[ptx] RFC: "tube-panorama"

[Pablo d'Angelo]

Hallo Sebastian,

Sebastian Nowozin schrieb am Freitag, den 05. März 2004:

> 
> my brother and me travelled some interesting road in Shanghai, that would make
> a very nice picture as very long linear picture. Like photographing each side
> of the road, complete as one very long picture. We thought and discussed a bit
> about it and then end up at a different kind of panorama image, we called
> "tube".
> 
> Here is some food for thought and questions how doable such kind of panorama
> is... please discuss/reject/approve it freely :)
> 
> 
> 
> Three-dimensional "tube-panorama"
> 
> The idea is this: Instead of having a single point around which a sphere with
> image information is constructed, one could have a "line", around which a
> "tube" is constructed. The ends of the tube are half spheres. So, with a
> viewer programm, one could do the usual viewing operations, plus travel on
> this "line" forth and back.
> 
> The idea sounds neat, but there are some problems there. (obviously not
> complete)
> 
> 
> 1. Recording the pictures

Lets put that issue aside for a second. Probably it is easiest to render
such a tube panorama with povray or another raytracer, to see how it looks,
before writing the programs needed to create the tube.

> 2. Constructing the panorama
> 
> There would have to be a new projection type, that would have half-spheres on
> both ends of the tube, and a cylinder/tube between the spheres. Besides that,
> I see no obvious problems. One would have to mark where the tube starts and
> ends, or this could maybe detected automatically. Each image would have to be
> assigned a new coordinate: the position on the line (maybe just 0.0 to 1.0,
> and maybe also automatically doable).

Ok, lets see an simple birdseye example:

       A-_                    [object A]
       |  \__
        \    \__
       B |      \___          [object B]
       |\|------\__ \
-------x-y-----------z----    [ tube line]

This illustrates (hmm, well... ;) the problem of this method when the real
world is not a tube.

A and B are some object, x,y,z are viewing positions along the tube line.
The ugly lines represent viewing lines.


I suppose the whole idea is that not only the "ring" take at a viewing
positions are to be shown.

Then, a viewer at x only sees object B, while the viewer at y sees B and a
part of A, and the viewer at z sees a large distance between A and B.


       A
       
       
       B
       |
-------x-y-----------z----    [ tube line]


       A                      [object A]
       |     
        \       
       B |                    [object B]
        \|           
-------x-y-----------z----    [ tube line]



       A-_                    [object A]
          \__
             \__
       B-___    \___          [object B]
            ---_____\
-------x-y-----------z----    [ tube line]


The problem is that the only correct information at a position is the "ring"
normal to the tube line. this information is not "strictly" valid for all
other viewing positions, and it will look very strange I to watch the scene
at z, but see B a bit like it would be seen at x.

For a longer line, one really needs a full 3D modelling. maybe multiple
tubes with different distances might work for some scenarios, a bit like in
the old jump-n-run game backgrounds.

Essentially, to look around at every point of the the line, one needs a full
pano at reach line point. The general problem to solve would be 3D
reconstrution. 

See http://www.esat.kuleuven.ac.be/~pollefey/ and especially:
http://www.esat.kuleuven.ac.be/~pollefey/SMILE2/tutorial.html
for some information. But this is a very hard problem.

> Unfortunately it would require some slight modifications to the viewing
> software to account and allow for sliding on the one-dimensional line.
> Questions: Would the viewer recognize its a "tube" immediately? Can it look as
> realistic as a normal spherical panorama does? How to linearly store the image
> as .JPG to be able to view it statically, without viewer, too? (maybe as four
> lines: two half-spheres, and two rows for the tube in length, each 180 degree)

The linear panorama will look very good without a viewer I think. So far, I
have only seen partial version of these, and not a full 360° deg one. 

ciao
  Pablo