Looking at the mesh_camera ... LOVE it, my dreams for novel cameras is realised!
However it seems that creating a mesh for reasonable resolution images is
problematic, rendering a 2Kx1K image, so the corresponding mesh is also 2Kx1K
and the parsing takes a long time. Am I missing something? Is there a way to
interpolate higher resolution images from lower resolution meshes?

Post a reply to this message

On 10/27/18 7:59 AM, Paul Bourke wrote:
> Looking at the mesh_camera ... LOVE it, my dreams for novel cameras is realised!
> However it seems that creating a mesh for reasonable resolution images is
> problematic, rendering a 2Kx1K image, so the corresponding mesh is also 2Kx1K
> and the parsing takes a long time. Am I missing something? Is there a way to
> interpolate higher resolution images from lower resolution meshes?
> 
> 
See the thread:

http://news.povray.org/povray.general/thread/%3C56d72b90%241%40news.povray.org%3E/

in which I posted some code taking advantage of one of the mesh camera's 
ability to work by mesh row - considerably speeding up parsing at the 
cost of fixed output orientation. Other than that I am not aware of any 
ability to speed things up or to interpolate larger images from smaller 
meshes. My experience with the mesh camera code is limited though so 
perhaps someone knows otherwise.

FYI. If you are running a 3.8 alpha release, you might be interested in 
the user_defined camera which uses functions.

http://wiki.povray.org/content/Reference:Camera#User_defined_projection

It too can be a little slow - in the render phase - compared to hard 
coded cameras. Especially true as the controlling functions get more 
complex. Here I got part way on a method storing doubles in df3 files 
with the thought of implementing simpler look up functions for stored 
vectors for different cameras. Stopped as the functions converting from 
the read df3 to doubles for vector values were themselves kind of slow. 
I thought it unlikely the result would be much faster than functions I 
was using in the user_defined camera. The method might though be faster 
than parsing for large mesh cameras in your aim is to create a bunch of 
complex cameras stored as values/meshes.

You might too check out Le Forgeron (Jérôme's) version of POV-Ray which 
has some additional cameras hard coded. See:

http://wiki.povray.org/content/User:Le_Forgeron

If working with ODS stuff in particular the following howto from the 
wiki related to the thread at top may be of interest:

http://wiki.povray.org/content/HowTo:ODS

Good to see you posting here again! I've certainly used a lot of 
information off your web sites over the years.

Bill P.

Post a reply to this message

Thanks for the pointers.
Not for ODS, but rather "inverse panoramas", some examples here
    http://paulbourke.net/miscellaneous/inversepano/
Multiplicity of viewpoints on the exterior of an object. The ones there
currently are cylindrical, doing spherical (equirectangular) now.

Post a reply to this message

Am 28.10.2018 um 05:23 schrieb Paul Bourke:
> Thanks for the pointers.
> Not for ODS, but rather "inverse panoramas", some examples here
>      http://paulbourke.net/miscellaneous/inversepano/
> Multiplicity of viewpoints on the exterior of an object. The ones there
> currently are cylindrical, doing spherical (equirectangular) now.

That's some fascinating work.

I'd certainly recommend `user_defined` camera for this purpose, as you 
"only" need to plug in formulas, don't have to wrap your brain around 
making a corresponding mesh, and don't get excessive memory requirements 
at high resolutions.

As long as the functions don't do anything particularly fancy, I don't 
expect render speed to be a problem. At the very least it should beat 
the mesh camera hands down anytime.

Post a reply to this message

Le 18-10-28 à 00:23, Paul Bourke a écrit :
> Thanks for the pointers.
> Not for ODS, but rather "inverse panoramas", some examples here
>      http://paulbourke.net/miscellaneous/inversepano/
> Multiplicity of viewpoints on the exterior of an object. The ones there
> currently are cylindrical, doing spherical (equirectangular) now.
> 
> 
> 

Maybe it could be obtained using one of the cylindrical cameras.
Set your object with an interior_tecture{rgbt 1} and surround it with a 
larger cylindrical mirror matching the cylindrical camera.

Did not try, but could be worth a few tests.

Post a reply to this message

These renders are truly amazing!

Where did you get those 3D models of those humans and other items?

---
Diese E-Mail wurde von AVG auf Viren geprüft.
http://www.avg.com

Post a reply to this message

On 29.10.2018 2:32, Alain wrote:

> Maybe it could be obtained using one of the cylindrical cameras.
> Set your object with an interior_tecture{rgbt 1} and surround it with a 
> larger cylindrical mirror matching the cylindrical camera.

no_image keyword might also be helpful to only see the reflection.

Post a reply to this message