from [povray.general]
Am 20.03.2017 um 11:38 schrieb muyu:

> Thanks. 1) Regarding my simulations, around 50% of the light is reflected in
> near-infrared. So physically the expected error should be much less than 1%.
> This should be neglible?

> 2) However, I found almost 10%  under-estimates compared with the reference
> using Monte Carlo method. What's the possible reasons for this? Should I
> increase the brightness or the light intensity?

No ideas at the moment. Maybe if you posted the resulting images, the
distribution of light within the canopy might give a hint.

You mentioned reflection above; did you also account for the transmitted
light?

Here it is the setting of the radiosity:
#version 3.7;
global_settings {
   assumed_gamma 1.0
   ambient_light 0.0
   adc_bailout 0.0
   max_trace_level 20

   radiosity {
      count 1500
      error_bound 0.01
      recursion_limit 20
      nearest_count 10
   }
}

// light--------------------------------------------------------------
light_source{<0,36.397,100> color rgb<0,1,0> parallel}

// ground ------------------------------------------------------------
plane { <0,0,1>, 0
        texture{ pigment{color rgb<0, 1, 0>}
                 finish {ambient 0 diffuse 0.1590 0.0000}
               } // end of texture
      } // end of plane

// leaves----------------------------------------------
#declare leaf_text = texture {
pigment {color rgb<0.0,1.0,0.0>}
finish {ambient 0 diffuse 0.4957 0.4409}}
disc {<9.5780,-10.5050,0.6040>, <0.7070,0.1830,0.6840>, 0.0500
texture {leaf_text}}
.....

diffuse 0.4957 is front diffuse, 0.4409 is the transmitted light? Should I set
rgbt or rgbf?

Is this the right way to consider the multi-scattering using Radiosity? I have
the impression that some processes are missing?

Thax in advance.

Post a reply to this message

Attachments:
Download 'canopy_2_camera.png' (651 KB)

Preview of image 'canopy_2_camera.png'

"muyu" <lsy### [at] gmailcom> wrote:
> from [povray.general]
> Am 20.03.2017 um 11:38 schrieb muyu:
>
> > Thanks. 1) Regarding my simulations, around 50% of the light is reflected in
> > near-infrared. So physically the expected error should be much less than 1%.
> > This should be neglible?
>
> > 2) However, I found almost 10%  under-estimates compared with the reference
> > using Monte Carlo method. What's the possible reasons for this? Should I
> > increase the brightness or the light intensity?
>
> No ideas at the moment. Maybe if you posted the resulting images, the
> distribution of light within the canopy might give a hint.
>
> You mentioned reflection above; did you also account for the transmitted
> light?
>
> Here it is the setting of the radiosity:
> #version 3.7;
> global_settings {
>    assumed_gamma 1.0
>    ambient_light 0.0
>    adc_bailout 0.0
>    max_trace_level 20
>
>    radiosity {
>       count 1500
>       error_bound 0.01
>       recursion_limit 20
>       nearest_count 10
>    }
> }
>
> // light--------------------------------------------------------------
> light_source{<0,36.397,100> color rgb<0,1,0> parallel}
>
> // ground ------------------------------------------------------------
> plane { <0,0,1>, 0
>         texture{ pigment{color rgb<0, 1, 0>}
>                  finish {ambient 0 diffuse 0.1590 0.0000}
>                } // end of texture
>       } // end of plane
>
> // leaves----------------------------------------------
> #declare leaf_text = texture {
> pigment {color rgb<0.0,1.0,0.0>}
> finish {ambient 0 diffuse 0.4957 0.4409}}
> disc {<9.5780,-10.5050,0.6040>, <0.7070,0.1830,0.6840>, 0.0500
> texture {leaf_text}}
> .....
>
> diffuse 0.4957 is front diffuse, 0.4409 is the transmitted light? Should I set
> rgbt or rgbf?
>
> Is this the right way to consider the multi-scattering using Radiosity? I have
> the impression that some processes are missing?
>
> Thax in advance.

Antialias = on
Antialias_Threshold = 0.1
Antialias_Depth = 2

Input_File_Name = "canopy_2_camera.pov"

width = 1000
height = 1000

Cyclic_Animation = on
Pause_when_Done = off

The reflectance of each pixel was calculated relative to the reference panel
with the same light source and camera setup
// the reference panel--------------------------------------------------------
plane { <0,0,1>, 0
        texture{ pigment{color rgb<0, 1, 0>}
                 finish {ambient 0 diffuse 1.0 0.0 conserve_energy}
               } // end of texture
      } // end of plane

Reflectance = (Simulation/Reference)^2.2

Is this calibration right? Thanks again.

Post a reply to this message

Attachments:
Download 'ref_plane.png' (6 KB)

Preview of image 'ref_plane.png'

On 03/20/2017 10:30 AM, muyu wrote:
>
> No ideas at the moment. Maybe if you posted the resulting images, the
> distribution of light within the canopy might give a hint.
>
> Here it is the setting of the radiosity:
> #version 3.7;
 >.....
> disc {<9.5780,-10.5050,0.6040>, <0.7070,0.1830,0.6840>, 0.0500
> texture {leaf_text}}
> .....
>
> diffuse 0.4957 is front diffuse, 0.4409 is the transmitted light? Should I set
> rgbt or rgbf?
>
> Is this the right way to consider the multi-scattering using Radiosity? I have
> the impression that some processes are missing?
>
> Thax in advance.
>
Not been following closely, but on seeing some images and code some a 
few thoughts pop into my head.

Just a thought, but in play here is perhaps that the 'disc' object has 
no thickness. This means rays which are nearly parallel with the plane 
of a disc will resolve somewhat noisily (more light transport than 
should be I suppose) and also that light will not attenuate/scatter in 
the leaf's thickness(1).

Perhaps give us a short description of your Monte Carlo simulation 
"standard." Is this also done in POV-Ray? A stand along program or?

Aside: I thought the second diffuse parameter required a proceeding 
comma, but if working OK for you must not be the case... :-)

Bill P.

(1) - More accurate leaves could be modeled in POV-Ray, but suspect it 
would be costly in terms of complexity and compute resource.

Post a reply to this message

Am 21.03.2017 um 15:12 schrieb William F Pokorny:

> Aside: I thought the second diffuse parameter required a proceeding
> comma, but if working OK for you must not be the case... :-)

All commas in POV-Ray's scene description language are optional --
unless the second expression has a leading `+` or `-` sign, which would
be interpreted as a mathematical operation without a comma; it is
therefore a good habit to make liberal use of commas.

Post a reply to this message

clipka <ano### [at] anonymousorg> wrote:
> Am 21.03.2017 um 15:12 schrieb William F Pokorny:
>
> > Aside: I thought the second diffuse parameter required a proceeding
> > comma, but if working OK for you must not be the case... :-)
>
> All commas in POV-Ray's scene description language are optional --
> unless the second expression has a leading `+` or `-` sign, which would
> be interpreted as a mathematical operation without a comma; it is
> therefore a good habit to make liberal use of commas.

Thanks. Do you have any suggestions to improve the accuracy of radiosity
simulation? The back side diffuse is well taken into account in Pov-ray? Should
I improve brightness or light intensity?

Post a reply to this message

William F Pokorny <ano### [at] anonymousorg> wrote:
> On 03/20/2017 10:30 AM, muyu wrote:
> >
> > No ideas at the moment. Maybe if you posted the resulting images, the
> > distribution of light within the canopy might give a hint.
> >
> > Here it is the setting of the radiosity:
> > #version 3.7;
>  >.....
> > disc {<9.5780,-10.5050,0.6040>, <0.7070,0.1830,0.6840>, 0.0500
> > texture {leaf_text}}
> > .....
> >
> > diffuse 0.4957 is front diffuse, 0.4409 is the transmitted light? Should I set
> > rgbt or rgbf?
> >
> > Is this the right way to consider the multi-scattering using Radiosity? I have
> > the impression that some processes are missing?
> >
> > Thax in advance.
> >
> Not been following closely, but on seeing some images and code some a
> few thoughts pop into my head.
>
> Just a thought, but in play here is perhaps that the 'disc' object has
> no thickness. This means rays which are nearly parallel with the plane
> of a disc will resolve somewhat noisily (more light transport than
> should be I suppose) and also that light will not attenuate/scatter in
> the leaf's thickness(1).
>
> Perhaps give us a short description of your Monte Carlo simulation
> "standard." Is this also done in POV-Ray? A stand along program or?
>
> Aside: I thought the second diffuse parameter required a proceeding
> comma, but if working OK for you must not be the case... :-)
>
> Bill P.
>
> (1) - More accurate leaves could be modeled in POV-Ray, but suspect it
> would be costly in terms of complexity and compute resource.

Thanks. The Monte Carlo simulation was generated by other softwares. I did not
know much details about it. But they were considered as the reference to
evaluate other methods.

1) Disc of negligible thickness...the same as in the Monte Carlo methods.

2) Now my simulations using Pov-ray radiosity were around 5 to 10% lower than
the reference even I set the recursion time as 20. I am wondering if Pov-ray
miss some physical processes or I can further improve the accuracy by tunning
some parameters. Brightness or light intensity etc...?

Thanks in advance.

Shouyang

Post a reply to this message

On 03/23/2017 05:54 AM, muyu wrote:
> William F Pokorny <ano### [at] anonymousorg> wrote:
>> On 03/20/2017 10:30 AM, muyu wrote:
>
> Thanks. The Monte Carlo simulation was generated by other softwares. I did not
> know much details about it. But they were considered as the reference to
> evaluate other methods.
>
> 1) Disc of negligible thickness...the same as in the Monte Carlo methods.
>
> 2) Now my simulations using Pov-ray radiosity were around 5 to 10% lower than
> the reference even I set the recursion time as 20. I am wondering if Pov-ray
> miss some physical processes or I can further improve the accuracy by tunning
> some parameters. Brightness or light intensity etc...?
>
> Thanks in advance.
>
> Shouyang
>
>
>

Christoph (clipka) & others understand the light and color modeling in 
POV-Ray far better than me. I know for example there are various diffuse 
models, but not in detail how they differ. See for example :

http://news.povray.org/povray.unofficial.patches/thread/%3C57be5290%241%40news.povray.org%3E/

for a pointer to an alternate version of POV-Ray (uber) offering 
additional methods.

Without knowing more about the standard simulation against which you are 
comparing, we're of course doing a good deal of guessing as to what 
might be different.

...

How much does POV-Ray itself vary run to run on changing the (random 
within a container?) disc placements? Is this difference reflected by 
the 5-10% range you mention or is this change due changes in other 
"leaf" parameters?

Another thought. With your POV-Ray disc placement are you avoiding 
collisions? If not, might the standard method be?

If I think of something else, I'll post.

Bill P.

Post a reply to this message

Am 20.03.2017 um 15:53 schrieb muyu:
> "muyu" <lsy### [at] gmailcom> wrote:
>> from [povray.general]
>> Am 20.03.2017 um 11:38 schrieb muyu:
>>
>>> Thanks. 1) Regarding my simulations, around 50% of the light is reflected in
>>> near-infrared. So physically the expected error should be much less than 1%.
>>> This should be neglible?
...
>> You mentioned reflection above; did you also account for the transmitted
>> light?
...
>> finish {ambient 0 diffuse 0.4957 0.4409}}

That's 0.9366 in total, which means that at each radiosity recursion
step, 93.66% of the light is /not/ absorbed.

Which means that after 20 recursion steps, 26.98% of the light is
/still/ not absorbed.

>> disc {<9.5780,-10.5050,0.6040>, <0.7070,0.1830,0.6840>, 0.0500
>> texture {leaf_text}}
>> .....
>>
>> diffuse 0.4957 is front diffuse, 0.4409 is the transmitted light? Should I set
>> rgbt or rgbf?

Don't use `rgbt` nor `rgbf`, unless your similated leaves are supposed
to transmit a portion of the light unscattered.

> The reflectance of each pixel was calculated relative to the reference panel
> with the same light source and camera setup
> // the reference panel--------------------------------------------------------
> plane { <0,0,1>, 0
>         texture{ pigment{color rgb<0, 1, 0>}
>                  finish {ambient 0 diffuse 1.0 0.0 conserve_energy}
>                } // end of texture
>       } // end of plane
> 
> Reflectance = (Simulation/Reference)^2.2

Are you using File_Gamma=2.2, or why is that exponent in there?

Post a reply to this message

clipka <ano### [at] anonymousorg> wrote:
> Am 20.03.2017 um 15:53 schrieb muyu:
> > "muyu" <lsy### [at] gmailcom> wrote:
> >> from [povray.general]
> >> Am 20.03.2017 um 11:38 schrieb muyu:
> >>
> >>> Thanks. 1) Regarding my simulations, around 50% of the light is reflected in
> >>> near-infrared. So physically the expected error should be much less than 1%.
> >>> This should be neglible?
> ...
> >> You mentioned reflection above; did you also account for the transmitted
> >> light?
> ...
> >> finish {ambient 0 diffuse 0.4957 0.4409}}
>
> That's 0.9366 in total, which means that at each radiosity recursion
> step, 93.66% of the light is /not/ absorbed.
>
> Which means that after 20 recursion steps, 26.98% of the light is
> /still/ not absorbed.
> >>
> >> diffuse 0.4957 is front diffuse, 0.4409 is the transmitted light? Should I set
> >> rgbt or rgbf?
>
> Don't use `rgbt` nor `rgbf`, unless your similated leaves are supposed
> to transmit a portion of the light unscattered.

For each leaf, 0.4957 is scatted directly. 0.4409 is transmitted and scatted. So
as I understood, I should not use any filter. The two sides diffuse is the right
choice. Am I right? So it means 20 times of recursion is not enough? Is it
possible to set larger value.Now for me, the rendering time is still
acceptable.

>
> > The reflectance of each pixel was calculated relative to the reference panel
> > with the same light source and camera setup
> > // the reference panel--------------------------------------------------------
> > plane { <0,0,1>, 0
> >         texture{ pigment{color rgb<0, 1, 0>}
> >                  finish {ambient 0 diffuse 1.0 0.0 conserve_energy}
> >                } // end of texture
> >       } // end of plane
> >
> > Reflectance = (Simulation/Reference)^2.2
>
> Are you using File_Gamma=2.2, or why is that exponent in there?

Here I did not set File_Gamma = 1.0. But the result should be same if I use this
Reflectance = (Simulation/Reference)^2.2.

Thanks again.

Shouyang

Post a reply to this message

Am 24.03.2017 um 00:14 schrieb muyu:

> For each leaf, 0.4957 is scatted directly. 0.4409 is transmitted and scatted. So
> as I understood, I should not use any filter. The two sides diffuse is the right
> choice. Am I right?

Yes, you are.

> So it means 20 times of recursion is not enough? Is it
> possible to set larger value.Now for me, the rendering time is still
> acceptable.

If building your own modified POV-Ray binaries is an option, you could
edit the definition of DEPTH_MAX in `source/core/lighting/radiosity.h`.


>>> The reflectance of each pixel was calculated relative to the reference panel
>>> with the same light source and camera setup
>>> // the reference panel--------------------------------------------------------
>>> plane { <0,0,1>, 0
>>>         texture{ pigment{color rgb<0, 1, 0>}
>>>                  finish {ambient 0 diffuse 1.0 0.0 conserve_energy}
>>>                } // end of texture
>>>       } // end of plane
>>>
>>> Reflectance = (Simulation/Reference)^2.2
>>
>> Are you using File_Gamma=2.2, or why is that exponent in there?
> 
> Here I did not set File_Gamma = 1.0. But the result should be same if I use this
> Reflectance = (Simulation/Reference)^2.2.

Note that for most output file formats (including JPEG and PNG) POV-Ray
3.7 uses the sRGB encoding transfer function by default (and for those
where it doesn't, it defaults to linear output or the ITU-R BT.709
transfer function). For the power-law with an exponent of 2.2 to give
correct results, you should use `File_Gamma=2.2`. (Or instead apply the
sRGB decoding transfer function to sim and reference separately, and
then divide.)

Also, make sure the software you are using to compute the reflectance
does indeed operate on gamma-encoded values.

Post a reply to this message