Hello raytracers,
I've occasionally come across object clipping in some parallel views
without setting any clipping plane. I've always assumed that the
camera for parallel views was at infinity and everything in the scene
was visible to it. I've included an example below that clips the
corner of a rectangle. Is there a workaround to this, say, a simple
conversion to a parallel view from very far away? Am I overlooking
something? I'm using Radiance 3.8, by the way.
Thanks in advance. Cheers,
--Dave
View file:------------------
rview -vtl -vp 33.4282 -12.544 6.49047 -vd -0.813537 0.363387 -0.45399
-vu -0.489383 -0.00613896 0.872047 -vh 100 -vv 100 -vo 0 -va 0 -vs 0
-vl 0
···
----------------------------
Rad file:-------------------
!gensky 4 1 12
void plastic white
0
0
5 1 1 1 0 0
white polygon plane
0
0
12
45 -15 0
45 15 0
-45 15 0
-45 -15 0
----------------------------
David Smith wrote:
Hello raytracers,
I've occasionally come across object clipping in some parallel views
without setting any clipping plane. I've always assumed that the
camera for parallel views was at infinity and everything in the scene
was visible to it. I've included an example below that clips the
corner of a rectangle.
your viewpoint is inside the box, which is why you're experiencing clipping even with no clipping plane specified. Just "back up" a little bit in your -vp specification, and you should be all good!
Hello raytracers,
I've occasionally come across object clipping in some parallel views
without setting any clipping plane. I've always assumed that the
camera for parallel views was at infinity and everything in the scene
was visible to it.
Everything in front of the camera, yes. The camera plane cuts through
the rectangle and only the parts in front of the plane are shown.
I've included an example below that clips the corner of a rectangle.
Is there a workaround to this, say, a simple conversion to a parallel
view from very far away?
The simple solution is to set the -o clipping plane to a negative value.
In your example '-o -8' is enough to show all of the rectangle.
For a view spec without clippings you have to add the inverse of your
view direction to your view point until everything is shown. Basically
you're moving the camera backwards but for a parallel view there is
no difference.
Regards,
Thomas
···
On 4 Nov 2008, at 16:27, David Smith wrote:
Rob and Thomas,
Ah, I was visualizing a sphere at infinity, centered on -vp with a
camera on the sphere pointed in the direction of -vd as opposed to a
plane within the model at -vp, with a normal of -vd that casts out
parallel rays. I misinterpreted the phrase "center of a parallel
projection" in the rpict man page for -vp. I guess the plane version
makes more sense when you start to throw clipping planes into the mix.
A negative -vo clipping plane an interesting concept, I hadn't
considered that.
I was looking at views from the sun, so I just subtracted a huge
multiple of the -vd vector from the -vp position and everything worked
out well.
Thank you very much,
--Dave