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

Re: [PanoToolsNG] Report on Circular Fisheye lens for APS-C Format

Expand Messages
  • Michel Thoby
    Hi Roger, ... Thanks a lot for translating and sharing this info with us. IMO some untold facts could be added to the article written by the Yasuhara
    Message 1 of 4 , Aug 4, 2012
      Hi Roger,

      Le 4 août 2012 à 06:49, Roger D. Williams a écrit :

      The August issue of Asahi Camera magazine (no relation to Asahi
      Pentax) carries a report on the MADOKA 180 lens.
      (....)
      The article was written by someone who isn't very familiar with fisheye
      lens and doesn't even mention their application to stitched panoramas.
      (...)
      Software that will "correct the huge distortion" is being developed, and will be limited to a 140-degree rectilinear FOV.
      (....)

      Roger W.

      Thanks a lot for translating and sharing this info with us. 

      IMO some untold facts could be added to the article written by the Yasuhara company...
      FYI There had been several discussions about it rolling lately on the Panoguide forum:

      This lens is very special as it's AFAIK the first  commercial fisheye lens with Orthographic projection that can be mounted on a commercial (APS-c or larger) camera. Orthographic could be approximately described as the "opposite" to Stereographic for fisheye lens.
      Because of this rare projection, there is a huge compression near the edge of the circular image and the FoV is absolutely restricted: a sole row of image in portrait mode cannot give an VFoV strictly equal to180 degrees because some few pixels shall always be missing before this ultimate limit.
      Theoretically, a full HFoV 360 degrees panorama can be got from only three shots around. But from few early intricate experiments (with real and synthetic images, all having the same fisheye orthographic radial mapping), I have concluded that it is unwise to do so. Four images around are OK though.
      Hints:
      1) The Optimizer function of current leading panorama stitching programs become (more often than not) unstable when valid Control Points very near the image circular boundary are manually placed and "forced" by the user.
      2) The 360 x 180 output panorama is good elsewhere but it is very distorted (and of a poor quality) near and around the Nadir" or around the Zenith" (or both): there is a small star-like blurry pattern surrounding the "empty" circular black hole(s).
      So, and in short: 
      Cropping the circular image might be implicitly required for panographic purpose too! If not done, the so-called "huge distortion" cannot be perfectly corrected (whatever the software is).
      Pitch above or below the horizontal (by, say, ten or fifteen degree) is recommended and it shall be required to shoot one (of more) supplementary image to fill (or to patch) the empty hole at the pole(s) in order to get a full spherical pano.

      Michel


    • Roger D Williams
      Good additional information, Michel! ... By the way, the article was written by a supposedly impartial journalist on the Asahi Camera staff, not by Yasuhara s
      Message 2 of 4 , Aug 4, 2012
        Good additional information, Michel!

        On Aug 4, 2012, at 6:27 PM, Michel Thoby <thobymichel@...> wrote:

        Hi Roger,

        Le 4 août 2012 à 06:49, Roger D. Williams a écrit :

        The August issue of Asahi Camera magazine (no relation to Asahi
        Pentax) carries a report on the MADOKA 180 lens.
        (....)
        The article was written by someone who isn't very familiar with fisheye
        lens and doesn't even mention their application to stitched panoramas.
        (...)
        Software that will "correct the huge distortion" is being developed, and will be limited to a 140-degree rectilinear FOV.
        (....)

        Roger W.

        Thanks a lot for translating and sharing this info with us. 

        IMO some untold facts could be added to the article written by the Yasuhara company...
        FYI There had been several discussions about it rolling lately on the Panoguide forum:

        By the way, the article was written by a supposedly impartial journalist on the Asahi Camera staff, not by Yasuhara's people. Mind you, they generally never write anything critical about the products they review...

        This lens is very special as it's AFAIK the first  commercial fisheye lens with Orthographic projection that can be mounted on a commercial (APS-c or larger) camera. Orthographic could be approximately described as the "opposite" to Stereographic for fisheye lens.
        Because of this rare projection, there is a huge compression near the edge of the circular image and the FoV is absolutely restricted: a sole row of image in portrait mode cannot give an VFoV strictly equal to180 degrees because some few pixels shall always be missing before this ultimate limit.
        Theoretically, a full HFoV 360 degrees panorama can be got from only three shots around. But from few early intricate experiments (with real and synthetic images, all having the same fisheye orthographic radial mapping), I have concluded that it is unwise to do so. Four images around are OK though.
        Hints:
        1) The Optimizer function of current leading panorama stitching programs become (more often than not) unstable when valid Control Points very near the image circular boundary are manually placed and "forced" by the user.
        2) The 360 x 180 output panorama is good elsewhere but it is very distorted (and of a poor quality) near and around the Nadir" or around the Zenith" (or both): there is a small star-like blurry pattern surrounding the "empty" circular black hole(s).
        So, and in short: 
        Cropping the circular image might be implicitly required for panographic purpose too! If not done, the so-called "huge distortion" cannot be perfectly corrected (whatever the software is).
        Pitch above or below the horizontal (by, say, ten or fifteen degree) is recommended and it shall be required to shoot one (of more) supplementary image to fill (or to patch) the empty hole at the pole(s) in order to get a full spherical 

        That all makes good sense. Looks as if I could get back to four shots at right angles on a hand-held monopod if they are all pointed slightly down to cover the nadir area provided I take one pointing up to fill in the zenith. Sounds good. Of course I would need a Sony NEX! I doubt if a Pentax mount could be made due to the very different geometry. Unfortunately I find the NEX line completely uncongenial and I cannot imagine buying one.

        Roger W.
      • prague
        Hi Michel, ... Is this because the 3rd degree polynomial lens model used by panotools is not sufficient to model this kind of projection? Maybe Tom Sharpless
        Message 3 of 4 , Aug 5, 2012
          Hi Michel,

          --- In PanoToolsNG@yahoogroups.com, Michel Thoby <thobymichel@...> wrote:
          >
          > 1) The Optimizer function of current leading panorama stitching programs become (more >often than not) unstable when valid Control Points very near the image circular boundary are >manually placed and "forced" by the user.

          Is this because the 3rd degree polynomial lens model used by panotools is not sufficient to model this kind of projection?

          Maybe Tom Sharpless could also comment, I know he's working on lens correction software with a different type of lens model...?


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