Loading ...
Sorry, an error occurred while loading the content.

Parquet flooring drives me crazy!

Expand Messages
  • obarakana
    From time to time I attempt making stitched rectilinear images of architectural subjects, usually interiors. Often in circumstances where I might previously
    Message 1 of 18 , Jul 30, 2008
    • 0 Attachment
      From time to time I attempt making stitched rectilinear images of
      architectural subjects, usually interiors. Often in circumstances
      where I might previously have used a 5x4 inch view camera.
      As a general rule, I use a Canon 5D with a 17mm lens (17-40 mm at its
      widest setting) mounted on a Manfrotto 303 SPH head. Each time I try
      to set the nodal point correctly. I use Autopano Pro to stitch the images.
      One of my regular clients occupies historic buildings with parquet
      floors and I have immense difficulty getting a seamless stitching of
      the individual floor boards. Even if I save the panorama as a PSD
      file, along with the layers, it is virtually impossible to retouch the
      errors by painting in details from a sepatate layer.
      In the preview window of APP everything looks ok and it is only when
      checking the output in PS that I can see the errors.
      All suggestions would be gratefully received. All I really want to do
      is make relatively simple rectilinear images by stitching in
      situations where, even with a wide-angle lens, I can't cover my
      subject in one shot.
      I am considering all options: should I try using a compact camera such
      as a Sigma DP1, on the assumption that the shorter lens to sensor
      distance will make setting the nodal point simpler?, or maybe use a
      fisheye and straightening out the image(s) on the basis that the fewer
      the images, the less chance of poor joins?
      Thanks in advance!

      David.
    • Mark D. Fink
      Hi David, Do you have some sample images for us to look at? That might help us troubleshoot what is going on. If you have the proper NPP, then stitching errors
      Message 2 of 18 , Jul 30, 2008
      • 0 Attachment
        Hi David,

        Do you have some sample images for us to look at? That might help us
        troubleshoot what is going on. If you have the proper NPP, then stitching
        errors should be minimal, if at all. I don't think switching to a point and
        shoot camera will help any.

        Once you set your NPP, how repeatable is the Manfrotto 303 SPH head? Can you
        reliably get it back exactly where you had it before? If so, then how are
        you determining if you are at the proper NPP? I used this method, which
        worked great: <http://www.johnhpanos.com/epcalib.htm>.

        Mark
        www.pinnacle-vr.com
        www.nyc.360cities.net
        www.northernlight.net


        >-----Original Message-----
        >From: PanoToolsNG@yahoogroups.com [mailto:PanoToolsNG@yahoogroups.com] On
        >Behalf Of obarakana
        >Sent: Wednesday, July 30, 2008 7:34 AM
        >To: PanoToolsNG@yahoogroups.com
        >Subject: [PanoToolsNG] Parquet flooring drives me crazy!
        >
        >From time to time I attempt making stitched rectilinear images of
        >architectural subjects, usually interiors. Often in circumstances
        >where I might previously have used a 5x4 inch view camera.
        >As a general rule, I use a Canon 5D with a 17mm lens (17-40 mm at its
        >widest setting) mounted on a Manfrotto 303 SPH head. Each time I try
        >to set the nodal point correctly. I use Autopano Pro to stitch the images.
        >One of my regular clients occupies historic buildings with parquet
        >floors and I have immense difficulty getting a seamless stitching of
        >the individual floor boards. Even if I save the panorama as a PSD
        >file, along with the layers, it is virtually impossible to retouch the
        >errors by painting in details from a sepatate layer.
        >In the preview window of APP everything looks ok and it is only when
        >checking the output in PS that I can see the errors.
        >All suggestions would be gratefully received. All I really want to do
        >is make relatively simple rectilinear images by stitching in
        >situations where, even with a wide-angle lens, I can't cover my
        >subject in one shot.
        >I am considering all options: should I try using a compact camera such
        >as a Sigma DP1, on the assumption that the shorter lens to sensor
        >distance will make setting the nodal point simpler?, or maybe use a
        >fisheye and straightening out the image(s) on the basis that the fewer
        >the images, the less chance of poor joins?
        >Thanks in advance!
        >
        >David.
        >
        >
        >
        >------------------------------------
        >
        >--
        >
        >
        >
      • paul womack
        ... I seem to remember a previous thread which gave information on how precise the NPP needed to be, which is related to this issue. BugBear
        Message 3 of 18 , Jul 30, 2008
        • 0 Attachment
          Mark D. Fink wrote:
          > Hi David,
          >
          > Do you have some sample images for us to look at? That might help us
          > troubleshoot what is going on. If you have the proper NPP, then stitching
          > errors should be minimal, if at all. I don't think switching to a point and
          > shoot camera will help any.
          >
          > Once you set your NPP, how repeatable is the Manfrotto 303 SPH head? Can you
          > reliably get it back exactly where you had it before? If so, then how are
          > you determining if you are at the proper NPP? I used this method, which
          > worked great: <http://www.johnhpanos.com/epcalib.htm>.

          I seem to remember a previous thread which gave information
          on how precise the NPP needed to be, which is related to
          this issue.

          BugBear
        • Mark D. Fink
          ... Wouldn t that be a relative issue? The closer the object is to the lens, the more precise your NPP has to be. (Thankfully, otherwise we d have MUCH more
          Message 4 of 18 , Jul 30, 2008
          • 0 Attachment
            >-----Original Message-----
            >From: PanoToolsNG@yahoogroups.com [mailto:PanoToolsNG@yahoogroups.com] On
            >Behalf Of paul womack
            >Sent: Wednesday, July 30, 2008 8:05 AM
            >To: PanoToolsNG@yahoogroups.com
            >Subject: Re: [PanoToolsNG] Parquet flooring drives me crazy!
            >
            >Mark D. Fink wrote:
            >> Hi David,
            >>
            >> Do you have some sample images for us to look at? That might help us
            >> troubleshoot what is going on. If you have the proper NPP, then stitching
            >> errors should be minimal, if at all. I don't think switching to a point
            >and
            >> shoot camera will help any.
            >>
            >> Once you set your NPP, how repeatable is the Manfrotto 303 SPH head? Can
            >you
            >> reliably get it back exactly where you had it before? If so, then how are
            >> you determining if you are at the proper NPP? I used this method, which
            >> worked great: <http://www.johnhpanos.com/epcalib.htm>.
            >
            >I seem to remember a previous thread which gave information
            >on how precise the NPP needed to be, which is related to
            >this issue.
            >
            > BugBear

            Wouldn't that be a relative issue? The closer the object is to the lens, the
            more precise your NPP has to be. (Thankfully, otherwise we'd have MUCH more
            trouble with pole panos!)

            For me, equally important is being able to precisely return to a particular
            NPP once you've gone through the trouble of finding it. <ADVERT>That's why I
            designed the Pinnacle VR, to not only give me precise adjustments, but also
            be able to return reliably to those positions.</ADVERT>

            Mark
            www.pinnacle-vr.com
            www.nyc.360cities.net
            www.northernlight.net
          • paul womack
            ... Heh. It was me! But the answer is due to Erik Krause. http://wiki.panotools.org/Parallax You have a Canon 5D with a 17mm lens Since the 5D is full frame,
            Message 5 of 18 , Jul 30, 2008
            • 0 Attachment
              paul womack wrote:
              > Mark D. Fink wrote:
              >> Hi David,
              >>
              >> Do you have some sample images for us to look at? That might help us
              >> troubleshoot what is going on. If you have the proper NPP, then stitching
              >> errors should be minimal, if at all. I don't think switching to a point and
              >> shoot camera will help any.
              >>
              >> Once you set your NPP, how repeatable is the Manfrotto 303 SPH head? Can you
              >> reliably get it back exactly where you had it before? If so, then how are
              >> you determining if you are at the proper NPP? I used this method, which
              >> worked great: <http://www.johnhpanos.com/epcalib.htm>.
              >
              > I seem to remember a previous thread which gave information
              > on how precise the NPP needed to be, which is related to
              > this issue.

              Heh. It was me!

              But the answer is due to Erik Krause.

              http://wiki.panotools.org/Parallax

              You have a "Canon 5D with a 17mm lens"
              Since the 5D is full frame, that's a
              angle of view of

              http://en.wikipedia.org/wiki/Angle_of_view#Calculating_a_camera.27s_angle_of_view
              perl -e 'use Math::Trig; print rad2deg(2 * atan(36/(2 * 17)))'

              1.62 radians or 93 degrees.

              Let's assume you're being ambitious, and want to get
              down to 1m distance from the subject,
              and that you're doing a 4 shot panorama with 25% overlap,
              giving a 14000 pixel wide panorama (roughly 4368 * (.75 * 3 + 1) )

              Assuming further ambition, we'll target no more than 2 pixel
              stitching errors.

              Erik's page gives us pixel error from NPP error,
              but I want the opposite.

              from Erik:

              pe = 2 * B * panowidth / 360
              B = atan(r * sin(a)/d)

              pe = 2 * atan(r * sin(a)/d) * panowidth / 360
              pe = atan(r * sin(a)/d) * panowidth / 180
              180 * pe / panowidth = atan(r * sin(a) / d)
              tan(180 * pe / panowidth) = r * sin(a) / d
              r = (d / sin(a)) * tan(180 * pe / panowidth)

              in our example d is 1000mm, panowidth is 14000, a
              is (.75 * 93 / 2) = 35, and pe is 2

              gives: (*)

              .8 mm as the required accuracy of your NPP.

              BugBear

              *

              perl -e 'use Math::Trig; printf("%g\n", (1000/sin(deg2rad(35))) * tan(deg2rad(180) * 2 / 14000))'
            • Erik Krause
              ... If the NPP is ok even autopano pro should be able to do a good stitch. However, PTGui has advanced correction options like viewpoint correction. Would be
              Message 6 of 18 , Jul 30, 2008
              • 0 Attachment
                David Carr-2 wrote:
                >
                > As a general rule, I use a Canon 5D with a 17mm lens (17-40 mm at its
                > widest setting) mounted on a Manfrotto 303 SPH head. Each time I try
                > to set the nodal point correctly. I use Autopano Pro to stitch the images.
                > One of my regular clients occupies historic buildings with parquet
                > floors and I have immense difficulty getting a seamless stitching of
                > the individual floor boards.
                >

                If the NPP is ok even autopano pro should be able to do a good stitch.
                However, PTGui has advanced correction options like viewpoint correction.
                Would be interesting to have a try...

                best regards


                -----
                Erik Krause
                http://www.erik-krause.de
                --
                View this message in context: http://www.nabble.com/Parquet-flooring-drives-me-crazy%21-tp18731256p18732971.html
                Sent from the PanoToolsNG mailing list archive at Nabble.com.
              • paul womack
                ... I don t that page can be right; it implies that you can avoid parallax errors in a 360 pano taken with a wide angle lens by simply taking more shots. I
                Message 7 of 18 , Jul 30, 2008
                • 0 Attachment
                  paul womack wrote:
                  > But the answer is due to Erik Krause.
                  >
                  > http://wiki.panotools.org/Parallax

                  I don't that page can be right; it implies
                  that you can avoid parallax errors in a 360
                  pano taken with a wide angle lens
                  by simply taking more shots.

                  I don't see how this would avoid the errors
                  (although it clearly minimises the errors between
                  2 adjacent shots)

                  BugBear
                • Sacha Griffin
                  You have to make a determination on whether they are program stitching errors or physical parallax stitching errors. If you are not 100% confident in no
                  Message 8 of 18 , Jul 30, 2008
                  • 0 Attachment
                    You have to make a determination on whether they are program stitching
                    errors or physical parallax stitching errors.

                    If you are not 100% confident in no parallax, then it's almost certainly
                    parallax.



                    On the other side 17mm rectilinear is ultra high resolution, for a sphere or
                    immersive application. For a normal panoramic photo, stick with it.

                    Poor joins are corrected by either learning the NPP process, which doesn't
                    require that much investment in time, or stepping up to a full fledged
                    panorama tools gui like ptgui.

                    One problem I see for such a long lens for a sphere, is this process. You
                    don't have a whole lot of turn needed in order to see "Close" parallax
                    errors which you need to correct to shoot a sphere.



                    For example, I have a 50mm lens that can not focus close and doing the turn
                    trick is pointless.



                    "Turn trick"

                    Setup your camera, lens, head on your best npp point next to a table with
                    object(s) close to the lens on the right or left side of the frame.

                    Snap a photo, and rotate the camera so the object is still in frame but on
                    the other side.

                    Check "In a computer" not the lcd.

                    Notice the background in relation to the edges of the near object. That's
                    your parallax.



                    The problem with a longer lens, is you can not rotate them enough, and they
                    usually aren't able to focus on objects very close, making the process
                    difficult for mm precision.



                    Most likely you have a easily in process and easy in time parallax problem.
                    17mm rectilinear should be doable.

                    Otherwise, switch to another program where you can optimize lens distortions
                    and sensor shift.





                    Sacha Griffin

                    Southern Digital Solutions LLC

                    http://www.southern-digital.com

                    http://www.seeit360.net

                    404-551-4275











                    From: PanoToolsNG@yahoogroups.com [mailto:PanoToolsNG@yahoogroups.com] On
                    Behalf Of obarakana
                    Sent: Wednesday, July 30, 2008 7:34 AM
                    To: PanoToolsNG@yahoogroups.com
                    Subject: [PanoToolsNG] Parquet flooring drives me crazy!



                    From time to time I attempt making stitched rectilinear images of
                    architectural subjects, usually interiors. Often in circumstances
                    where I might previously have used a 5x4 inch view camera.
                    As a general rule, I use a Canon 5D with a 17mm lens (17-40 mm at its
                    widest setting) mounted on a Manfrotto 303 SPH head. Each time I try
                    to set the nodal point correctly. I use Autopano Pro to stitch the images.
                    One of my regular clients occupies historic buildings with parquet
                    floors and I have immense difficulty getting a seamless stitching of
                    the individual floor boards. Even if I save the panorama as a PSD
                    file, along with the layers, it is virtually impossible to retouch the
                    errors by painting in details from a sepatate layer.
                    In the preview window of APP everything looks ok and it is only when
                    checking the output in PS that I can see the errors.
                    All suggestions would be gratefully received. All I really want to do
                    is make relatively simple rectilinear images by stitching in
                    situations where, even with a wide-angle lens, I can't cover my
                    subject in one shot.
                    I am considering all options: should I try using a compact camera such
                    as a Sigma DP1, on the assumption that the shorter lens to sensor
                    distance will make setting the nodal point simpler?, or maybe use a
                    fisheye and straightening out the image(s) on the basis that the fewer
                    the images, the less chance of poor joins?
                    Thanks in advance!

                    David.





                    [Non-text portions of this message have been removed]
                  • Erik Krause
                    ... Not avoid but minimize. ... Well, it should be clearer perhaps. Relevant for the amount of parallax is the displacement of the NPP relative to the optical
                    Message 9 of 18 , Jul 30, 2008
                    • 0 Attachment
                      paul womack wrote:
                      >
                      > paul womack wrote:
                      >> But the answer is due to Erik Krause.
                      >>
                      >> http://wiki.panotools.org/Parallax
                      >
                      > I don't that page can be right; it implies
                      > that you can avoid parallax errors in a 360
                      > pano taken with a wide angle lens
                      > by simply taking more shots.
                      >

                      Not avoid but minimize.


                      paul womack wrote:
                      >
                      > I don't see how this would avoid the errors
                      >

                      Well, it should be clearer perhaps. Relevant for the amount of parallax is
                      the displacement of the NPP relative to the optical axis of the adjacent
                      shot. Since the distance to the rotation axis stays constant this
                      displacement depends on the rotation angle only. This is how the formula
                      works. Of course this is valid for adjacent images only.

                      And this is why you can shoot stereo panoramas (where the camera is heavily
                      displaced) without parallax errors if you take 120 shots around like in the
                      "Strip assembly.." thread <http://www.nabble.com/-to18682318.html>

                      best regards

                      -----
                      Erik Krause
                      http://www.erik-krause.de
                      --
                      View this message in context: http://www.nabble.com/Parquet-flooring-drives-me-crazy%21-tp18731256p18734689.html
                      Sent from the PanoToolsNG mailing list archive at Nabble.com.
                    • paul womack
                      ... And (implicitly) only use 3 degrees from each shot, which I think is the qualification missing from the wiki page. BugBear
                      Message 10 of 18 , Jul 30, 2008
                      • 0 Attachment
                        Erik Krause wrote:
                        >
                        > And this is why you can shoot stereo panoramas (where the camera is heavily
                        > displaced) without parallax errors if you take 120 shots around like in the
                        > "Strip assembly.." thread <http://www.nabble.com/-to18682318.html>

                        And (implicitly) only use 3 degrees from each shot, which I think
                        is the qualification missing from the wiki page.

                        BugBear
                      • obarakana
                        Thanks everyone! Sorry, I had just deleted the unsatisfactory photos so I had nothing to show... I suspect the problem is at least partly due to a lack of
                        Message 11 of 18 , Jul 30, 2008
                        • 0 Attachment
                          Thanks everyone!

                          Sorry, I had just deleted the unsatisfactory photos so I had nothing
                          to show...

                          I suspect the problem is at least partly due to a lack of precision
                          establishing the nodal point. The entrance pupil database looks to be
                          a very useful tool. I had just been lining up the nearest and farthest
                          objects in the viewfinder and checking that their relative positions
                          didn't (appear to) move when the camera was rotated. Apparently this
                          wasn't good enough! I will have to spend more time here to learn...
                        • Sacha Griffin
                          Exactly what I expected! Been there done that. J Just use the same method, but load them into photoshop this time and you should be set. Sacha Griffin Southern
                          Message 12 of 18 , Jul 30, 2008
                          • 0 Attachment
                            Exactly what I expected! Been there done that. J

                            Just use the same method, but load them into photoshop this time and you
                            should be set.





                            Sacha Griffin

                            Southern Digital Solutions LLC

                            http://www.southern-digital.com

                            http://www.seeit360.net

                            404-551-4275







                            From: PanoToolsNG@yahoogroups.com [mailto:PanoToolsNG@yahoogroups.com] On
                            Behalf Of obarakana
                            Sent: Wednesday, July 30, 2008 5:00 PM
                            To: PanoToolsNG@yahoogroups.com
                            Subject: [PanoToolsNG] Re: Parquet flooring drives me crazy!



                            Thanks everyone!

                            Sorry, I had just deleted the unsatisfactory photos so I had nothing
                            to show...

                            I suspect the problem is at least partly due to a lack of precision
                            establishing the nodal point. The entrance pupil database looks to be
                            a very useful tool. I had just been lining up the nearest and farthest
                            objects in the viewfinder and checking that their relative positions
                            didn't (appear to) move when the camera was rotated. Apparently this
                            wasn't good enough! I will have to spend more time here to learn...





                            [Non-text portions of this message have been removed]
                          • paul womack
                            ... It s good enough IF (and only if) you check for them not moving at pixel-for-pixel resolution. related question. Given that the maths for parallax appears
                            Message 13 of 18 , Jul 31, 2008
                            • 0 Attachment
                              obarakana wrote:
                              > Thanks everyone!
                              >
                              > Sorry, I had just deleted the unsatisfactory photos so I had nothing
                              > to show...
                              >
                              > I suspect the problem is at least partly due to a lack of precision
                              > establishing the nodal point. The entrance pupil database looks to be
                              > a very useful tool. I had just been lining up the nearest and farthest
                              > objects in the viewfinder and checking that their relative positions
                              > didn't (appear to) move when the camera was rotated. Apparently this
                              > wasn't good enough! I will have to spend more time here to learn...

                              It's good enough IF (and only if) you check for
                              them not moving at pixel-for-pixel resolution.

                              related question. Given that the maths
                              for parallax appears reasonably simple, is it possible
                              to take two angularly separated photographs,
                              and use the measured AMOUNT of parallax
                              error (and some maths) to make a calculated
                              correction to the NPP and get it right in one iteration,
                              and to high precision?

                              BugBear
                            • Erik Krause
                              ... No, that doesn t matter. As I wrote before the displacement of the NPP sideways from the adjacent shots optical axis is relevant. You can test it yourself:
                              Message 14 of 18 , Jul 31, 2008
                              • 0 Attachment
                                paul womack wrote:
                                >
                                > Erik Krause wrote:
                                >>
                                >> And this is why you can shoot stereo panoramas (where the camera is
                                >> heavily
                                >> displaced) without parallax errors if you take 120 shots around like in
                                >> the
                                >> "Strip assembly.." thread <http://www.nabble.com/-to18682318.html>
                                >
                                > And (implicitly) only use 3 degrees from each shot, which I think
                                > is the qualification missing from the wiki page.
                                >

                                No, that doesn't matter. As I wrote before the displacement of the NPP
                                sideways from the adjacent shots optical axis is relevant. You can test it
                                yourself: Adjust your pano head such that the camera doesn't rotate around
                                the NPP. Put some objects near the camera and have a structured background.
                                Then look in the viewfinder or on the display and rotate the camera slowly.
                                You see that the objects in the foreground move gradually and all the same
                                no matter whether they are in the center or near the image border.

                                If you use a fisheye and you look very thoroughly you might spot a
                                difference in moving speed from the center to the borders due to fisheye NPP
                                shift. In this case the distance between NPP and rotation axis changes. The
                                formula however stays valid, only that r and alpha change instead of only
                                alpha. Since the NPP is the center of perspective (the point from which the
                                camera sees) all boils down to simple geometry.

                                best regards


                                -----
                                Erik Krause
                                http://www.erik-krause.de
                                --
                                View this message in context: http://www.nabble.com/Parquet-flooring-drives-me-crazy%21-tp18731256p18751798.html
                                Sent from the PanoToolsNG mailing list archive at Nabble.com.
                              • Eric O'Brien
                                Since we re talking about identical areas of a FLOOR failing to stitch well, it doesn t seem that it would be a *parallax* error, right? Therefore, the
                                Message 15 of 18 , Aug 4, 2008
                                • 0 Attachment
                                  Since we're talking about identical areas of a FLOOR failing to
                                  stitch well, it doesn't seem that it would be a *parallax* error,
                                  right? Therefore, the problem probably is not an incorrectly placed
                                  "nodal point."

                                  It is more likely that the "warping coefficients" (or whatever we
                                  call them) are not sufficiently precise. I believe that AutopanoPro
                                  has some way to manually add control points. You might try to find a
                                  view where existing control points are shown, and manually add some
                                  toward the edges of your images.

                                  I'm assuming that APP will then use your added points and generate
                                  better lens characterization values. You might try asking for help
                                  on the APP forums.

                                  eo


                                  On Jul 30, 2008, at 2:00 PM, obarakana wrote:

                                  > Thanks everyone!
                                  >
                                  > Sorry, I had just deleted the unsatisfactory photos so I had nothing
                                  > to show...
                                  >
                                  > I suspect the problem is at least partly due to a lack of precision
                                  > establishing the nodal point. The entrance pupil database looks to be
                                  > a very useful tool. I had just been lining up the nearest and farthest
                                  > objects in the viewfinder and checking that their relative positions
                                  > didn't (appear to) move when the camera was rotated. Apparently this
                                  > wasn't good enough! I will have to spend more time here to learn...
                                  >
                                  > ------------------------------------
                                  >
                                • Sacha Griffin
                                  Floors are generally the closest place to the camera, and are most affected by parallax. Ceilings are generally featureless, and stitching errors won t be
                                  Message 16 of 18 , Aug 4, 2008
                                  • 0 Attachment
                                    Floors are generally the closest place to the camera, and are most affected
                                    by parallax. Ceilings are generally featureless, and stitching errors won't
                                    be noticeable.

                                    However, you are semi-right. Moderate parallax, will also throw off good
                                    lens optimization as the control points on images with parallax are now out
                                    place from where they should be.

                                    Obarakana, stated he used his viewfinder only to check npp placement, and
                                    didn't shoot and view the images on a computer.

                                    The problem with using a viewfinder only, on determining parallax from an
                                    test in which an image is close to the camera, and the background is
                                    notated, is that you can also move your eye in relation to the viewfinder,
                                    to compensate the apparent parallax error.

                                    Meaning you can move your eye from true straight on while looking through
                                    the viewfinder, without realizing it.. and correcting the true parallax of
                                    the npp placement of the lens and thus considering your npp optimization is
                                    complete.





                                    Sacha Griffin

                                    Southern Digital Solutions LLC

                                    http://www.southern-digital.com

                                    http://www.seeit360.net

                                    404-551-4275







                                    From: PanoToolsNG@yahoogroups.com [mailto:PanoToolsNG@yahoogroups.com] On
                                    Behalf Of Eric O'Brien
                                    Sent: Monday, August 04, 2008 4:40 AM
                                    To: PanoToolsNG@yahoogroups.com
                                    Subject: Re: [PanoToolsNG] Re: Parquet flooring drives me crazy!



                                    Since we're talking about identical areas of a FLOOR failing to
                                    stitch well, it doesn't seem that it would be a *parallax* error,
                                    right? Therefore, the problem probably is not an incorrectly placed
                                    "nodal point."

                                    It is more likely that the "warping coefficients" (or whatever we
                                    call them) are not sufficiently precise. I believe that AutopanoPro
                                    has some way to manually add control points. You might try to find a
                                    view where existing control points are shown, and manually add some
                                    toward the edges of your images.

                                    I'm assuming that APP will then use your added points and generate
                                    better lens characterization values. You might try asking for help
                                    on the APP forums.

                                    eo

                                    On Jul 30, 2008, at 2:00 PM, obarakana wrote:

                                    > Thanks everyone!
                                    >
                                    > Sorry, I had just deleted the unsatisfactory photos so I had nothing
                                    > to show...
                                    >
                                    > I suspect the problem is at least partly due to a lack of precision
                                    > establishing the nodal point. The entrance pupil database looks to be
                                    > a very useful tool. I had just been lining up the nearest and farthest
                                    > objects in the viewfinder and checking that their relative positions
                                    > didn't (appear to) move when the camera was rotated. Apparently this
                                    > wasn't good enough! I will have to spend more time here to learn...
                                    >
                                    > ------------------------------------
                                    >





                                    [Non-text portions of this message have been removed]
                                  • PanoToolsNG.10.m8@spamgourmet.com
                                    Another thing to watch for with repeating patterns(eg. parquet flooring), is that the auto(or manual) control points do point to the correct places. It is
                                    Message 17 of 18 , Aug 4, 2008
                                    • 0 Attachment
                                      Another thing to watch for with repeating patterns(eg. parquet flooring), is
                                      that the auto(or manual) control points "do" point to the correct places. It
                                      is very easy for the control points to be pointing to slightly different
                                      places in adjacent images.

                                      Cheers,
                                      Darren.

                                      )-----Original Message-----
                                      )From: Sacha Griffin - sachagriffin@...
                                      )[mailto:+panotoolsng+m8+7d2fc24864.sachagriffin#southern-digita
                                      l.com@...]
                                      )Sent: Monday, 4 August 2008 22:00
                                      )To: PanoToolsNG@yahoogroups.com
                                      )Subject: RE: [PanoToolsNG] Re: Parquet flooring drives me
                                      )crazy! (panotoolsng: sachagriffin@... exclusive)
                                      )
                                      )Floors are generally the closest place to the camera, and are
                                      )most affected by parallax. Ceilings are generally featureless,
                                      )and stitching errors won't be noticeable.
                                      )
                                      )However, you are semi-right. Moderate parallax, will also
                                      )throw off good lens optimization as the control points on
                                      )images with parallax are now out place from where they should be.
                                      )
                                      )Obarakana, stated he used his viewfinder only to check npp
                                      )placement, and didn't shoot and view the images on a computer.
                                      )
                                      )The problem with using a viewfinder only, on determining
                                      )parallax from an test in which an image is close to the
                                      )camera, and the background is notated, is that you can also
                                      )move your eye in relation to the viewfinder, to compensate the
                                      )apparent parallax error.
                                      )
                                      )Meaning you can move your eye from true straight on while
                                      )looking through the viewfinder, without realizing it.. and
                                      )correcting the true parallax of the npp placement of the lens
                                      )and thus considering your npp optimization is complete.
                                      )
                                      )
                                      )
                                      )
                                      )
                                      )Sacha Griffin
                                      )
                                      )Southern Digital Solutions LLC
                                      )
                                      )http://www.southern-digital.com
                                      )
                                      )http://www.seeit360.net
                                      )
                                      )404-551-4275
                                      )
                                      )
                                      )
                                      )
                                      )
                                      )
                                      )
                                      )From: PanoToolsNG@yahoogroups.com
                                      )[mailto:PanoToolsNG@yahoogroups.com] On Behalf Of Eric O'Brien
                                      )Sent: Monday, August 04, 2008 4:40 AM
                                      )To: PanoToolsNG@yahoogroups.com
                                      )Subject: Re: [PanoToolsNG] Re: Parquet flooring drives me crazy!
                                      )
                                      )
                                      )
                                      )Since we're talking about identical areas of a FLOOR failing
                                      )to stitch well, it doesn't seem that it would be a *parallax*
                                      )error, right? Therefore, the problem probably is not an
                                      )incorrectly placed "nodal point."
                                      )
                                      )It is more likely that the "warping coefficients" (or whatever
                                      )we call them) are not sufficiently precise. I believe that
                                      )AutopanoPro has some way to manually add control points. You
                                      )might try to find a view where existing control points are
                                      )shown, and manually add some toward the edges of your images.
                                      )
                                      )I'm assuming that APP will then use your added points and
                                      )generate better lens characterization values. You might try
                                      )asking for help on the APP forums.
                                      )
                                      )eo
                                      )
                                      )On Jul 30, 2008, at 2:00 PM, obarakana wrote:
                                      )
                                      )> Thanks everyone!
                                      )>
                                      )> Sorry, I had just deleted the unsatisfactory photos so I had nothing
                                      )> to show...
                                      )>
                                      )> I suspect the problem is at least partly due to a lack of precision
                                      )> establishing the nodal point. The entrance pupil database
                                      )looks to be
                                      )> a very useful tool. I had just been lining up the nearest
                                      )and farthest
                                      )> objects in the viewfinder and checking that their relative positions
                                      )> didn't (appear to) move when the camera was rotated. Apparently this
                                      )> wasn't good enough! I will have to spend more time here to learn...
                                      )>
                                      )> ------------------------------------
                                      )>
                                      )
                                      )
                                      )
                                      )
                                      )
                                      )[Non-text portions of this message have been removed]
                                      )
                                      )
                                      )------------------------------------
                                      )
                                      )--
                                      )
                                      )
                                      )
                                    • paul womack
                                      ... I ve had issues of this type with bricks/stone blocks on buildings, as well as leaded windows. They do provide wonderful control points - if you re
                                      Message 18 of 18 , Aug 5, 2008
                                      • 0 Attachment
                                        PanoToolsNG.10.m8@... wrote:
                                        > Another thing to watch for with repeating patterns(eg. parquet flooring), is
                                        > that the auto(or manual) control points "do" point to the correct places. It
                                        > is very easy for the control points to be pointing to slightly different
                                        > places in adjacent images.

                                        I've had issues of this type with bricks/stone blocks on buildings,
                                        as well as "leaded" windows.

                                        They do provide wonderful control points - if you're careful.

                                        BugBear
                                      Your message has been successfully submitted and would be delivered to recipients shortly.