Bob Hughes wrote:
> 
> I'm still not getting this in my mind right.  Seems like a sort of
> perspective camera to me, like say you were wanting a rectangular
> region of 4x3 units at 0 z then double that at 1 z (8x6) and half that
> (2x1.5) at -1 z.... am I even close?

Not at all. the 4x3 stay always the same apparent size, at 0 z, -1 z or
1 z,
but using (u,v) coordinates for the rendered pictures, (v is up), i
think
the correlation is somethings like this:

 u = x + p.z
 v = y + q.z

so that xyz<0,0,0> is at say uv<0,0>
xyz<2,0,0> is at uv<2,0>
but xyz<2,0,2> is at uv<3,1>

whereas the orthographic camera would have given for xyz<2,0,2> the
uv<2,0>

In fact, there is no perspective points or lines: all the rays are
really parallele
(thus the apparent size do not change with the z, just like the
orthographic camera),
but instead of a full face compression, you have a partial look of the
depth.

> Someone is bound to have a better grasp on your concept than I so
> don't despair : ) just yet anyway.
> To let you know what I'm thinking this would be like I'd have to say
> it's a slice of 3D turned to one side a bit.  If so why not use
> orthographic on a rotated section of a model?  
Well that's may be the idea, 
The <x,y,0> should really be directly seen as <u=x,v=y>, keeping the
lengths
and the right angle.

In fact, It seems I need to have a direction which is not perpendicular
to
the image right and up. 

Eh! that's it. I just remove the look_at and it works !!!!
(ok, with a warning about the non perpendicular, but that's what I
want!)

my camera is (for full cabinet)
camera
{ orthographic
  location  <40.0, 40.5, -40.0>
  direction <-1,-1,1>
  up        5*y
  right     5*4/3*x
  //look_at   <0.0, 0.0,  0.0>
}

I will post a rendered image in p.b.i soon

>I may be way off about
> this, I know, so sorry if I'm confused.
> 
> PS  see p.b.i. for J. Grimberts demo gif.
> 
> Bob
>

Post a reply to this message