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

Stereo pano -- accurate calibration idea -- extra NPP camera.

Expand Messages
  • panovrx
    Shooting stereo panos with two cameras there are two main approaches to stitching calibration. One is to not have to do it by using a pixel accurate head and a
    Message 1 of 26 , Aug 6, 2011
    • 0 Attachment
      Shooting stereo panos with two cameras there are two main approaches to stitching calibration. One is to not have to do it by using a pixel accurate head and a digital level so the process is one of slice assembly or slice assembly + blending. Ie. your y, p and r are known values.
      The other approach is to use regular stitching software and try to mitigate the NPP character of the image sequences.-- By constraining point finding to distant areas by masking, by rigorous elimination of large deviation pairs, and by (somehow) restricting point finding to successive image pairs only. Still in some cases -- eg. a close wall behind the cameras -- this is not going to work. How to do it then?
      By using an extra camera -- between the two primary cameras (it could be between but above them.) It would not have to be the same camera/lens.
      You then just have to find the (fixed) r, p, y deviations of the L and R cameras from the central one.

      A related concept would be to have the L camera, say, being NPP and you use its values (compensated by the fixed deviations) for the R camera which is twice as non-NPP as usual. So you have to use twice as many shots as usual for a give scene depth range. But maybe that is not such a problem with static scenes at least.

      PeterM
    • Quentin Desouza
      Has anyone tried experimenting with this technology? http://www.cultofmac.com/new-mind-blowing-3d-mapping-technology-for-ipad-ces-2011/76046 Quentin,
      Message 2 of 26 , Aug 6, 2011
      • 0 Attachment
        Has anyone tried experimenting with this technology?
        http://www.cultofmac.com/new-mind-blowing-3d-mapping-technology-for-ipad-ces-2011/76046

        Quentin,
      • Wim Koornneef
        ... Hello Peter, Thanks for sharing this idea. Using a third camera for the sole purpopose of getting the y,r,p of the shoot could be a helpful solution,
        Message 3 of 26 , Aug 7, 2011
        • 0 Attachment
          panovrx-2 wrote:
          > ....By using an extra camera -- between the two primary cameras (it could
          > be between but above them.) It would not have to be the same camera/lens.
          > You then just have to find the (fixed) r, p, y deviations of the L and R
          > cameras from the central one....
          >

          Hello Peter,

          Thanks for sharing this idea.

          Using a third camera for the sole purpopose of getting the y,r,p of the
          shoot could be a helpful solution, unfortunately it will add extra weight to
          the already havey gear of a dual cam setup. Optimizing of the lens params
          d,e of the other cams will be still neccesary so I doubt this is the road to
          go for 3D panos.

          Best,
          Wim


          --
          View this message in context: http://panotoolsng.586017.n4.nabble.com/Stereo-pano-accurate-calibration-idea-extra-NPP-camera-tp3724211p3724609.html
          Sent from the PanoToolsNG mailing list archive at Nabble.com.
        • panovrx
          ... A version of the concept that only uses two cameras is this: Have your two cameras on a rail so that the two camera setup can be slid sideways over the
          Message 4 of 26 , Aug 7, 2011
          • 0 Attachment
            --- In PanoToolsNG@yahoogroups.com, Wim Koornneef <wim.koornneef@...> wrote:
            >
            >
            > panovrx-2 wrote:
            > > ....By using an extra camera -- between the two primary cameras (it could
            > > be between but above them.) It would not have to be the same camera/lens.
            > > You then just have to find the (fixed) r, p, y deviations of the L and R
            > > cameras from the central one....
            > >
            >
            > Hello Peter,
            >
            > Thanks for sharing this idea.
            >
            > Using a third camera for the sole purpopose of getting the y,r,p of the
            > shoot could be a helpful solution, unfortunately it will add extra weight to
            > the already havey gear of a dual cam setup. Optimizing of the lens params
            > d,e of the other cams will be still neccesary so I doubt this is the road to
            > go for 3D panos.
            >
            > Best,
            > Wim
            >
            >
            > --
            > View this message in context: http://panotoolsng.586017.n4.nabble.com/Stereo-pano-accurate-calibration-idea-extra-NPP-camera-tp3724211p3724609.html
            > Sent from the PanoToolsNG mailing list archive at Nabble.com.
            >

            A version of the concept that only uses two cameras is this: Have your two cameras on a rail so that the two camera setup can be slid sideways over the panorama head. So at one end of the rail movement both cameras are equidistant from the rotation axis with the NPPs of the cameras forming a straight line with the rotation axis. Then the cameras can be slid so the Left camera say has its NPP over the rotation axis. Many heads are reproducible (in the short term) in their dedent settings but not necessarily accurate otherwise. So now you shoot a sequence with the now NPP L camera and calibrate the rig movement for a 360 rotation (at 5 degree increments say) and then youslide back the cameras to the stereo pano shooting position(equidistant from the rotation axis). Your r, p, y template from the calibration sequence should still apply to the L camera and you can parallelize the L and R frames by calibrating the first pair of L and R and then the template should apply to R frames too.

            PeterM
          • panovrx
            ... Here btw is a recent stereo (anaglyph) panorama I made (using a standard setup). E: chanaglyph customh.htm This is from 34 shots per camera, in a 15 second
            Message 5 of 26 , Aug 7, 2011
            • 0 Attachment
              --- In PanoToolsNG@yahoogroups.com, "panovrx" <panovrx@...> wrote:
              >
              >
              >
              > --- In PanoToolsNG@yahoogroups.com, Wim Koornneef <wim.koornneef@> wrote:
              > >
              > >
              > > panovrx-2 wrote:
              > > > ....By using an extra camera -- between the two primary cameras (it could
              > > > be between but above them.) It would not have to be the same camera/lens.
              > > > You then just have to find the (fixed) r, p, y deviations of the L and R
              > > > cameras from the central one....
              > > >
              > >
              > > Hello Peter,
              > >
              > > Thanks for sharing this idea.
              > >
              > > Using a third camera for the sole purpopose of getting the y,r,p of the
              > > shoot could be a helpful solution, unfortunately it will add extra weight to
              > > the already havey gear of a dual cam setup. Optimizing of the lens params
              > > d,e of the other cams will be still neccesary so I doubt this is the road to
              > > go for 3D panos.
              > >
              > > Best,
              > > Wim
              > >
              > >
              > > --
              > > View this message in context: http://panotoolsng.586017.n4.nabble.com/Stereo-pano-accurate-calibration-idea-extra-NPP-camera-tp3724211p3724609.html
              > > Sent from the PanoToolsNG mailing list archive at Nabble.com.
              > >
              >
              > A version of the concept that only uses two cameras is this: Have your two cameras on a rail so that the two camera setup can be slid sideways over the panorama head. So at one end of the rail movement both cameras are equidistant from the rotation axis with the NPPs of the cameras forming a straight line with the rotation axis. Then the cameras can be slid so the Left camera say has its NPP over the rotation axis. Many heads are reproducible (in the short term) in their dedent settings but not necessarily accurate otherwise. So now you shoot a sequence with the now NPP L camera and calibrate the rig movement for a 360 rotation (at 5 degree increments say) and then youslide back the cameras to the stereo pano shooting position(equidistant from the rotation axis). Your r, p, y template from the calibration sequence should still apply to the L camera and you can parallelize the L and R frames by calibrating the first pair of L and R and then the template should apply to R frames too.
              >
              > PeterM
              >


              Here btw is a recent stereo (anaglyph) panorama I made (using a standard setup).
              E:\chanaglyph\customh.htm

              This is from 34 shots per camera, in a 15 second rotation. This is using SRaw1 with the intervalometer set to 0.4sec intervals.
              The interocular distance is 10cm. The cameras are about 4.5m high. So you can see 34 shots is enough for relatively good stitching at that height though the depth is not very strong with that lens separation.

              For blending of action stereo scenes with crowds, traffic I have settled on a multistage workflow currently where I use PTGui's masking tools to create a narrow vertical band in each source frame so that there is about 50 percent overlap in coverage of any given area -- ie. the horizontal masking is down to about 20 degrees here. Then I retouch those masks (in the L view say) so that action areas are only visible in one frame. Then I export these PTGui masked images as equi360 tifs with alpha. Then I reload these equi360 tifs into a new PTGui project and stitch with Smartblend. The problem with stitching from Smartblend directly is that it ignores PTGui's "masks" and grabs detail from various frames instead of closely adjacent frames - which makes parallax artefacts. By giving Smartblend only narrow retouched alpha images to deal with it constrains its blending action to where you want it.

              But how do you copy PTGui masks from L to the R images if you have calibrated the R images differently from the L (for finetuning after applying an L template,say). Well you can open two instances of PTGui and copy/paste the corresponding masks from the L to the R images one at a time. It would be good if you could do a mass copy of PTGui masks independent of their associated images.

              PeterM
            • panovrx
              ... Should be http://www.mediavr.com/customh.htm
              Message 6 of 26 , Aug 7, 2011
              • 0 Attachment
                --- In PanoToolsNG@yahoogroups.com, "panovrx" <panovrx@...> wrote:
                >
                >
                >
                > --- In PanoToolsNG@yahoogroups.com, "panovrx" <panovrx@> wrote:
                > >
                > >
                > >
                > > --- In PanoToolsNG@yahoogroups.com, Wim Koornneef <wim.koornneef@> wrote:
                > > >
                > > >
                > > > panovrx-2 wrote:
                > > > > ....By using an extra camera -- between the two primary cameras (it could
                > > > > be between but above them.) It would not have to be the same camera/lens.
                > > > > You then just have to find the (fixed) r, p, y deviations of the L and R
                > > > > cameras from the central one....
                > > > >
                > > >
                > > > Hello Peter,
                > > >
                > > > Thanks for sharing this idea.
                > > >
                > > > Using a third camera for the sole purpopose of getting the y,r,p of the
                > > > shoot could be a helpful solution, unfortunately it will add extra weight to
                > > > the already havey gear of a dual cam setup. Optimizing of the lens params
                > > > d,e of the other cams will be still neccesary so I doubt this is the road to
                > > > go for 3D panos.
                > > >
                > > > Best,
                > > > Wim
                > > >
                > > >
                > > > --
                > > > View this message in context: http://panotoolsng.586017.n4.nabble.com/Stereo-pano-accurate-calibration-idea-extra-NPP-camera-tp3724211p3724609.html
                > > > Sent from the PanoToolsNG mailing list archive at Nabble.com.
                > > >
                > >
                > > A version of the concept that only uses two cameras is this: Have your two cameras on a rail so that the two camera setup can be slid sideways over the panorama head. So at one end of the rail movement both cameras are equidistant from the rotation axis with the NPPs of the cameras forming a straight line with the rotation axis. Then the cameras can be slid so the Left camera say has its NPP over the rotation axis. Many heads are reproducible (in the short term) in their dedent settings but not necessarily accurate otherwise. So now you shoot a sequence with the now NPP L camera and calibrate the rig movement for a 360 rotation (at 5 degree increments say) and then youslide back the cameras to the stereo pano shooting position(equidistant from the rotation axis). Your r, p, y template from the calibration sequence should still apply to the L camera and you can parallelize the L and R frames by calibrating the first pair of L and R and then the template should apply to R frames too.
                > >
                > > PeterM
                > >
                >
                >
                > Here btw is a recent stereo (anaglyph) panorama I made (using a standard setup).
                > E:\chanaglyph\customh.htm
                >
                > This is from 34 shots per camera, in a 15 second rotation. This is using SRaw1 with the intervalometer set to 0.4sec intervals.
                > The interocular distance is 10cm. The cameras are about 4.5m high. So you can see 34 shots is enough for relatively good stitching at that height though the depth is not very strong with that lens separation.
                >
                > For blending of action stereo scenes with crowds, traffic I have settled on a multistage workflow currently where I use PTGui's masking tools to create a narrow vertical band in each source frame so that there is about 50 percent overlap in coverage of any given area -- ie. the horizontal masking is down to about 20 degrees here. Then I retouch those masks (in the L view say) so that action areas are only visible in one frame. Then I export these PTGui masked images as equi360 tifs with alpha. Then I reload these equi360 tifs into a new PTGui project and stitch with Smartblend. The problem with stitching from Smartblend directly is that it ignores PTGui's "masks" and grabs detail from various frames instead of closely adjacent frames - which makes parallax artefacts. By giving Smartblend only narrow retouched alpha images to deal with it constrains its blending action to where you want it.
                >
                > But how do you copy PTGui masks from L to the R images if you have calibrated the R images differently from the L (for finetuning after applying an L template,say). Well you can open two instances of PTGui and copy/paste the corresponding masks from the L to the R images one at a time. It would be good if you could do a mass copy of PTGui masks independent of their associated images.
                >
                > PeterM
                >
                Should be
                http://www.mediavr.com/customh.htm
              • panovrx
                ... Also it occurs to me that having a NPP camera with a synced non-NPP camera can be used for static scenes to obtain variable disparity stereo panos from
                Message 7 of 26 , Aug 9, 2011
                • 0 Attachment
                  --- In PanoToolsNG@yahoogroups.com, "panovrx" <panovrx@...> wrote:
                  >

                  > A related concept would be to have the L camera, say, being NPP and you use its values (compensated by the fixed deviations) for the R camera which is twice as non-NPP as usual. So you have to use twice as many shots as usual for a give scene depth range. But maybe that is not such a problem with static scenes at least.
                  >
                  > PeterM
                  >
                  Also it occurs to me that having a NPP camera with a synced non-NPP camera can be used for static scenes to obtain variable disparity stereo panos from the other single camera non-NPP sequence. Thusly: (imagining the L camera to be NPP)
                  apply the obtained template from the L camera to the R camera images (which have a fixed offset applied from calibrating the first L and R to each other). From the R images export individual layers of 360equis. These are now very well aligned. Trim slices left and right of the center of the image area in each of the aligned R image equis and assemble/blend in the usual single camera stereo pano rotation ways. By taking slices at various distances from the center of the image areas you can can get whatever parallax you need.

                  PeterM
                • panovrx
                  ... Here is a rig I put together to try having one camera NPP and the other not NPP. http://www.mediavr.com/nppstereo.jpg (anaglyph) I had the turntable set
                  Message 8 of 26 , Aug 13, 2011
                  • 0 Attachment
                    --- In PanoToolsNG@yahoogroups.com, "panovrx" <panovrx@...> wrote:
                    >
                    > Shooting stereo panos with two cameras there are two main approaches to stitching calibration. One is to not have to do it by using a pixel accurate head and a digital level so the process is one of slice assembly or slice assembly + blending. Ie. your y, p and r are known values.
                    > The other approach is to use regular stitching software and try to mitigate the NPP character of the image sequences.-- By constraining point finding to distant areas by masking, by rigorous elimination of large deviation pairs, and by (somehow) restricting point finding to successive image pairs only. Still in some cases -- eg. a close wall behind the cameras -- this is not going to work. How to do it then?
                    > By using an extra camera -- between the two primary cameras (it could be between but above them.) It would not have to be the same camera/lens.
                    > You then just have to find the (fixed) r, p, y deviations of the L and R cameras from the central one.
                    >
                    > A related concept would be to have the L camera, say, being NPP and you use its values (compensated by the fixed deviations) for the R camera which is twice as non-NPP as usual. So you have to use twice as many shots as usual for a give scene depth range. But maybe that is not such a problem with static scenes at least.
                    >
                    > PeterM
                    >


                    Here is a rig I put together to try having one camera NPP and the other not NPP.
                    http://www.mediavr.com/nppstereo.jpg (anaglyph)

                    I had the turntable set to one revolution in 50 seconds and 2 fps (= 100 shots per camera). I calibrated the first L and R to each other and produced aligned pairs of fullframe fisheye perspective images for each camera. Then I calibrated the NPP series of these for 360 and applied that as a template to the non-NPP images of the other camera. This is the unretouched result
                    http://www.mediavr.com/stairsana1.jpg (anaglyph)


                    this was shot on a monopod and even so the stitching for both is ok --
                    this is much quicker than other calibration methods I have tried -- the major disadvantage is you need more shots (I have ordered some faster CF cards to see if I can get the rotation time down and still shoot raw).

                    PeterM
                  • panovrx
                    ... Here is another one from yesterday which is nicer I think http://www.mediavr.com/pathlr.jpg (anaglyph) This is unretouched and stitched and processed from
                    Message 9 of 26 , Aug 13, 2011
                    • 0 Attachment
                      > > A related concept would be to have the L camera, say, being NPP and you use its values (compensated by the fixed deviations) for the R camera which is twice as non-NPP as usual. So you have to use twice as many shots as usual for a give scene depth range. But maybe that is not such a problem with static scenes at least.
                      > >
                      > > PeterM
                      > >
                      >
                      >
                      > Here is a rig I put together to try having one camera NPP and the other not NPP.
                      > http://www.mediavr.com/nppstereo.jpg (anaglyph)
                      >
                      > I had the turntable set to one revolution in 50 seconds and 2 fps (= 100 shots per camera). I calibrated the first L and R to each other and produced aligned pairs of fullframe fisheye perspective images for each camera. Then I calibrated the NPP series of these for 360 and applied that as a template to the non-NPP images of the other camera. This is the unretouched result
                      > http://www.mediavr.com/stairsana1.jpg (anaglyph)
                      >
                      >
                      > this was shot on a monopod and even so the stitching for both is ok --
                      > this is much quicker than other calibration methods I have tried -- the major disadvantage is you need more shots (I have ordered some faster CF cards to see if I can get the rotation time down and still shoot raw).
                      >
                      > PeterM
                      >

                      Here is another one from yesterday which is nicer I think
                      http://www.mediavr.com/pathlr.jpg (anaglyph)

                      This is unretouched and stitched and processed from raw in less than an hour.

                      So if you have two cameras and one is NPP and the cameras are mounted rigidly together and they are synced accurately stereo pano production can be automatic (if there is not too much action, wind etc). For scenes with closer detail a 3 camera rig might be better for action scenes or (for static scenes) a 2 camera rig with the second camera above the lower NPP camera and displaced forward of the rotation axis -- for single camera L/R slice assembly/blending -- would be the way to go.

                      PeterM
                    • panovrx
                      ... I did a trial in my kitchen of the single camera L/R slice assembly/blending concept using a second camera NPP sequence for the calibration. Here is the
                      Message 10 of 26 , Aug 13, 2011
                      • 0 Attachment
                        --- In PanoToolsNG@yahoogroups.com, "panovrx" <panovrx@...> wrote:
                        >
                        >
                        > > > A related concept would be to have the L camera, say, being NPP and you use its values (compensated by the fixed deviations) for the R camera which is twice as non-NPP as usual. So you have to use twice as many shots as usual for a give scene depth range. But maybe that is not such a problem with static scenes at least.
                        > > >
                        > > > PeterM
                        > > >
                        > >
                        > >
                        > > Here is a rig I put together to try having one camera NPP and the other not NPP.
                        > > http://www.mediavr.com/nppstereo.jpg (anaglyph)
                        > >
                        > > I had the turntable set to one revolution in 50 seconds and 2 fps (= 100 shots per camera). I calibrated the first L and R to each other and produced aligned pairs of fullframe fisheye perspective images for each camera. Then I calibrated the NPP series of these for 360 and applied that as a template to the non-NPP images of the other camera. This is the unretouched result
                        > > http://www.mediavr.com/stairsana1.jpg (anaglyph)
                        > >
                        > >
                        > > this was shot on a monopod and even so the stitching for both is ok --
                        > > this is much quicker than other calibration methods I have tried -- the major disadvantage is you need more shots (I have ordered some faster CF cards to see if I can get the rotation time down and still shoot raw).
                        > >
                        > > PeterM
                        > >
                        >
                        > Here is another one from yesterday which is nicer I think
                        > http://www.mediavr.com/pathlr.jpg (anaglyph)
                        >
                        > This is unretouched and stitched and processed from raw in less than an hour.
                        >
                        > So if you have two cameras and one is NPP and the cameras are mounted rigidly together and they are synced accurately stereo pano production can be automatic (if there is not too much action, wind etc). For scenes with closer detail a 3 camera rig might be better for action scenes or (for static scenes) a 2 camera rig with the second camera above the lower NPP camera and displaced forward of the rotation axis -- for single camera L/R slice assembly/blending -- would be the way to go.
                        >
                        > PeterM
                        >

                        I did a trial in my kitchen of the single camera L/R slice assembly/blending concept using a second camera NPP sequence for the calibration. Here is the rig I put together to test it
                        http://www.mediavr.com/nppstereo2.jpg (anaglyph).

                        The bottom camera is NPP, the top camera is forward of the rotation axis but otherwise parallel and aligned with the bottom camera. I aligned the cameras vertically (tilt) with a digital level and aligned them horizontally (yaw) by seeing that they were pointing in the same direction using the 10X LiveView. I used a Manfrotto rotator and used 5 degree increments (72 shots).

                        Then the workflow was this:
                        Calibrate the bottom NPP sequence.
                        Save as a template.
                        Apply this template to the top camera sequence.
                        From now on we can forget about the bottom camera images.

                        Export the top camera images as individual layers of 360 equis. (72X)
                        Load these 360 equis into PTGui.

                        .. this is maybe the non-obvious bit ..
                        Export 60 degree wide equi sections of each of these equis using the Fill Yaw feature of PTGui to just get a cropped image section of each equi -- using 5 degrees as the increment.i.e. Export as 1200 by 3600 images (using 0 roll and 0 pitch and 0/5/10/15 etc yaw).

                        Crop each of the 60 degree wide equis from the top camera from the left side so you have a 10 degree slice (200 pixels by 3600 pixels).

                        Load these into PTGui and tell PTGui the fov is 10 degrees and and use Fill Yaw to give them the correct Yaw ( Roll and Pitch are 0). Stitch into a 360 by 180 equi. This is the L final panorama.

                        Now you have to do the last two steps again cropping this time from the Right of the 60 degree wide equis. Now you have the R final panorama.

                        And it works good! For smaller spaces than a kitchen you would need more shots. How to do that with a Manfrotto head. No problem. Use the rotator section at the top of the head to do an offset of approx 2.5 degrees and do another set of 72 shots. It is not important that the offset be exact since the NPP character of the bottom cameras shots will still give an accurate solution.

                        PeterM
                      • Wim Koornneef
                        Hello Peter, Using 2 cameras for the shoot, one for calibration purposes and the other for the images is -a- solution to make things easier. But it is not
                        Message 11 of 26 , Aug 14, 2011
                        • 0 Attachment
                          Hello Peter,

                          Using 2 cameras for the shoot, one for calibration purposes and the other
                          for the images is -a- solution to make things easier.

                          But it is not -the- ultimate solution.

                          I have shot 3d 360x180 panos in small spaces with a single camera and I
                          never got stuck when optimizing the projekt so when I take in account the
                          havier gear, the extra mounting, the leveling issues, the synchronizing
                          issues, and other issues you have to deal with when using a second camera
                          then I really don't see the advantages.
                          Of course optimizing a projekt is much more simple with the camera in NPP
                          but when a camera is shifted forwards out of NPP just 3, 4 or 5 cm (the max
                          for small spaces) optimizing isn't so hard to do if you know the drill.

                          And even when using the template of the camera in NPP then there still is
                          the need of calibrating the lens params d and e for the second camera (the
                          one shifted out of NPP).
                          You can apply the v, a, b, c params of the first NPP camera for the second
                          camera but not d and e.
                          Perhaps your lenses are mounted so tight that there is never ever a small
                          movement of the lens in the bajonet mounting of the body and then you can
                          get re-use the d and e values of a previous calibrated projekt shot with the
                          second camera, but if there is any chance that the lens can move a bit in
                          the bajonet mount then you still have to optimize the template of the second
                          camera for d and e params after each shoot.

                          I understand why you first output full equis to make blended slices (I used
                          this method in the past) but you can do the same, and imo easier to do, when
                          using the mask option of PTGui 9 for this, just optimize the projekt and
                          then make 2 templates, one with a left mask and the other with a right mask
                          in it and after running both templates your done.

                          You can make the masks in Photoshop (colors red and black, save as PNG) or
                          make them with the mask tools of PTGui 9. You only have to make the left and
                          right mask once and then you can re-use them for other projekts
                          The shape of the masks is not critical so there is no need to fine tune the
                          shape. Btw, you only have to make one mask and then flip it in Photoshop to
                          get the other one.
                          Masks can be copied in one image and pasted in all other images.
                          In case there are small moving objects in the scene they can easily be
                          masked out in the masks.

                          I am running Smartblend with Wine on OSX and the PTGui masks are fully
                          accepted by Smartblend so I guess this will be the same on Windows.
                          I know there are more roads then just one going to Rome so in the end the
                          road you choose is a matter of personal preferences but may I advice you to
                          give the mask road a try ?

                          Success,
                          Wim



                          --
                          View this message in context: http://panotoolsng.586017.n4.nabble.com/Stereo-pano-accurate-calibration-idea-extra-NPP-camera-tp3724211p3742615.html
                          Sent from the PanoToolsNG mailing list archive at Nabble.com.
                        • panovrx
                          ... Hi Wim Thanks for the feedback! I have the impression (Win7) that Smartblend ignores PTGui masks which is why I make and reimport the separate layers with
                          Message 12 of 26 , Aug 14, 2011
                          • 0 Attachment
                            --- In PanoToolsNG@yahoogroups.com, Wim Koornneef <wim.koornneef@...> wrote:
                            >
                            > Hello Peter,
                            >
                            > Using 2 cameras for the shoot, one for calibration purposes and the other
                            > for the images is -a- solution to make things easier.
                            >
                            > But it is not -the- ultimate solution.
                            >
                            > I have shot 3d 360x180 panos in small spaces with a single camera and I
                            > never got stuck when optimizing the projekt so when I take in account the
                            > havier gear, the extra mounting, the leveling issues, the synchronizing
                            > issues, and other issues you have to deal with when using a second camera
                            > then I really don't see the advantages.
                            > Of course optimizing a projekt is much more simple with the camera in NPP
                            > but when a camera is shifted forwards out of NPP just 3, 4 or 5 cm (the max
                            > for small spaces) optimizing isn't so hard to do if you know the drill.
                            >
                            > And even when using the template of the camera in NPP then there still is
                            > the need of calibrating the lens params d and e for the second camera (the
                            > one shifted out of NPP).
                            > You can apply the v, a, b, c params of the first NPP camera for the second
                            > camera but not d and e.
                            > Perhaps your lenses are mounted so tight that there is never ever a small
                            > movement of the lens in the bajonet mounting of the body and then you can
                            > get re-use the d and e values of a previous calibrated projekt shot with the
                            > second camera, but if there is any chance that the lens can move a bit in
                            > the bajonet mount then you still have to optimize the template of the second
                            > camera for d and e params after each shoot.
                            >
                            > I understand why you first output full equis to make blended slices (I used
                            > this method in the past) but you can do the same, and imo easier to do, when
                            > using the mask option of PTGui 9 for this, just optimize the projekt and
                            > then make 2 templates, one with a left mask and the other with a right mask
                            > in it and after running both templates your done.
                            >
                            > You can make the masks in Photoshop (colors red and black, save as PNG) or
                            > make them with the mask tools of PTGui 9. You only have to make the left and
                            > right mask once and then you can re-use them for other projekts
                            > The shape of the masks is not critical so there is no need to fine tune the
                            > shape. Btw, you only have to make one mask and then flip it in Photoshop to
                            > get the other one.
                            > Masks can be copied in one image and pasted in all other images.
                            > In case there are small moving objects in the scene they can easily be
                            > masked out in the masks.
                            >
                            > I am running Smartblend with Wine on OSX and the PTGui masks are fully
                            > accepted by Smartblend so I guess this will be the same on Windows.
                            > I know there are more roads then just one going to Rome so in the end the
                            > road you choose is a matter of personal preferences but may I advice you to
                            > give the mask road a try ?
                            >
                            > Success,
                            > Wim
                            >
                            >
                            >
                            Hi Wim

                            Thanks for the feedback!

                            I have the impression (Win7) that Smartblend ignores PTGui masks which is why I make and reimport the separate layers with alpha.

                            My fisheye circles on all lens/camera combos seem to be very close anyway and I prealign pairwise (by levelling the first L and aligning it to the first R) the original shots and convert to FFFisheye (90*190). So maybe that is why I dont need to worry about d and e.

                            Yes that is right about the masks for making the slices but having the equis gives you the option of getting the parallax level you want (by taking slices more or less from the centre of frame).

                            PeterM
                          • panovrx
                            ... Here are a couple of unretouched anaglyph 360s done using this single camera slice assembly/blending + extra NNP camera for calibration approach (from 72
                            Message 13 of 26 , Aug 15, 2011
                            • 0 Attachment
                              --- In PanoToolsNG@yahoogroups.com, "panovrx" <panovrx@...> wrote:

                              > I did a trial in my kitchen of the single camera L/R slice assembly/blending concept using a second camera NPP sequence for the calibration. Here is the rig I put together to test it
                              > http://www.mediavr.com/nppstereo2.jpg (anaglyph).
                              >
                              > The bottom camera is NPP, the top camera is forward of the rotation axis but otherwise parallel and aligned with the bottom camera. I aligned the cameras vertically (tilt) with a digital level and aligned them horizontally (yaw) by seeing that they were pointing in the same direction using the 10X LiveView. I used a Manfrotto rotator and used 5 degree increments (72 shots).
                              >
                              > Then the workflow was this:
                              > Calibrate the bottom NPP sequence.
                              > Save as a template.
                              > Apply this template to the top camera sequence.
                              > From now on we can forget about the bottom camera images.
                              >
                              > Export the top camera images as individual layers of 360 equis. (72X)
                              > Load these 360 equis into PTGui.
                              >
                              > .. this is maybe the non-obvious bit ..
                              > Export 60 degree wide equi sections of each of these equis using the Fill Yaw feature of PTGui to just get a cropped image section of each equi -- using 5 degrees as the increment.i.e. Export as 1200 by 3600 images (using 0 roll and 0 pitch and 0/5/10/15 etc yaw).
                              >
                              > Crop each of the 60 degree wide equis from the top camera from the left side so you have a 10 degree slice (200 pixels by 3600 pixels).
                              >
                              > Load these into PTGui and tell PTGui the fov is 10 degrees and and use Fill Yaw to give them the correct Yaw ( Roll and Pitch are 0). Stitch into a 360 by 180 equi. This is the L final panorama.
                              >
                              > Now you have to do the last two steps again cropping this time from the Right of the 60 degree wide equis. Now you have the R final panorama.
                              >
                              > And it works good! For smaller spaces than a kitchen you would need more shots. How to do that with a Manfrotto head. No problem. Use the rotator section at the top of the head to do an offset of approx 2.5 degrees and do another set of 72 shots. It is not important that the offset be exact since the NPP character of the bottom cameras shots will still give an accurate solution.
                              >
                              > PeterM
                              >

                              Here are a couple of unretouched anaglyph 360s done using this single camera slice assembly/blending + extra NNP camera for calibration approach (from 72 shots)
                              (of scenes in a forest)
                              http://www.mediavr.com/canyonana1.jpg
                              http://www.mediavr.com/canyonana2.jpg

                              You get good control of amount of parallax this way but your L and R slices are effectively taken at different times so the scene needs to be very still. Here there is some breeze and water movement so there is a bit of retinal rivalry happening -- but for static interiors this approach is very useful. Then the only problem might be flare differences in the L and R slices in a pair.

                              PeterM
                            • panovrx
                              Here is a city scene in stereo with lots of people in a park http://www.mediavr.com/stereo/wynyardpark3.html (anaglyph) This uses this canera setup: -- with
                              Message 14 of 26 , Sep 7, 2011
                              • 0 Attachment
                                Here is a city scene in stereo with lots of people in a park
                                http://www.mediavr.com/stereo/wynyardpark3.html (anaglyph)

                                This uses this canera setup: -- with one camera rotating on the No Parallax Point and the other rotating offaxis
                                http://www.mediavr.com/nppstereo.jpg (anaglyph)

                                This incorporates Google Map (done with Panotour Pro) -- you can see it looks a little strange having a 2D map next to a 3d panorama. Maybe if I used "floating windows"
                                http://www.3diran3d.ir/Floating-Window/
                                this would make the map and panorama look more integrated.

                                This pano was 50 shots per camera approx -- for 360-- in 25 sec (ie. 2fps) at full size Raw (on 5DMkII). I used a pair of fast CF cards (Lexar Professional 600X) and this boosts the maximum synced continuous raw rate to 2 fps. I use the PClix intervalometer for the sub- 1sec timing. This is a useful gadget for testing CF card performance too.

                                The vertical range is restricted as I was using a masking hood over each lens -- to reduce flare.

                                PeterM
                              • Sacha Griffin
                                I think that s your best one yet. Thanks for sharing. -sacha From: PanoToolsNG@yahoogroups.com [mailto:PanoToolsNG@yahoogroups.com] On Behalf Of panovrx Sent:
                                Message 15 of 26 , Sep 8, 2011
                                • 0 Attachment

                                  I think that’s your best one yet. Thanks for sharing.

                                   

                                  -sacha

                                   

                                  From: PanoToolsNG@yahoogroups.com [mailto:PanoToolsNG@yahoogroups.com] On Behalf Of panovrx
                                  Sent: Wednesday, September 07, 2011 10:39 PM
                                  To: PanoToolsNG@yahoogroups.com
                                  Subject: [PanoToolsNG] Re: Stereo pano -- accurate calibration idea -- extra NPP camera.

                                   

                                   

                                  Here is a city scene in stereo with lots of people in a park
                                  http://www.mediavr.com/stereo/wynyardpark3.html (anaglyph)

                                  This uses this canera setup: -- with one camera rotating on the No Parallax Point and the other rotating offaxis
                                  http://www.mediavr.com/nppstereo.jpg (anaglyph)

                                  This incorporates Google Map (done with Panotour Pro) -- you can see it looks a little strange having a 2D map next to a 3d panorama. Maybe if I used "floating windows"
                                  http://www.3diran3d.ir/Floating-Window/
                                  this would make the map and panorama look more integrated.

                                  This pano was 50 shots per camera approx -- for 360-- in 25 sec (ie. 2fps) at full size Raw (on 5DMkII). I used a pair of fast CF cards (Lexar Professional 600X) and this boosts the maximum synced continuous raw rate to 2 fps. I use the PClix intervalometer for the sub- 1sec timing. This is a useful gadget for testing CF card performance too.

                                  The vertical range is restricted as I was using a masking hood over each lens -- to reduce flare.

                                  PeterM

                                • enridp
                                  I always enjoy reading your ideas and experiences with 3D Peter. Thanks for sharing !!!
                                  Message 16 of 26 , Sep 8, 2011
                                  • 0 Attachment
                                    I always enjoy reading your ideas and experiences with 3D Peter.
                                    Thanks for sharing !!!

                                    --- In PanoToolsNG@yahoogroups.com, "panovrx" <panovrx@...> wrote:
                                    >
                                    > Here is a city scene in stereo with lots of people in a park
                                    > http://www.mediavr.com/stereo/wynyardpark3.html (anaglyph)
                                    >
                                    > This uses this canera setup: -- with one camera rotating on the No Parallax Point and the other rotating offaxis
                                    > http://www.mediavr.com/nppstereo.jpg (anaglyph)
                                    >
                                    > This incorporates Google Map (done with Panotour Pro) -- you can see it looks a little strange having a 2D map next to a 3d panorama. Maybe if I used "floating windows"
                                    > http://www.3diran3d.ir/Floating-Window/
                                    > this would make the map and panorama look more integrated.
                                    >
                                    > This pano was 50 shots per camera approx -- for 360-- in 25 sec (ie. 2fps) at full size Raw (on 5DMkII). I used a pair of fast CF cards (Lexar Professional 600X) and this boosts the maximum synced continuous raw rate to 2 fps. I use the PClix intervalometer for the sub- 1sec timing. This is a useful gadget for testing CF card performance too.
                                    >
                                    > The vertical range is restricted as I was using a masking hood over each lens -- to reduce flare.
                                    >
                                    > PeterM
                                    >
                                  • panovrx
                                    ... This is with the same setup http://www.mediavr.com/stereo/martin2.html except it includes some embedded closeup 3d views (anaglyph) linked to hotspots. If
                                    Message 17 of 26 , Sep 10, 2011
                                    • 0 Attachment
                                      --- In PanoToolsNG@yahoogroups.com, "panovrx" <panovrx@...> wrote:
                                      >
                                      > Here is a city scene in stereo with lots of people in a park
                                      > http://www.mediavr.com/stereo/wynyardpark3.html (anaglyph)
                                      >
                                      > This uses this canera setup: -- with one camera rotating on the No Parallax Point and the other rotating offaxis
                                      > http://www.mediavr.com/nppstereo.jpg (anaglyph)
                                      >
                                      > This incorporates Google Map (done with Panotour Pro) -- you can see it looks a little strange having a 2D map next to a 3d panorama. Maybe if I used "floating windows"
                                      > http://www.3diran3d.ir/Floating-Window/
                                      > this would make the map and panorama look more integrated.
                                      >
                                      > This pano was 50 shots per camera approx -- for 360-- in 25 sec (ie. 2fps) at full size Raw (on 5DMkII). I used a pair of fast CF cards (Lexar Professional 600X) and this boosts the maximum synced continuous raw rate to 2 fps. I use the PClix intervalometer for the sub- 1sec timing. This is a useful gadget for testing CF card performance too.
                                      >
                                      > The vertical range is restricted as I was using a masking hood over each lens -- to reduce flare.

                                      This is with the same setup
                                      http://www.mediavr.com/stereo/martin2.html
                                      except it includes some embedded closeup 3d views (anaglyph) linked to hotspots. If I could figure out how to make warped krpano hotspots do the flyout hotspot image effect seamlessly in 3d that would be a very cool effect I think.

                                      PeterM
                                    • enridp
                                      Hi peter, what do you mean with make warped krpano hotspots do the flyout hotspot image effect seamlessly in 3d ? To some type of transition or animation? Or
                                      Message 18 of 26 , Sep 11, 2011
                                      • 0 Attachment
                                        Hi peter, what do you mean with "make warped krpano hotspots do the flyout hotspot image effect seamlessly in 3d"?
                                        To some type of transition or animation?
                                        Or just to make a 3D hotspot that merges with the proper scene?
                                        For the last, maybe you can create the 3D hotspot with your stereo program (photoshop or anywhere you are using), and then use that image instead the polyline created by krpano.


                                        --- In PanoToolsNG@yahoogroups.com, "panovrx" <panovrx@...> wrote:
                                        >
                                        >
                                        >
                                        > --- In PanoToolsNG@yahoogroups.com, "panovrx" <panovrx@> wrote:
                                        > >
                                        > > Here is a city scene in stereo with lots of people in a park
                                        > > http://www.mediavr.com/stereo/wynyardpark3.html (anaglyph)
                                        > >
                                        > > This uses this canera setup: -- with one camera rotating on the No Parallax Point and the other rotating offaxis
                                        > > http://www.mediavr.com/nppstereo.jpg (anaglyph)
                                        > >
                                        > > This incorporates Google Map (done with Panotour Pro) -- you can see it looks a little strange having a 2D map next to a 3d panorama. Maybe if I used "floating windows"
                                        > > http://www.3diran3d.ir/Floating-Window/
                                        > > this would make the map and panorama look more integrated.
                                        > >
                                        > > This pano was 50 shots per camera approx -- for 360-- in 25 sec (ie. 2fps) at full size Raw (on 5DMkII). I used a pair of fast CF cards (Lexar Professional 600X) and this boosts the maximum synced continuous raw rate to 2 fps. I use the PClix intervalometer for the sub- 1sec timing. This is a useful gadget for testing CF card performance too.
                                        > >
                                        > > The vertical range is restricted as I was using a masking hood over each lens -- to reduce flare.
                                        >
                                        > This is with the same setup
                                        > http://www.mediavr.com/stereo/martin2.html
                                        > except it includes some embedded closeup 3d views (anaglyph) linked to hotspots. If I could figure out how to make warped krpano hotspots do the flyout hotspot image effect seamlessly in 3d that would be a very cool effect I think.
                                        >
                                        > PeterM
                                        >
                                      • jrgen_schrader
                                        I m impressed, excellent work and technique. Also the images from those hotspots really pop . Cheers Jürgen
                                        Message 19 of 26 , Sep 11, 2011
                                        • 0 Attachment
                                          I'm impressed, excellent work and technique.
                                          Also the images from those hotspots really "pop".

                                          Cheers
                                          Jürgen

                                          --- In PanoToolsNG@yahoogroups.com, "panovrx" <panovrx@...> wrote:
                                          >
                                          >
                                          >
                                          > --- In PanoToolsNG@yahoogroups.com, "panovrx" <panovrx@> wrote:
                                          > >
                                          > > Here is a city scene in stereo with lots of people in a park
                                          > > http://www.mediavr.com/stereo/wynyardpark3.html (anaglyph)
                                          > >
                                          > > This uses this canera setup: -- with one camera rotating on the No Parallax Point and the other rotating offaxis
                                          > > http://www.mediavr.com/nppstereo.jpg (anaglyph)
                                          > >
                                          > > This incorporates Google Map (done with Panotour Pro) -- you can see it looks a little strange having a 2D map next to a 3d panorama. Maybe if I used "floating windows"
                                          > > http://www.3diran3d.ir/Floating-Window/
                                          > > this would make the map and panorama look more integrated.
                                          > >
                                          > > This pano was 50 shots per camera approx -- for 360-- in 25 sec (ie. 2fps) at full size Raw (on 5DMkII). I used a pair of fast CF cards (Lexar Professional 600X) and this boosts the maximum synced continuous raw rate to 2 fps. I use the PClix intervalometer for the sub- 1sec timing. This is a useful gadget for testing CF card performance too.
                                          > >
                                          > > The vertical range is restricted as I was using a masking hood over each lens -- to reduce flare.
                                          >
                                          > This is with the same setup
                                          > http://www.mediavr.com/stereo/martin2.html
                                          > except it includes some embedded closeup 3d views (anaglyph) linked to hotspots. If I could figure out how to make warped krpano hotspots do the flyout hotspot image effect seamlessly in 3d that would be a very cool effect I think.
                                          >
                                          > PeterM
                                          >
                                        • panovrx
                                          http://www.mediavr.com/newtownanaglyph2/newtownanaglyph2.htm The shutter lag of a Canon5D is listed as 75ms -- close to the 5DMkII s listed 73ms -- so I
                                          Message 20 of 26 , Oct 10, 2011
                                          • 0 Attachment
                                            http://www.mediavr.com/newtownanaglyph2/newtownanaglyph2.htm

                                            The shutter lag of a Canon5D is listed as 75ms -- close to the 5DMkII's listed 73ms -- so I thought I would test the sync potential of a 5D and my 5DMkIIs together (with spliced cable releases) -- and it is fine (so long at least as you use a manual lens -- or a lens with taped over contacts - on the 5D). So I have been testing three camera configurations (side by side in a row here) lately. With a 5d/Sigma15mm as the NPP camera and two 5DMkII/10.5Nikkors as the stereo pair. Interlens separation for the stereo cameras was about 20cm.

                                            This, with a tripod too to carry, is starting to become a heavy load
                                            but it does give reliable calibration with busy street scenes, especially, for hyperstereo scenes. I set the 5d to M JPG and the 5dMkIIs to Raw, and set the intervalometer to 0.5sec and they will stay in sync indefinitely.

                                            The 5D/Sigma sequence is calibrated and that template is applied to the stereo camera sequences (and the lens definition changed). This works pretty well straight off, but I get more resolution by then generating control points for the stereo sequences afresh using the prepositioning the Sigma template provides and recalibrating them again as a fine tune. It is quicker and more reliable mostly than trying to calibrate them without the Sigma template first -- especially with action, hyperstereo scenes.

                                            Here there were about 200 shots per camera (about 4 rotations) at 2 fps. I kept the cameras going around as there was so much traffic and I knew I needed multiple records of every area of the scene if it was to be stitchable.

                                            PeterM




                                            --- In PanoToolsNG@yahoogroups.com, "panovrx" <panovrx@...> wrote:
                                            >
                                            > Shooting stereo panos with two cameras there are two main approaches to stitching calibration. One is to not have to do it by using a pixel accurate head and a digital level so the process is one of slice assembly or slice assembly + blending. Ie. your y, p and r are known values.
                                            > The other approach is to use regular stitching software and try to mitigate the NPP character of the image sequences.-- By constraining point finding to distant areas by masking, by rigorous elimination of large deviation pairs, and by (somehow) restricting point finding to successive image pairs only. Still in some cases -- eg. a close wall behind the cameras -- this is not going to work. How to do it then?
                                            > By using an extra camera -- between the two primary cameras (it could be between but above them.) It would not have to be the same camera/lens.
                                            > You then just have to find the (fixed) r, p, y deviations of the L and R cameras from the central one.

                                            >
                                            > PeterM
                                            >
                                          • Sacha Griffin
                                            Superb. Not an artifact to be found. I must say I like your color shots a bit more. Sacha From: PanoToolsNG@yahoogroups.com
                                            Message 21 of 26 , Oct 10, 2011
                                            • 0 Attachment

                                              Superb. Not an artifact to be found.

                                              I must say I like your color shots a bit more.

                                               

                                              Sacha

                                               

                                               

                                              From: PanoToolsNG@yahoogroups.com [mailto:PanoToolsNG@yahoogroups.com] On Behalf Of panovrx
                                              Sent: Monday, October 10, 2011 8:24 PM
                                              To: PanoToolsNG@yahoogroups.com
                                              Subject: [PanoToolsNG] Re: Stereo pano -- accurate calibration idea -- extra NPP camera.

                                               

                                               

                                              http://www.mediavr.com/newtownanaglyph2/newtownanaglyph2.htm

                                              The shutter lag of a Canon5D is listed as 75ms -- close to the 5DMkII's listed 73ms -- so I thought I would test the sync potential of a 5D and my 5DMkIIs together (with spliced cable releases) -- and it is fine (so long at least as you use a manual lens -- or a lens with taped over contacts - on the 5D). So I have been testing three camera configurations (side by side in a row here) lately. With a 5d/Sigma15mm as the NPP camera and two 5DMkII/10.5Nikkors as the stereo pair. Interlens separation for the stereo cameras was about 20cm.

                                              This, with a tripod too to carry, is starting to become a heavy load
                                              but it does give reliable calibration with busy street scenes, especially, for hyperstereo scenes. I set the 5d to M JPG and the 5dMkIIs to Raw, and set the intervalometer to 0.5sec and they will stay in sync indefinitely.

                                              The 5D/Sigma sequence is calibrated and that template is applied to the stereo camera sequences (and the lens definition changed). This works pretty well straight off, but I get more resolution by then generating control points for the stereo sequences afresh using the prepositioning the Sigma template provides and recalibrating them again as a fine tune. It is quicker and more reliable mostly than trying to calibrate them without the Sigma template first -- especially with action, hyperstereo scenes.

                                              Here there were about 200 shots per camera (about 4 rotations) at 2 fps. I kept the cameras going around as there was so much traffic and I knew I needed multiple records of every area of the scene if it was to be stitchable.

                                              PeterM

                                              --- In PanoToolsNG@yahoogroups.com, "panovrx" <panovrx@...> wrote:
                                              >
                                              > Shooting stereo panos with two cameras there are two main approaches to stitching calibration. One is to not have to do it by using a pixel accurate head and a digital level so the process is one of slice assembly or slice assembly + blending. Ie. your y, p and r are known values.
                                              > The other approach is to use regular stitching software and try to mitigate the NPP character of the image sequences.-- By constraining point finding to distant areas by masking, by rigorous elimination of large deviation pairs, and by (somehow) restricting point finding to successive image pairs only. Still in some cases -- eg. a close wall behind the cameras -- this is not going to work. How to do it then?
                                              > By using an extra camera -- between the two primary cameras (it could be between but above them.) It would not have to be the same camera/lens.
                                              > You then just have to find the (fixed) r, p, y deviations of the L and R cameras from the central one.

                                              >
                                              > PeterM
                                              >

                                            • Ken Warner
                                              I get a ghost in the red side. The blue side is clean. Maybe a little less cyan and more blue on the blue side would eliminate the ghost.
                                              Message 22 of 26 , Oct 10, 2011
                                              • 0 Attachment
                                                I get a ghost in the red side. The blue side is clean. Maybe a little less cyan and more blue on the blue side would eliminate the ghost.

                                                Sacha Griffin wrote:
                                                > Superb. Not an artifact to be found.
                                                >
                                                > I must say I like your color shots a bit more.
                                                >
                                                >
                                                >
                                                > Sacha
                                                >
                                                >
                                                >
                                                >
                                                >
                                                > From: PanoToolsNG@yahoogroups.com [mailto:PanoToolsNG@yahoogroups.com] On
                                                > Behalf Of panovrx
                                                > Sent: Monday, October 10, 2011 8:24 PM
                                                > To: PanoToolsNG@yahoogroups.com
                                                > Subject: [PanoToolsNG] Re: Stereo pano -- accurate calibration idea -- extra
                                                > NPP camera.
                                                >
                                                >
                                                >
                                                >
                                                >
                                                > http://www.mediavr.com/newtownanaglyph2/newtownanaglyph2.htm
                                                >
                                                > The shutter lag of a Canon5D is listed as 75ms -- close to the 5DMkII's
                                                > listed 73ms -- so I thought I would test the sync potential of a 5D and my
                                                > 5DMkIIs together (with spliced cable releases) -- and it is fine (so long at
                                                > least as you use a manual lens -- or a lens with taped over contacts - on
                                                > the 5D). So I have been testing three camera configurations (side by side in
                                                > a row here) lately. With a 5d/Sigma15mm as the NPP camera and two
                                                > 5DMkII/10.5Nikkors as the stereo pair. Interlens separation for the stereo
                                                > cameras was about 20cm.
                                                >
                                                > This, with a tripod too to carry, is starting to become a heavy load
                                                > but it does give reliable calibration with busy street scenes, especially,
                                                > for hyperstereo scenes. I set the 5d to M JPG and the 5dMkIIs to Raw, and
                                                > set the intervalometer to 0.5sec and they will stay in sync indefinitely.
                                                >
                                                > The 5D/Sigma sequence is calibrated and that template is applied to the
                                                > stereo camera sequences (and the lens definition changed). This works pretty
                                                > well straight off, but I get more resolution by then generating control
                                                > points for the stereo sequences afresh using the prepositioning the Sigma
                                                > template provides and recalibrating them again as a fine tune. It is quicker
                                                > and more reliable mostly than trying to calibrate them without the Sigma
                                                > template first -- especially with action, hyperstereo scenes.
                                                >
                                                > Here there were about 200 shots per camera (about 4 rotations) at 2 fps. I
                                                > kept the cameras going around as there was so much traffic and I knew I
                                                > needed multiple records of every area of the scene if it was to be
                                                > stitchable.
                                                >
                                                > PeterM
                                                >
                                                > --- In PanoToolsNG@yahoogroups.com <mailto:PanoToolsNG%40yahoogroups.com> ,
                                                > "panovrx" <panovrx@...> wrote:
                                                >> Shooting stereo panos with two cameras there are two main approaches to
                                                > stitching calibration. One is to not have to do it by using a pixel accurate
                                                > head and a digital level so the process is one of slice assembly or slice
                                                > assembly + blending. Ie. your y, p and r are known values.
                                                >> The other approach is to use regular stitching software and try to
                                                > mitigate the NPP character of the image sequences.-- By constraining point
                                                > finding to distant areas by masking, by rigorous elimination of large
                                                > deviation pairs, and by (somehow) restricting point finding to successive
                                                > image pairs only. Still in some cases -- eg. a close wall behind the cameras
                                                > -- this is not going to work. How to do it then?
                                                >> By using an extra camera -- between the two primary cameras (it could be
                                                > between but above them.) It would not have to be the same camera/lens.
                                                >> You then just have to find the (fixed) r, p, y deviations of the L and R
                                                > cameras from the central one.
                                                >
                                                >> PeterM
                                                >>
                                                >
                                                >
                                                >
                                                >
                                              • Wim Koornneef
                                                ... Ghosting is most times caused when monitors are not color calibrated or when the color gamut of the screen is to narrow. For the first reason TV sets are
                                                Message 23 of 26 , Oct 10, 2011
                                                • 0 Attachment
                                                  DemonDuck wrote:
                                                  >
                                                  > I get a ghost in the red side. The blue side is clean. Maybe a little
                                                  > less cyan and more blue on the blue side would eliminate the ghost.

                                                  Ghosting is most times caused when monitors are not color calibrated or when
                                                  the color gamut of the screen is to narrow.
                                                  For the first reason TV sets are giving lots of problems problems and for
                                                  the second f.e. the iPad and iPhone.
                                                  Depending on the anaglyph system (red/cyan, amber/blue, grey/full/half
                                                  color) this problem can be more or less.

                                                  If your monitor is calibrated properly then I am afraid there is nothing
                                                  that can be done other then viewing grey anaglyphs.

                                                  Wim



                                                  --
                                                  View this message in context: http://panotoolsng.586017.n4.nabble.com/Stereo-pano-accurate-calibration-idea-extra-NPP-camera-tp3724211p3892981.html
                                                  Sent from the PanoToolsNG mailing list archive at Nabble.com.
                                                • onezebra1
                                                  I have 2 Canon 5D-2 cameras that I need to set up with a better way for tripping them. I was afraid to try a spliced cable releases, is it ok to run a spliced
                                                  Message 24 of 26 , Oct 11, 2011
                                                  • 0 Attachment
                                                    I have 2 Canon 5D-2 cameras that I need to set up with a better way for tripping them. I was afraid to try a spliced cable releases, is it ok to run a spliced cable hooking the 2 cameras together where I can use only 1 wireless remote to trip them?

                                                    As for viewing these photos, I got 3 different pairs of glasses and only one set gave me a perfect 3D view.

                                                    --- In PanoToolsNG@yahoogroups.com, Wim Koornneef <wim.koornneef@...> wrote:
                                                    >
                                                    >
                                                    > DemonDuck wrote:
                                                    > >
                                                    > > I get a ghost in the red side. The blue side is clean. Maybe a little
                                                    > > less cyan and more blue on the blue side would eliminate the ghost.
                                                    >
                                                    > Ghosting is most times caused when monitors are not color calibrated or when
                                                    > the color gamut of the screen is to narrow.
                                                    > For the first reason TV sets are giving lots of problems problems and for
                                                    > the second f.e. the iPad and iPhone.
                                                    > Depending on the anaglyph system (red/cyan, amber/blue, grey/full/half
                                                    > color) this problem can be more or less.
                                                    >
                                                    > If your monitor is calibrated properly then I am afraid there is nothing
                                                    > that can be done other then viewing grey anaglyphs.
                                                    >
                                                    > Wim
                                                    >
                                                    >
                                                    >
                                                    > --
                                                    > View this message in context: http://panotoolsng.586017.n4.nabble.com/Stereo-pano-accurate-calibration-idea-extra-NPP-camera-tp3724211p3892981.html
                                                    > Sent from the PanoToolsNG mailing list archive at Nabble.com.
                                                    >
                                                  • Ken Warner
                                                    I wonder if color sliders for red and blue colors could be used to adjust the anaglyph colors to the monitor/glasses???
                                                    Message 25 of 26 , Oct 11, 2011
                                                    • 0 Attachment
                                                      I wonder if color sliders for red and blue colors could be used to adjust the anaglyph colors to the monitor/glasses???

                                                      onezebra1 wrote:
                                                      > I have 2 Canon 5D-2 cameras that I need to set up with a better way for tripping them. I was afraid to try a spliced cable releases, is it ok to run a spliced cable hooking the 2 cameras together where I can use only 1 wireless remote to trip them?
                                                      >
                                                      > As for viewing these photos, I got 3 different pairs of glasses and only one set gave me a perfect 3D view.
                                                      >
                                                      > --- In PanoToolsNG@yahoogroups.com, Wim Koornneef <wim.koornneef@...> wrote:
                                                      >>
                                                      >> DemonDuck wrote:
                                                      >>> I get a ghost in the red side. The blue side is clean. Maybe a little
                                                      >>> less cyan and more blue on the blue side would eliminate the ghost.
                                                      >> Ghosting is most times caused when monitors are not color calibrated or when
                                                      >> the color gamut of the screen is to narrow.
                                                      >> For the first reason TV sets are giving lots of problems problems and for
                                                      >> the second f.e. the iPad and iPhone.
                                                      >> Depending on the anaglyph system (red/cyan, amber/blue, grey/full/half
                                                      >> color) this problem can be more or less.
                                                      >>
                                                      >> If your monitor is calibrated properly then I am afraid there is nothing
                                                      >> that can be done other then viewing grey anaglyphs.
                                                      >>
                                                      >> Wim
                                                      >>
                                                      >>
                                                      >>
                                                      >> --
                                                      >> View this message in context: http://panotoolsng.586017.n4.nabble.com/Stereo-pano-accurate-calibration-idea-extra-NPP-camera-tp3724211p3892981.html
                                                      >> Sent from the PanoToolsNG mailing list archive at Nabble.com.
                                                      >>
                                                      >
                                                      >
                                                      >
                                                    • panovrx
                                                      http://www.mediavr.com/newtownanaglyph2/newtownanaglyph2.htm ... Yes this is what I do - without any issues so far. Actually with this particular setup I have
                                                      Message 26 of 26 , Oct 11, 2011
                                                      • 0 Attachment
                                                        http://www.mediavr.com/newtownanaglyph2/newtownanaglyph2.htm

                                                        --- In PanoToolsNG@yahoogroups.com, "onezebra1" <onezebra1@...> wrote:
                                                        >
                                                        >
                                                        > I have 2 Canon 5D-2 cameras that I need to set up with a better way for tripping them. I was afraid to try a spliced cable releases, is it ok to run a spliced cable hooking the 2 cameras together where I can use only 1 wireless remote to trip them?
                                                        >

                                                        Yes this is what I do - without any issues so far. Actually with this particular setup I have the intervalometer (PClix) switched on and connected to a solid state relay and the remote release receiver is wired into that. I do this because you can only start that intervalometer manually.

                                                        Re shooting 5D MkIIs in long continuous bursts. This will quickly reduce the shutter life of your cameras but no more I suspect than just shooting the equivalent number of shots randomly. I dont think firing two cameras together makes it worse.

                                                        5ds are maybe different. At one point I had two 5Ds which I used for timelapse etc and both lost their mirrors at some point .ie. the mirrors came detached in the middle of long bursts. Not when they were in pairs.
                                                        This is a bad sound to be hearing! but it was not a serious repair problem -- unlike a shutter repair. Maybe they have fixed the glue on the 5d MkII mirrors

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