If you entered 38mm this might be just rounded. So let hugin optimize for hfov, maybe it works then. I've had this case several times. regards Joachim --~--~-~--~~~---~--~~ 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 hugin-ptx@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/hugin-ptx -~--~~~~--~~--~--~---
[hugin-ptx] Re: hugin creating random panoramas
> most of the time hugin forms the panoramas > totally wrong and on the preview i get all kinds of shapes and circles > and diagonals Could you explain this a bit more, by showing a screenshot for example? I've had a problem recently which kind of sounds like what you're trying to explain, I think. In my case, there were very nice control points between the right images, but also some between images which aren't adjacent at all. This caused Hugin to screw up the alignment completely. After removing these wrong matches, the problem was solved. It took quite a lot of effort to remove all these CPs by hand though. Feature idea: a graph showing which images are matched to which could be a nice idea, with an option to remove "edges" in the graph, to speed up the process of removing such wrong image match CPs quite a bit. --~--~-~--~~~---~--~~ 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 hugin-ptx@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/hugin-ptx -~--~~~~--~~--~--~---
