Hi Pov Ray Team,

I would like to know if you plan to implement another rendering technique
outside
you raytracing and GI solution?
My personal favourite would be MLT or Bidirectional Pathracing, it would
give Pov Ray the cutting edge :-D

Thor

Post a reply to this message

McHannemann schrieb:
> Hi Pov Ray Team,
> 
> I would like to know if you plan to implement another rendering technique
> outside
> you raytracing and GI solution?

There are no intentions for adding techniques that are incompatible to 
raytracing like for example methods that require the objects to be 
represented as meshes.

> My personal favourite would be MLT or Bidirectional Pathracing, it would
> give Pov Ray the cutting edge :-D

Suggestions like that are usually not leading to much, especially if you 
  cannot give anything to support the claim that technique X would be 
useful (i.e. efficient and effective) and feasible to implement for POV-Ray.

-- Christoph

Post a reply to this message

McHannemann wrote:
> Hi Pov Ray Team,
> 
> I would like to know if you plan to implement another rendering technique
> outside
> you raytracing and GI solution?
> My personal favourite would be MLT or Bidirectional Pathracing, it would
> give Pov Ray the cutting edge :-D
> 
> Thor
> 
> 

 From a quick look at Wikipedia, it seems that POV-Ray almost implements 
bidirectional pathtracing as it is.

First of all, POV is already a forward raytracer.
Second, it already supports radiosity (monte carlo global illumination, 
I believe the technical term is).

So what would it really add?

...Chambers

Post a reply to this message

Ben Chambers schrieb:
> 
>  From a quick look at Wikipedia, it seems that POV-Ray almost implements 
> bidirectional pathtracing as it is.

No, POV-Ray radiosity only traces rays in direction camera -> light 
source and not the other way round.  Also POV-Ray uses an irradiance 
cache to speed up computations and to avoid high frequency noise.  This 
only works reasonably well for diffuse light transport.  So path tracing 
would allow rendering some lighting situations that can't be 
realistically done in POV-Ray right now but OTOH it would be much less 
efficient for most cases which can be handled by POV-Ray radiosity.

-- Christoph

Post a reply to this message

Ben Chambers <ben### [at] pacificwebguycom> wrote:
> First of all, POV is already a forward raytracer.

  The most common agreement is that the "forward" and "backward" terms
refer to the direction in which light travels.

  In other words, a "forward raytracer" is a raytracer which calculates
how light traverses from light sources towards the scene, and from there
towards the camera.

  A "backward raytracer" does the opposite: It calculates in the opposite
direction how light travels, ie. it starts from the camera towards the
scene, and from there to the light sources. This is how POV-Ray (and
almost all raytracers) work.

  POV-Ray's stochastic global illumination algorithm is still "backwards
raytracing" because it traces from surfaces to other surfaces and from
there to the light sources (ie. it still traces back how light traverses).
POV-Ray's photon mapping algorithm is a genuine "forward raytracing"
feature.

  Of course there are some books which mix these terms, causing confusion.

-- 
                                                          - Warp

Post a reply to this message

Warp <war### [at] tagpovrayorg> wrote:
> Ben Chambers <ben### [at] pacificwebguycom> wrote:
> > First of all, POV is already a forward raytracer.
>
>   The most common agreement is that the "forward" and "backward" terms
> refer to the direction in which light travels.
>
>   In other words, a "forward raytracer" is a raytracer which calculates
> how light traverses from light sources towards the scene, and from there
> towards the camera.
>
>   A "backward raytracer" does the opposite: It calculates in the opposite
> direction how light travels, ie. it starts from the camera towards the
> scene, and from there to the light sources. This is how POV-Ray (and
> almost all raytracers) work.

Hmm, I've normally seen it where "raytracer" implies from-the-camera-out
(camera -> scene -> light source), and where "backwards raytracer" is used
to mean light source -> scene -> camera.
Charles

Post a reply to this message

Warp wrote:
>   POV-Ray's stochastic global illumination algorithm is still "backwards
> raytracing" because it traces from surfaces to other surfaces and from
> there to the light sources (ie. it still traces back how light traverses).
> POV-Ray's photon mapping algorithm is a genuine "forward raytracing"
> feature.

Duh, I know that :)

Actually, this brings up one point I've run into in the past: since 
POV-Ray computes radiosity samples by tracing rays from the camera, and 
then lighting them, I've run into some issues in scenes lit entirely by 
radiosity that don't light up the way they should (for instance, if a 
large object with high ambience is partially (or entirely) out of view).

I suppose that building a radiosity cache, by doing "forward tracing" 
(from the light source) pretrace, would be beneficial in such scenes. 
However, I can't imagine its very high on the priority list.

...Chambers

Post a reply to this message

What Ben mentioned is what I had as a problem,
scenes with rely heavy on lighting objects and interiors lit from outside
and a lamp have sever problems, and I could just calculate in another
renderer correct where pov ray left most of the areas completly black, you
can have a look at kerkythea as example they have even comparsion shots
between the render technologies.

and... I didn`t want to sound offensive christopher, I just wanted to know
if you think about implementing additional methods...
and if you like to get more information on it I would love to help and
supply what I can get my fingers on!

Thorsten

Post a reply to this message

> and... I didn`t want to sound offensive christopher, I just wanted to know
> if you think about implementing additional methods...
> and if you like to get more information on it I would love to help and
> supply what I can get my fingers on!

Since you can save_file and load_file radiosity data, you
could place the camera at the major light sources during
several pre-renders to flesh out the radiosity data more.
This should result in the same results as bidirectional path
tracing.
I did a test with 2 lights and a sphere, there are differences
but very slight. I'll put the differenced picture up in p.b.i
(it's tweaked to make the difference be visible)

Post a reply to this message

McHannemann schrieb:
> What Ben mentioned is what I had as a problem,
> scenes with rely heavy on lighting objects and interiors lit from outside
> and a lamp have sever problems, and I could just calculate in another
> renderer correct where pov ray left most of the areas completly black, you
> can have a look at kerkythea as example they have even comparsion shots
> between the render technologies.
> 
> and... I didn`t want to sound offensive christopher, 

This wasn't regarded as offence, but as i said we all know about the 
techniques that exist to approach the global illumination problem. 
Therefore mentioning some nice sounding names does not lead to much.

> I just wanted to know
> if you think about implementing additional methods...
> and if you like to get more information on it I would love to help and
> supply what I can get my fingers on!

As also said if you'd come up with some practical examples (i.e. 
demonstrating usability and efficiency) of global illumination 
techniques in comparison with POV-Ray's current system that would be 
great.  But you should take care this is a valid comparison (i.e. no 
mesh only scene compared to a complex CSG in POV-Ray) and giving render 
times is essential of course.  Gilles made such a comparison for C4D 
some time ago (the result was that the techniques seem very similar in 
results although there are also differences).

-- Christoph

Post a reply to this message