Re: [hugin-ptx] Re: Vertical line detector
On Wed, 2 Nov 2011 22:57:08 -0700 (PDT), JohnPW wrote: If I don't misunderstand your post . . . Yes, one wants a level horizon, but the horizon is an imaginary line and is only evident in certain perfect situations without obstructions (flat desert, seascape, etc.) In most natural situations like landscapes, the viewer's senses are not greatly offended by a slightly misaligned horizon and it can estimated and adjusted without difficulty. The manmade environment, where most pictures are taken, is positively chock full of obviously vertical and horizontal lines. When all these cues do not align properly it is can be offensively obvious and discordant, so you want them to be properly aligned. in a rectilinear projection, only a vertical line always appears vertical and straight (like the longitude lines on a globe.) The only horizontal line that is level and straight is the horizon (which really isn't a line at all since it's theoretical projection and is actually a huge circle. All the truly straight, non vertical lines actually appear slightly curved, to a greater or lesser extent. The only horizontal lines that appear close to straight are on objects that directly face the viewer, and only those that happen to be coincident with the horizon, and these only over a short distance. (Sorry if I have misunderstood and I'm stating the obvious to you.) I ask because I took a panorama from a tower in Provincetown, MA, at the tip of Cape Cod. About 3/4 of the horizon from that spot is the ocean, and a misalignment of 1 pixel was very apparent; I had to correct it by editing the final output. I suppose the proposed new Apple Inc. campus building (which will be huge and round with a round courtyard in the middle) would offer an opportunity to use horizontals for alignment. :-) http://theinspirationroom.com/daily/design/2011/8/apple_city_rendering_1.jpg On Nov 2, 9:18 pm, Robert Krawitz r...@alum.mit.edu wrote: On Wed, 2 Nov 2011 20:04:03 +, Bruno Postle wrote: On Wed 02-Nov-2011 at 11:47 +0500, Emad ud din Bhatt wrote: can we use vertical line detector by Setting equirect pitch to 90 degree and than call vertical line detector. Than set pano pitch to -90 and call vertical line detector again? Yes (or roll), but you would have to manually change all the 'vertical control points' to 'horizontal'. This doesn't solve the fundamental problem: vertical lines are parallel, but horizontal lines generally are not - This is why horizontal control points have very limited use for levelling panoramas. I don't understand this -- usually one wants a level horizon? -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
[hugin-ptx] Re: Vertical line detector
Well that's one of those easy situations. When you can see the horizon clearly like that, horizontal CPs distributed about the pano on the horizon should give you great results. Contrary to how one might casually think, placing them far apart (with very wide angle images) produces diminishing accuracy. As they get closer to 180º apart the effect of minor errors in point placement is amplified, just as it would be by placing them closer to 0º apart. Was this at Long Point Light or were you up on a communications tower? On Nov 3, 7:15 am, Robert Krawitz r...@alum.mit.edu wrote: I ask because I took a panorama from a tower in Provincetown, MA, at the tip of Cape Cod. About 3/4 of the horizon from that spot is the ocean, and a misalignment of 1 pixel was very apparent; I had to correct it by editing the final output. -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
[hugin-ptx] Re: Vertical line detector
BTW my sig would have to be more like Unix doesn't dictate how I work, it dictates how I don't work (but only when I consciously try to use it.) ;-) I love that it's there, but I like having that Aqua interface softening the ride! -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
Re: [hugin-ptx] Re: Vertical line detector
On Thu, 3 Nov 2011 07:06:01 -0700 (PDT), JohnPW wrote: Well that's one of those easy situations. When you can see the horizon clearly like that, horizontal CPs distributed about the pano on the horizon should give you great results. Contrary to how one might casually think, placing them far apart (with very wide angle images) produces diminishing accuracy. As they get closer to 180º apart the effect of minor errors in point placement is amplified, just as it would be by placing them closer to 0º apart. One would think it would be simple, but it wasn't. There was one error of about 1 pixel I was never able to get rid of, and I had to play some games in GIMP to clean it up to my satisfaction. Even 1 pixel error is very noticeable for something like the sky-sae interface. Was this at Long Point Light or were you up on a communications tower? Provincetown Monument. See http://rlk.smugmug.com/Other/Landscapes/4851912_XB4SmT/1079379016_sm6Jy (the monument itself is http://rlk.smugmug.com/Travel/Provincetown-MA-October-2010/14616061_32XQRG/1087222681_A2kfU). -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
[hugin-ptx] Re: Vertical line detector
Ah the Pilgrim monument. I haven't been there. Very nice panoramas. They all look very well done to me. You clearly have very high production standards. When I was working for the NPS I did a similar style panorama from the top of the Cape Lookout Light (not even remotely as nice as yours though.) It was my first 360ş panorama and I wasn't sure it would work since I took it from the outer catwalk, but I figured the closest objects were far enough away that it wouldn't be a problem. It wasn't that great (3.5Mp jpegs, no fusion or TCA correction etc., but this was a few years ago and my colleagues (who had never seen stitched panos before) were amazed and perplexed by how I got such nice images using a little digital point and shoot (although the results would probably not be acceptable to anyone in this forum.) I stitched them together with well aged Apple QTVR Studio Software running on system 7 via Rosetta on a G3 PowerMac. Although the whole western sky was blown out, I was actually pleasantly surprised myself. I should go back and run it through Hugin. I'm sure there would be a little noticeable improvement. On Nov 3, 9:23 am, Robert Krawitz r...@alum.mit.edu wrote: On Thu, 3 Nov 2011 07:06:01 -0700 (PDT), JohnPW wrote: Well that's one of those easy situations. When you can see the horizon clearly like that, horizontal CPs distributed about the pano on the horizon should give you great results. Contrary to how one might casually think, placing them far apart (with very wide angle images) produces diminishing accuracy. As they get closer to 180ş apart the effect of minor errors in point placement is amplified, just as it would be by placing them closer to 0ş apart. One would think it would be simple, but it wasn't. There was one error of about 1 pixel I was never able to get rid of, and I had to play some games in GIMP to clean it up to my satisfaction. Even 1 pixel error is very noticeable for something like the sky-sae interface. Was this at Long Point Light or were you up on a communications tower? Provincetown Monument. Seehttp://rlk.smugmug.com/Other/Landscapes/4851912_XB4SmT/1079379016_sm6Jy (the monument itself ishttp://rlk.smugmug.com/Travel/Provincetown-MA-October-2010/14616061_3...). -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
Re: [hugin-ptx] Re: Vertical line detector
On Thu, 3 Nov 2011 09:35:04 -0700 (PDT), JohnPW wrote: Ah the Pilgrim monument. I haven't been there. Very nice panoramas. They all look very well done to me. You clearly have very high production standards. Thanks! When I was working for the NPS I did a similar style panorama from the top of the Cape Lookout Light (not even remotely as nice as yours though.) It was my first 360ş panorama and I wasn't sure it would work since I took it from the outer catwalk, but I figured the closest objects were far enough away that it wouldn't be a problem. It wasn't that great (3.5Mp jpegs, no fusion or TCA correction etc., but this was a few years ago and my colleagues (who had never seen stitched panos before) were amazed and perplexed by how I got such nice images using a little digital point and shoot (although the results would probably not be acceptable to anyone in this forum.) I stitched them together with well aged Apple QTVR Studio Software running on system 7 via Rosetta on a G3 PowerMac. Although the whole western sky was blown out, I was actually pleasantly surprised myself. I should go back and run it through Hugin. I'm sure there would be a little noticeable improvement. I had the same problem on the Pilgrim monument. There are only four spots, at the center of each side, where there's a clear view without glass and bars getting in the way. It's fortunate that I have an 8-16 mm lens; I don't think even a 10 or 11 mm lens would have provided enough overlap for a good stitch and a 12 mm lens probably wouldn't have been wide enough, period. I did have to fix some things up by hand where the parallax error was too great (the parking lot at the bottom had some problems that I had to fix manually, in addition to the horizon problem I mentioned earlier). I actually generally do use JPEGs, and I haven't done TCA correction or predefined lens models (which are likely to be accurate only at one particular focal length, anyway). And all too often I do them hand-held. But when I look at my panoramas, I generally don't see a lot of TCA problems. As for RAW vs. JPEG, the 7D does a very good job of in-camera processing. If the light's such that I'm going to have serious dynamic range problems, I probably need more than the additional one or two stops I'll get from my own RAW processing. The P-town panorama, for example, did have dynamic range problems, but simple exposure bracketing and fusion worked very well. For this one: http://rlk.smugmug.com/Other/Landscapes/4851912_XB4SmT/1488875261_xzm3Fzn I really did have to use RAW, though (and fix up a lot of sky by hand, also). There is one little trick I sometimes play that I haven't seen mentioned anywhere to reduce the aspect ratio and get more foreground detail. With wide angle lenses, the final output is somewhat torpedo or barrel shaped due to the projection onto a planar surface. I make a second pass with Hugin, treating the first stage panorama as having been shot by a cylindrical lens (like a Spinshot camera) of between 20 and 35 mm focal length and then re-projecting it as rectilinear, which applies a pincushion effect. Panorama stitching really is a lot of fun. -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
[hugin-ptx] Re: Vertical line detector
So how do folks do their hand adjustments? . . . [I guess I should start a new topic for this.] On Nov 3, 12:47 pm, Robert Krawitz r...@alum.mit.edu wrote: I had the same problem on the Pilgrim monument. There are only four spots, at the center of each side, where there's a clear view without glass and bars getting in the way. It's fortunate that I have an 8-16 mm lens; I don't think even a 10 or 11 mm lens would have provided enough overlap for a good stitch and a 12 mm lens probably wouldn't have been wide enough, period. I did have to fix some things up by hand where the parallax error was too great (the parking lot at the bottom had some problems that I had to fix manually, in addition to the horizon problem I mentioned earlier). I actually generally do use JPEGs, and I haven't done TCA correction or predefined lens models (which are likely to be accurate only at one particular focal length, anyway). And all too often I do them hand-held. But when I look at my panoramas, I generally don't see a lot of TCA problems. As for RAW vs. JPEG, the 7D does a very good job of in-camera processing. If the light's such that I'm going to have serious dynamic range problems, I probably need more than the additional one or two stops I'll get from my own RAW processing. The P-town panorama, for example, did have dynamic range problems, but simple exposure bracketing and fusion worked very well. -- I have no choice but to use jpeg on my Nikon CP4500 (Tiff isn't really a reasonable choice time wise and the camera has no RAW.) For this one: http://rlk.smugmug.com/Other/Landscapes/4851912_XB4SmT/1488875261_xzm... I really did have to use RAW, though (and fix up a lot of sky by hand, also). There is one little trick I sometimes play that I haven't seen mentioned anywhere to reduce the aspect ratio and get more foreground detail. With wide angle lenses, the final output is somewhat torpedo or barrel shaped due to the projection onto a planar surface. I make a second pass with Hugin, treating the first stage panorama as having been shot by a cylindrical lens (like a Spinshot camera) of between 20 and 35 mm focal length and then re-projecting it as rectilinear, which applies a pincushion effect. -- I'd have to see to understand (my fault not yours.) Perhaps you should make a tutorial? :-) -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
Re: [hugin-ptx] Re: Vertical line detector
can we use vertical line detector by Setting equirect pitch to 90 degree and than call vertical line detector. Than set pano pitch to -90 and call vertical line detector again? On Fri, Oct 28, 2011 at 7:09 PM, Battle battlebr...@gmail.com wrote: I'll reiterate Thomas' comments. I do a lot of buildings. I would like horizontal and vertical line detectors. Seems like it ought to be a simple programming loop that would reuse existing vertical line feature search on a 90 degree rotated basis calling them horizontal instead of vertical. I realize I'm over simplifying things, but its not a new invention, but an adaption of an existing one. I probably don't have the programming skills to help, but I would like to see the extension of the vertical line find to horizontal as 99 of what I do is rectilinear projections of architecture. Battle On Oct 27, 8:49 am, Jeffrey Martin 360cit...@gmail.com wrote: I'm bringing this thread up again... Does anyone find the possibility of a horizon detector interesting, for cases where there are not vertical lines, but there is a horizon? This would help to automatically level more panos Jeffrey -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx -- *Emaad* www.flickr.com/emaad -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
Re: [hugin-ptx] Re: Vertical line detector
On Wed 02-Nov-2011 at 11:47 +0500, Emad ud din Bhatt wrote: can we use vertical line detector by Setting equirect pitch to 90 degree and than call vertical line detector. Than set pano pitch to -90 and call vertical line detector again? Yes (or roll), but you would have to manually change all the 'vertical control points' to 'horizontal'. This doesn't solve the fundamental problem: vertical lines are parallel, but horizontal lines generally are not - This is why horizontal control points have very limited use for levelling panoramas. The case of straightening an elevation of a building is an exception. For this someone can write a switch for linefind that enables detection of horizontal lines in panoramas with FoV 180°, but it still shouldn't be the default in Hugin. -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
Re: [hugin-ptx] Re: Vertical line detector
Greetings all, I'm new here. This is all very interesting to me. My project at the moment is to photograph the school where I work, parts of which can be seen here: http://www.i-magz.com/thomas-more.html In front of the school is a strip of grass, then a narrow road, then a thick row of trees. Photography involves a lot of tilting the camera at very steep angles! My best effort so far was to move along the building taking photos in portrait format, 1 low and 1 high in each location. Hugin does avery good job but gets defeated by the serious curves on the verticals. It also gets very confused by the parts of the building that are invisible on one shot but visible on the next. I would be very grateful for any comments, suggestions or criticisms. Thank you Ian On Wed, Nov 2, 2011 at 8:04 PM, Bruno Postle br...@postle.net wrote: On Wed 02-Nov-2011 at 11:47 +0500, Emad ud din Bhatt wrote: can we use vertical line detector by Setting equirect pitch to 90 degree and than call vertical line detector. Than set pano pitch to -90 and call vertical line detector again? Yes (or roll), but you would have to manually change all the 'vertical control points' to 'horizontal'. This doesn't solve the fundamental problem: vertical lines are parallel, but horizontal lines generally are not - This is why horizontal control points have very limited use for levelling panoramas. The case of straightening an elevation of a building is an exception. For this someone can write a switch for linefind that enables detection of horizontal lines in panoramas with FoV 180°, but it still shouldn't be the default in Hugin. -- 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_FAQhttp://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 hugin-ptx+unsubscribe@** googlegroups.com hugin-ptx%2bunsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/** group/hugin-ptx http://groups.google.com/group/hugin-ptx -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
[hugin-ptx] Re: Vertical line detector
If I don't misunderstand your post . . . Yes, one wants a level horizon, but the horizon is an imaginary line and is only evident in certain perfect situations without obstructions (flat desert, seascape, etc.) In most natural situations like landscapes, the viewer's senses are not greatly offended by a slightly misaligned horizon and it can estimated and adjusted without difficulty. The manmade environment, where most pictures are taken, is positively chock full of obviously vertical and horizontal lines. When all these cues do not align properly it is can be offensively obvious and discordant, so you want them to be properly aligned. in a rectilinear projection, only a vertical line always appears vertical and straight (like the longitude lines on a globe.) The only horizontal line that is level and straight is the horizon (which really isn't a line at all since it's theoretical projection and is actually a huge circle. All the truly straight, non vertical lines actually appear slightly curved, to a greater or lesser extent. The only horizontal lines that appear close to straight are on objects that directly face the viewer, and only those that happen to be coincident with the horizon, and these only over a short distance. (Sorry if I have misunderstood and I'm stating the obvious to you.) I suppose the proposed new Apple Inc. campus building (which will be huge and round with a round courtyard in the middle) would offer an opportunity to use horizontals for alignment. :-) http://theinspirationroom.com/daily/design/2011/8/apple_city_rendering_1.jpg On Nov 2, 9:18 pm, Robert Krawitz r...@alum.mit.edu wrote: On Wed, 2 Nov 2011 20:04:03 +, Bruno Postle wrote: On Wed 02-Nov-2011 at 11:47 +0500, Emad ud din Bhatt wrote: can we use vertical line detector by Setting equirect pitch to 90 degree and than call vertical line detector. Than set pano pitch to -90 and call vertical line detector again? Yes (or roll), but you would have to manually change all the 'vertical control points' to 'horizontal'. This doesn't solve the fundamental problem: vertical lines are parallel, but horizontal lines generally are not - This is why horizontal control points have very limited use for levelling panoramas. I don't understand this -- usually one wants a level horizon? -- Robert Krawitz r...@alum.mit.edu Tall Clubs International -- http://www.tall.org/or 1-888-IM-TALL-2 Member of the League for Programming Freedom -- http://ProgFree.org 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
[hugin-ptx] Re: Vertical line detector
I'll reiterate Thomas' comments. I do a lot of buildings. I would like horizontal and vertical line detectors. Seems like it ought to be a simple programming loop that would reuse existing vertical line feature search on a 90 degree rotated basis calling them horizontal instead of vertical. I realize I'm over simplifying things, but its not a new invention, but an adaption of an existing one. I probably don't have the programming skills to help, but I would like to see the extension of the vertical line find to horizontal as 99 of what I do is rectilinear projections of architecture. Battle On Oct 27, 8:49 am, Jeffrey Martin 360cit...@gmail.com wrote: I'm bringing this thread up again... Does anyone find the possibility of a horizon detector interesting, for cases where there are not vertical lines, but there is a horizon? This would help to automatically level more panos Jeffrey -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
Re: [hugin-ptx] Re: Vertical line detector
I'm bringing this thread up again... Does anyone find the possibility of a horizon detector interesting, for cases where there are not vertical lines, but there is a horizon? This would help to automatically level more panos Jeffrey -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
[hugin-ptx] Re: Vertical line detector
Still, there is certainly use for a *horizon* detector. in many panos there is water - the ocean - and no vertical lines. here, the only way to level the pano is to set this horizon line. this is not an edge case - this is a really big case :) if you check some panos on 360cities you'll see what i mean. additionally, these panos are really hard for anyone to level, unless they deeply understand stitching software i didn't have a chance to try this vertical line detector yet but i'm very intrigued :) i'll try it soon! Thanks Thomas for helping to make it a reality :) :) Jeffrey -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
Re: [hugin-ptx] Re: Vertical line detector
Hi Thomas, 2011/9/17 T. Modes thomas.mo...@gmx.de It's an edge detector. The same as in calibrate_lens_gui. If it would consider only parallel lines, it would not handles perspective distortion (stürzende Linien in German) correctly. (A view from a non-programming user): Isn't it possible to make linefind a two-step edge detector. First search for vertical lines, then rotate the algorithm itself and search for horizontal lines. I see that you have vertical lines functions and structures inside findlines.cpp. Could it be done with some coordination changes in the algorithms to make them search for horizontal lines? Harry -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
[hugin-ptx] Re: Vertical line detector
Hi (A view from a non-programming user): Isn't it possible to make linefind a two-step edge detector. First search for vertical lines, then rotate the algorithm itself and search for horizontal lines. I see that you have vertical lines functions and structures inside findlines.cpp. Could it be done with some coordination changes in the algorithms to make them search for horizontal lines? That is possible. But the question is: for which application this feature is necessary? For generating a normal panorama the detection of horizontal lines would only help if there is a defined horizon or the projection is rectilinear. For all other projection there is maximal one horizontal line - the horizon - adding horizontal lines will result in the case often in distortions. I can imagine, that it helps for correcting perspective distortions, but this is also limited to the rectilinear projection. Thomas -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
[hugin-ptx] Re: Vertical line detector
On 18 Sep., 10:51, T. Modes thomas.mo...@gmx.de wrote: Hi (A view from a non-programming user): Isn't it possible to make linefind a two-step edge detector. First search for vertical lines, then rotate the algorithm itself and search for horizontal lines. I see that you have vertical lines functions and structures inside findlines.cpp. Could it be done with some coordination changes in the algorithms to make them search for horizontal lines? That is possible. But the question is: for which application this feature is necessary? Just one example: I've done panoramas of dams, and the only really reliable lines were along the banisters. When I manually set horizontal CPs on them, they made my pano perfectly level. The few verticals I had were short, few and unreliable (lampposts). And while we're at it - why not support lines which are neither horizontal nor vertical? There is a CP type for them, and they can be very helpful - plus, they can contain more than just two image points, which makes them ideal for nudging bits from several images to a common line (I've used them to good effect when stitching mosaics of scans of large images). So if the line CPG detects a very long straight edge, it might output a line cp with points every few hundred pixels. All of this is supposing, of course, that the detector is taking into account the input's projection - e.g. in my dam case it's stereographic, so only lines through the center are straight, just where they're worst- suited for levelling the panorama (if I remember correctly, it's best to have two vertical line CPs near the left or right margin of an image) For generating a normal panorama the detection of horizontal lines would only help if there is a defined horizon or the projection is rectilinear. For all other projection there is maximal one horizontal line - the horizon - adding horizontal lines will result in the case often in distortions. This implies linefind is working on the source images, instead of reprojected source images. This wouldn't be good news for fisheye users. If I understand hugin's ways right, vericality/horizontality is refering to the points when transformed to equirect (so, the optimization is done in pano space) - and in equirect all horizontal lines come out horizontal (to be more precise, I'm talking about lines which in a perfect equirect or rectilinear or cylindrical projection would appear horizontal, not merely lines which are horizontal in the scene but may not project to horizontal because of perspective), no matter where they are, so every horizontal can be helpful Kay -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
[hugin-ptx] Re: Vertical line detector
On 18 Sep., 11:16, kfj _...@yahoo.com wrote: and in equirect all horizontal lines come out horizontal (to be more precise, I'm talking about lines which in a perfect equirect or rectilinear or cylindrical projection would appear horizontal, not merely lines which are horizontal in the scene but may not project to horizontal because of perspective) ... oops - tahat's nonesense. You're right, Thomas, horizontal remain horizontal only in rectilinear, or otherwise on the horizon itself. Kay -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
Re: [hugin-ptx] Re: Vertical line detector
2011/9/18 T. Modes thomas.mo...@gmx.de Hi (A view from a non-programming user): Isn't it possible to make linefind a two-step edge detector. First search for vertical lines, then rotate the algorithm itself and search for horizontal lines. I see that you have vertical lines functions and structures inside findlines.cpp. Could it be done with some coordination changes in the algorithms to make them search for horizontal lines? That is possible. But the question is: for which application this feature is necessary? For generating a normal panorama the detection of horizontal lines would only help if there is a defined horizon or the projection is rectilinear. For all other projection there is maximal one horizontal line - the horizon - adding horizontal lines will result in the case often in distortions. I can imagine, that it helps for correcting perspective distortions, but this is also limited to the rectilinear projection. Thomas You are right. I was just thinking further along the line of what is possible. However, in case you really want to correct perspective distortions in buildings or so, the image is such that you want to do it by hand anyway (at least most of the time). Harry -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
[hugin-ptx] Re: Vertical line detector
Hi Kay, Sorry to keep pestering you about it, but if no images are selected in the images tab and you click to use linefind, it looks at no images - rather than at all images, which is the default behaviour with all other CPGs if no images are selectd. So maybe you can modify your hack yet again. Fixed again. As far as linefind itself goes, I have mixed results with it. I am currently scanning some maps, and they have black border lines and a separate black framing box, not to mention the UTM grid in thin red lines. I ran linefeed on the whole set of 15 A4 300 dpi scans, and it found 8 vertical lines in total, all of which were on the black frame, and usually very short (few tens of pixels) - it found none of the red grid lines. Is the detector really a line detector, or more one for vertical edges, like two planes bordering on each other vertically? It's an edge detector. The same as in calibrate_lens_gui. If it would consider only parallel lines, it would not handles perspective distortion (stürzende Linien in German) correctly. And is there a possibility to also have it scan for horizontal lines? That would be great for my scans. It reports only vertical lines. But there is a workaround: Run linefind to find vertical lines. Then change the roll value of all images to 90 (or -90) and run linefind again. This will find the horizontal lines. Thomas -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
Re: [hugin-ptx] Re: Vertical line detector
T. Modes wrote: Hi Kay, Sorry to keep pestering you about it, but if no images are selected in the images tab and you click to use linefind, it looks at no images - rather than at all images, which is the default behaviour with all other CPGs if no images are selectd. So maybe you can modify your hack yet again. Fixed again. As far as linefind itself goes, I have mixed results with it. I am currently scanning some maps, and they have black border lines and a separate black framing box, not to mention the UTM grid in thin red lines. I ran linefeed on the whole set of 15 A4 300 dpi scans, and it found 8 vertical lines in total, all of which were on the black frame, and usually very short (few tens of pixels) - it found none of the red grid lines. Is the detector really a line detector, or more one for vertical edges, like two planes bordering on each other vertically? It's an edge detector. The same as in calibrate_lens_gui. If it would consider only parallel lines, it would not handles perspective distortion (stürzende Linien in German) correctly. And is there a possibility to also have it scan for horizontal lines? That would be great for my scans. It reports only vertical lines. But there is a workaround: Run linefind to find vertical lines. Then change the roll value of all images to 90 (or -90) and run linefind again. This will find the horizontal lines. Then wouldn't that leave the set of horizontal lines identified as vertical lines? -- Gnome Nomad gnomeno...@gmail.com wandering the landscape of god http://www.cafepress.com/otherend/ -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
[hugin-ptx] Re: Vertical line detector
On 17 Sep., 19:10, Gnome Nomad gnomeno...@gmail.com wrote: It reports only vertical lines. But there is a workaround: Run linefind to find vertical lines. Then change the roll value of all images to 90 (or -90) and run linefind again. This will find the horizontal lines. Then wouldn't that leave the set of horizontal lines identified as vertical lines? Well, yes. But if I have to go to the length of rotating the images 90 degrees (easy in hsi), I might as well do that before I scan for verticals and run a simple find-and-replace on the CP type field in the pto after the horizontals are found. It's even quite feasible to write a little python script as a go-between - see my cpg_glue script at http://bazaar.launchpad.net/~kfj/+junk/script/view/head:/main/cpg_glue.py which does the two linefind calls etc. Kay -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
Re: [hugin-ptx] Re: Vertical line detector
Hi Thomas, I have observed now that vertical linefind is an integral part of cpfind. I don't like that very much as it actually disturbs some panorama's. I think it should be an option, be it a default option.But is should be possible to do without, something like --noverticallinefind (or a better name). I have quite some panoramas of landscapes containing trees. I suppose some edge like algorithm is used and with trees that are vertical and quite straight, but not absolutely vertical and not completely straight I get horrible distortions. This is very obvious in the Assistant. It requires me to recenter, resize to fit, restraighten and to autocrop again. So please make this a (default) option that can be switched off. Harry -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
[hugin-ptx] Re: Vertical line detector
Hi Harry, I have observed now that vertical linefind is an integral part of cpfind. That's wrong. Linefind is an own program and not integrated into cpfind. I don't like that very much as it actually disturbs some panorama's. I think it should be an option, be it a default option.But is should be possible to do without, something like --noverticallinefind (or a better name). Linefind is integrated into the assistant workflow. I made this step optional, you can switch it off in the preferences. Thomas -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
Re: [hugin-ptx] Re: Vertical line detector
2011/9/15 Harry van der Wolf hvdw...@gmail.com 2011/9/15 T. Modes thomas.mo...@gmx.de Hi Harry, I have observed now that vertical linefind is an integral part of cpfind. That's wrong. Linefind is an own program and not integrated into cpfind. So I thought but it seemed not from the output. I don't like that very much as it actually disturbs some panorama's. I think it should be an option, be it a default option.But is should be possible to do without, something like --noverticallinefind (or a better name). Linefind is integrated into the assistant workflow. I made this step optional, you can switch it off in the preferences. Thomas Where is it in Preferences? I can't find it. Harry OK, I see. A new commit to the trunk. I will sync and build it. Harry -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
Re: [hugin-ptx] Re: Vertical line detector
2011/9/15 T. Modes thomas.mo...@gmx.de Hi Harry, I have observed now that vertical linefind is an integral part of cpfind. That's wrong. Linefind is an own program and not integrated into cpfind. So I thought but it seemed not from the output. I don't like that very much as it actually disturbs some panorama's. I think it should be an option, be it a default option.But is should be possible to do without, something like --noverticallinefind (or a better name). Linefind is integrated into the assistant workflow. I made this step optional, you can switch it off in the preferences. Thomas Where is it in Preferences? I can't find it. Harry -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
Re: [hugin-ptx] Re: Vertical line detector
2011/9/15 Harry van der Wolf hvdw...@gmail.com 2011/9/15 Harry van der Wolf hvdw...@gmail.com 2011/9/15 T. Modes thomas.mo...@gmx.de Hi Harry, I have observed now that vertical linefind is an integral part of cpfind. That's wrong. Linefind is an own program and not integrated into cpfind. So I thought but it seemed not from the output. I don't like that very much as it actually disturbs some panorama's. I think it should be an option, be it a default option.But is should be possible to do without, something like --noverticallinefind (or a better name). Linefind is integrated into the assistant workflow. I made this step optional, you can switch it off in the preferences. Thomas Where is it in Preferences? I can't find it. Harry OK, I see. A new commit to the trunk. I will sync and build it. Harry Works OK. thanks. Harry -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
Re: [hugin-ptx] Re: Vertical line detector
Can I have a copy of your built OS X one too, if you can put it up online somewhere? Thanks. On 13 September 2011 20:06, Harry van der Wolf hvdw...@gmail.com wrote: Hi Thomas, Just built the 5551. Did 2 panos and linefind works fine on OSX as well. Will do some more tests. Harry -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx -- *Ian Tindale* -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
Re: [hugin-ptx] Re: Vertical line detector
2011/9/14 Ian Tindale ian.tind...@gmail.com Can I have a copy of your built OS X one too, if you can put it up online somewhere? Thanks. Tonight I will put it on my website as usual. Harry -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
[hugin-ptx] Re: Vertical line detector
On 11 Sep., 19:41, T. Modes thomas.mo...@gmx.de wrote: Hi group, the default branch contains now a vertical line detector. It tries to generate vertical control points for levelling of the pano. I have another slight issue with linefind. When I use it from hugin, sometimes I want it to look only at a single image. If I activate just one image in the images tab and run any CPG, the CPG is fed all images instead of just the single one. With most CPGs it is of course futile to only look at a single image, because, after all, one wants CPS between several images - the only reason to apply a CPG to a single image would be for side-effects, like to generate a key file. With vertical line CPs, it's perfectly sensible to just process one image from the set, though. So my issue is with hugins's interaction with this specific CPG rather than with the CPG itself. I wonder if this wouldn't be easy to fix for someone who's inside the code? Kay -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
[hugin-ptx] Re: Vertical line detector
Hi Kay, So my issue is with hugins's interaction with this specific CPG rather than with the CPG itself. I wonder if this wouldn't be easy to fix for someone who's inside the code? added special handling for linefind to work with single image. (It's more like a hack, it works only for cpg which have linefind in the program name) Thomas -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
[hugin-ptx] Re: Vertical line detector
On 13 Sep., 19:43, T. Modes thomas.mo...@gmx.de wrote: Hi Kay, So my issue is with hugins's interaction with this specific CPG rather than with the CPG itself. I wonder if this wouldn't be easy to fix for someone who's inside the code? added special handling for linefind to work with single image. (It's more like a hack, it works only for cpg which have linefind in the program name) I suppose new CPGs needing this special treatment won't suddenly turn up out of the blue in large numbers, so a hack should be perfectly adequate ;) It works here. Thanks! Kay -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
Re: [hugin-ptx] Re: Vertical line detector
Hi Thomas, Just built the 5551. Did 2 panos and linefind works fine on OSX as well. Will do some more tests. Harry -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
[hugin-ptx] Re: Vertical line detector
On 11 Sep., 19:41, T. Modes thomas.mo...@gmx.de wrote: Hi group, the default branch contains now a vertical line detector. It tries to generate vertical control points for levelling of the pano. ... Compiles, integrates and runs here: Betriebssystem: Linux 2.6.38-11-generic-tuxonice i686 Architektur: 32 bit Freier Speicher: -1828108 kiB Hugin Version: Pre-Release 2011.3.0.4826832f8d18 Ressourcen-Pfad: /usr/local/share/hugin/xrc/ Datenpfad: /usr/local/share/hugin/data/ Libraries wxWidgets: 2.8.11.0 ... but the result seems strange to me - maybe there's an error? I fed it a pto with nine images from my stereographic fisheye. It found thre vertical lines in three different images: (command line output, same if run from hugin) $ linefind -o result.pto input.pto linefind is searching for vertical lines Working on image IMG_1050.tif Found 0 vertical lines Working on image IMG_1051.tif Found 1 vertical lines Working on image IMG_1052.tif Found 0 vertical lines Working on image IMG_1053.tif Found 1 vertical lines Working on image IMG_1054.tif Found 0 vertical lines Working on image IMG_1055.tif Found 0 vertical lines Working on image IMG_1056.tif Found 0 vertical lines Working on image IMG_1057.tif Found 1 vertical lines Working on image IMG_1058.tif Found 0 vertical lines Written output to result.pto but in the resulting pto file, result.pto, all three vertical CPs are assigned to image number zero: # control points c n0 N0 x2847.22729509834 y2096.36641460524 X2851.72865032147 Y2405.76855173728 t1 c n0 N0 x-2.47460425806207 y2128.94518372973 X-3.35545983791303 Y2407.96021470891 t1 c n0 N0 x2851.06479346968 y1874.90595527298 X2843.63811550857 Y2193.76486309095 t1 The use of negative coordinates also seems a bit strange. Am I missing something? Kay -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
[hugin-ptx] Re: Vertical line detector
@Kay but in the resulting pto file, result.pto, all three vertical CPs are assigned to image number zero: # control points c n0 N0 x2847.22729509834 y2096.36641460524 X2851.72865032147 Y2405.76855173728 t1 c n0 N0 x-2.47460425806207 y2128.94518372973 X-3.35545983791303 Y2407.96021470891 t1 c n0 N0 x2851.06479346968 y1874.90595527298 X2843.63811550857 Y2193.76486309095 t1 The use of negative coordinates also seems a bit strange. Both issues are fixed in default branch. @brian_ims Kornel is on the right way. Maybe autooptimiser is still running. Check taskmanager and kill autooptimiser.exe or reboot your system. Thomas -- 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 hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx