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

Re: [PanoToolsNG] stitching of shots from the inside of a sphere

Expand Messages
  • John Riley
    Looks like some serious drusen in that eye! Macular degeneration? The fundus pictures my ophthalmologist takes look like they have a lot bigger field than test
    Message 1 of 5 , Nov 16, 2012
    • 0 Attachment
      Looks like some serious drusen in that eye! Macular degeneration? The fundus pictures my ophthalmologist takes look like they have a lot bigger field than test and I am pretty sure they are a single shot, though last time they did take a second, off-axis shot to see more.

      Sorry that I'm no help with the stitching.

      John

      John Riley
      4Pi-VR Media Solutions
      http://4pi-vr.com
      johnriley@...
      (h)864-461-3504
      (c)864-431-7075
      (w)864-503-5775

      On Nov 16, 2012, at 8:50 AM, tderveaux wrote:

       

      Hi all,

      I am trying to stitch ophthalmic images (fundus pictures)  in Hugin. The fundus camera's own software stitching process is far from ideal (bad vignetting removal for instance).

      Montages are created from multiple single circular shots, each subtending 50° of the retina, the inside of a sphere if you will.  Although Hugin has done a great job at stitching panoramas for ordinary outdoors shots for me, it has difficulties dealing with these images. I was wondering if I somehow have to communicate to Hugin that these shots are in fact already 2D projections of a spherical surface? Or should it just take them as they are?

      Examples:
      https://www.dropbox.com/sh/r1h70wmn4f5pfab/WGW6edwQjh

      How would I best feed these into panorama software? Other options of getting the best projection from the inside of a sphere are welcome of course.

      Thanks

      Thierry



    • Erik Krause
      ... I d use circular fisheye as input projection and allow for all lens distortions to optimize. To get further help you should provide at least two
      Message 2 of 5 , Nov 17, 2012
      • 0 Attachment
        Am 16.11.2012 14:50, schrieb tderveaux:
        > Examples:
        > https://www.dropbox.com/sh/r1h70wmn4f5pfab/WGW6edwQjh
        >
        > How would I best feed these into panorama software? Other options of
        > getting the best projection from the inside of a sphere are welcome of
        > course.

        I'd use circular fisheye as input projection and allow for all lens
        distortions to optimize. To get further help you should provide at least
        two overlapping images (better all).

        --
        Erik Krause
        http://www.erik-krause.de
      • Thomas
        Hugin actually assumes all input images are patches of a sphere seen from the inside (the panosphere, or sphere of vision) so should naturally do well at your
        Message 3 of 5 , Nov 19, 2012
        • 0 Attachment
          Hugin actually assumes all input images are patches of a sphere seen from the inside (the panosphere, or sphere of vision) so should naturally do well at your job. It can be made to stitch flat mosaics, but I doubt it could ever handle a sphere seen from outside.

          I agree with Erik that the most important thing is plenty of overlap between images. Specifying that the source projection is spherical might be a good starting point, but the lens that took the pictures has to considered too. If it is rectilinear, then what you have is some rectilinear projections of patches of a sphere, and a rectilinear correction might give a less distorted final picture. Ideally you should calibrate the camera by taking a 360 degree set of pictures of something other than an eye, having plenty of local detail, with a 50% overlap between images. Align those with optimization of all lens parameters, and use the resulting correction parameters for stitching the eyeball images with lens optimization disabled.

          --- In PanoToolsNG@yahoogroups.com, Erik Krause <erik.krause@...> wrote:
          >
          > Am 16.11.2012 14:50, schrieb tderveaux:
          > > Examples:
          > > https://www.dropbox.com/sh/r1h70wmn4f5pfab/WGW6edwQjh
          > >
          > > How would I best feed these into panorama software? Other options of
          > > getting the best projection from the inside of a sphere are welcome of
          > > course.
          >
          > I'd use circular fisheye as input projection and allow for all lens
          > distortions to optimize. To get further help you should provide at least
          > two overlapping images (better all).
          >
          > --
          > Erik Krause
          > http://www.erik-krause.de
          >
        • Erik Krause
          ... I don t think this would help. One element of the lens is the eye itself, which you can t calibrate. Furthermore there most likely is non-symmetric
          Message 4 of 5 , Nov 19, 2012
          • 0 Attachment
            Am 19.11.2012 18:07, schrieb Thomas:
            > Specifying that the source projection is spherical might be a good
            > starting point, but the lens that took the pictures has to considered
            > too. If it is rectilinear, then what you have is some rectilinear
            > projections of patches of a sphere, and a rectilinear correction
            > might give a less distorted final picture. Ideally you should
            > calibrate the camera by taking a 360 degree set of pictures of
            > something other than an eye, having plenty of local detail, with a
            > 50% overlap between images.

            I don't think this would help. One element of the "lens" is the eye
            itself, which you can't calibrate. Furthermore there most likely is
            non-symmetric distortion due to the different angles the ophtalmoscope
            (or fundus camera) is placed relative to the lens of the eye. So this
            images are not taken from the middle of the sphere, but from an opening
            in the sphere and with a "lens" where part of the lens elements are
            tilted relative to the others. All this would give a pretty non-uniform
            mapping.

            May be this could be corrected using mosaic mode (in hugin) respectively
            viewpoint correction (in PTGui). I guess a perfect fit would only be
            possible with morphing software.

            However, I would like to try on those images. Thierry, could you provide
            them?

            --
            Erik Krause
            http://www.erik-krause.de
          Your message has been successfully submitted and would be delivered to recipients shortly.