I would like to use several different cameras in the same scene, where cameras
are associated to specific objects. The effect  I have in mind can be described
as follows:

Suppose your scene contains two towers in front of the camera, one to the left,
the other to the right. I would like them to render so that both towers appear
as if the camera was directly in front of each them both. This will result in a
slightly unrealistic perspective. Note that rotating the towers towards the
camera does not have the same
effect.

One can achieve this by hand by rendering two images of the individual towers,
each with the camera directly in front of the respective tower, and the rest of
the scene sent to alpha,
and  then blend the two images with the scene without the towers.

If one could associate a camera to an object so that rays that hit thet object
are
recorded by the associated camera, this would simplify the process, in
particular in complex scenes.

Matthias

Post a reply to this message

Am 07.12.2018 um 23:18 schrieb Matthias:

> If one could associate a camera to an object so that rays that hit thet object
> are
> recorded by the associated camera, this would simplify the process, in
> particular in complex scenes.

There's a conceptual problem with this: It is the camera that defines 
where a ray originates and in which direction it goes. So you have to 
associate a camera /before/ you test which object(s) the ray hits.

Post a reply to this message

clipka <ano### [at] anonymousorg> wrote:
> Am 07.12.2018 um 23:18 schrieb Matthias:

> There's a conceptual problem with this: It is the camera that defines
> where a ray originates and in which direction it goes. So you have to
> associate a camera /before/ you test which object(s) the ray hits.

Hi Matthias,

I haven't played around much with the different types of cameras, but perhaps
you could do on or both of two things.

I know that people have been playing around with stereograms and anaglyphs -
since those "overlay two slightly different views", then perhaps that's a way to
do what you want - only with two very different views.

You may also want to investigate the mesh camera and user-defined camera in
POV-Ray.

Also, I recall that with hgPOVRay, Jerome Grimbert (LeForgeron) and others were
experimenting with multiple camera views in a single render - it might have been
MegaPOV or UberPOV....

Perhaps check out one of those forks and see if that gets you what you need.

Post a reply to this message

Am 08.12.2018 um 14:36 schrieb Bald Eagle:

> Also, I recall that with hgPOVRay, Jerome Grimbert (LeForgeron) and others were
> experimenting with multiple camera views in a single render - it might have been
> MegaPOV or UberPOV....

I don't think MegaPOV had such a thing - though it had a "camera 
pigment" that could be used to such effect.

UberPOV definitely doesn't have such a thing.

Post a reply to this message

in news:web.5c0bc8e573c42217765e06870@news.povray.org Bald Eagle wrote:

> stereograms and anaglyphs

http://paulbourke.net/stereographics/povcameras/

ingo

Post a reply to this message

Le 08/12/2018 à 14:36, Bald Eagle a écrit :
> Hi Matthias,
> 
> I haven't played around much with the different types of cameras, but perhaps
> you could do on or both of two things.
> 
> I know that people have been playing around with stereograms and anaglyphs -
> since those "overlay two slightly different views", then perhaps that's a way to
> do what you want - only with two very different views.
> 
> You may also want to investigate the mesh camera and user-defined camera in
> POV-Ray.
> 
> Also, I recall that with hgPOVRay, Jerome Grimbert (LeForgeron) and others were
> experimenting with multiple camera views in a single render - it might have been
> MegaPOV or UberPOV....
> 

From http://wiki.povray.org/content/User:Le_Forgeron/cameras
you might be interested in:
* stereo (side by side), bottom of page
* omni_directional_stereo (ODS, full sphere view, twice, on top of each
other)
* grid (as many camera as you want, and for each whatever you want), top
of page

The two first ones assumes a pair of eyes, horizontally spaced (along
right vector), and then allow some tuning of their parameters.

The grid replaces the picture to produce with a, wait for it, grid of
cameras. The split of the picture is regular, but can be different for
each direction (e.g. 3x1, 2x2 and 1x3 are possible). Each element of the
grid get its own camera identifier. And recursion is allowed, you can
have a grid in a grid... and so on.

Post a reply to this message