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

Re: Combining Projection Modes

Expand Messages
  • dmgalpha
    ... Outdoors panos, without straight lines are easy to projection into something that is believable. I will say that aeither of the two examples above you can
    Message 1 of 11 , Feb 24, 2007
      --- In PanoToolsNG@yahoogroups.com, "goetzgeorg" <goetzgeorg@...> wrote:
      > May be I should first be more precise in what I mean
      > with "combination".
      > What I don´t intend at all is a form of collage or montage, where you
      > have to blend differently rendered parts of an image. Seamless
      > blending this way would reqire finding traces of points that fall to
      > the same pixel coordinates. I´m afraid so for most projections those
      > possible blending lines seem to be restricted to horizons or middle
      > vertical axis. In days of film I used a similar approach in blending
      > in the unprojected left and right half of the first and last image.
      > As can be seen in this panos.
      > http://www.panhorizonte.de/Fruehsommer/slides/Klatschmohnfeld.html
      > http://www.panhorizonte.de/Fruehsommer/slides/Hitzegewitter-am-
      > Meissner.html

      Outdoors panos, without straight lines are easy to projection into
      something that is believable. I will say that aeither of the two
      examples above you can be properly projected with an architectural or
      a simply a lambert or miller.

      It is far more interesting and challenging to render an interior pano
      where most of the lines in it are orthogonal.

      > But there are also aesthetical and practical ones.

      That is my main thesis while exploring new projections. I totally agree.

      > What will be neither fish nor fowl for a navigatior, could be very
      > useful if you want to get a well-balanced pano.

      One of the main problems is photographs and paintings is that their
      perpective is dependent on where the viewer is. For example, if one
      could print a rectilinear of 180 degrees to, say 3 meters wide, the
      perception of its accuracy will change dramatically depending on how
      far the viewer is to it. If the viewes is close enough to the center
      of the pano, it might look natural.

      There are two fundamental problems to flat panos:

      1. You don't know where the viewer is going to be with respect to the
      image (which part are they going to be in front of --but you can give
      clues) Interestingly enough I visited an exhibit this week where the
      painting is meant to be seen--from a pespective point of view, from
      the side, that is, you stand slightly to the right of the frame and it
      becomes natural).

      2. you don't know how magnified or far the viewer is going to be from
      the image.

      > Here I think of a wide-angle, but no 180/360 pano, which shows a
      > specific pull so that it would not look too fishy.
      > It would be a pano which is neither extremely squeezed nor
      > compressed around the borders. And finally you can crop it to an
      > appropriate aspect ratio.
      > To answer aesthetical purposes it may be better to adjust a
      > projection that is e.g. 35% rectilinear, 35% equirectangular and 30%
      > cylindrical or whatever meets your needs.
      > Say you prefer more perspectiv pull. Then you could increase the
      > rectilinear percentage and the pixels would be more spread towards
      > the pure rectilinear projection.

      I don't think cylindrical is a good choice. It is not conformal, and
      tends to deform objects. Both equi and cyl are ok as long as you don't
      stray too far from the equator.

      > By the by:
      > Miller, Mercator, Peirce-Quincuncial etc. are welcome to play with.
      > May be soon we will see some of it also implemented in PTGUI.
      > (I´m looking forward to your Peirce Quincuncial-example.)
      > > In my opinion the most promising solution is to say: this is the
      > >most important area, please render it as rectilinear (which is the
      > >one most natural for people).

      You can already play with some of them using hugin. And if you compile
      tlalli you can already use Peirce.

      > > I also believe that after rectilinear conformal projections are the
      > > most promising. So perhaps what we want to combine is rectilinear in
      > > the centre and mercator (or miller as a compromise) around. But I
      > > still believe it will be very difficult to get a 360/180.
      > Would this tends to result in multi-view perspectives as a form of
      > collage-projections?
      > May be there is another way to go. Say you want to preserve such an
      > area. It would be possible to define this area as a set of image-
      > points, which will be rendered 100% rectilinear. Besides this an area
      > could be defined acting as a ramp, where the other coefficiets will
      > come into play gradually.

      this needs to be investigated. The problem is that you will disturb
      the viewer. Some are will be natural and another will be unnatural. In
      some areas straight lines will be straight and in others they will
      not. This is, in my opinion, worse than no straight lines at all, but
      this assertion needs to be tested with real people.

      That is why corners are a good place for a chance in projection, as
      the brain expects a shift from one area to the other.

      I'll upload some images later today.


      > Best regards,
      > Georg
    Your message has been successfully submitted and would be delivered to recipients shortly.