Images too bright with gendaylit

Hello,

when I use gendaylit to create a sun my images are too bright.

The command used is

gendaylit -ang 90 0 -W 900 200

void light solar
0
3 7.576e+006 7.576e+006 7.576e+006

solar source sun
0
4 -0.000000 -0.052336 0.998630 0.533000

This will put the sun to the zenith with reasonable values
for direct normal and diffuse irradiance in W/m^2.

Correct so far?

I removed the skyfunc, just the sun source is active.

I put together a simple scene:
- plastic cylinders as coordinate origin,
- large brown ground plate
- dielectric spheres (I just love those :slight_smile:

Here is an image with renderings of different sun settings:

http://imgur.com/hOqCpmY

When I use the sun as stated above the image on the right
side is rendered. Much too bright.

Reducing the parameters of void light solar:

middle image (7.576e+005 7.576e+005 7.576e+005 )
left image (7.576e+004 7.576e+004 7.576e+004)

Viewed with default parameters of rvu.

Do you have an idea what I am doing wrong
or might have missed?

Kind regards,
Joe

Hi Joe,

why do you think it is too bright? Is the luminance to high? I guess you just have to adapt your exposure setting for the image.

The physics for gendaylit and your parameters is correct:

1. test of luminance of the sun (ray in sun direction):

gendaylit -ang 90 0 -W 900 200 >sky.rad
oconv -f sky.rad >sky.oct

echo 0 0 0 -0.000000 -0.052336 0.998630 |rtrace -h sky.oct | rcalc -e '$1=47.4*$1+120*$2+11.6*$3'
result: 1.356104e+09 cd/m2

This value looks reasonable - the max is about 1.6 e9 cd/m2

2. test of irradiance (sensor also looks towards sun):

gendaylit -ang 90 0 -O 1 -W 900 200 >sky_irr.rad
oconv -f sky_irr.rad >sky_irr.oct
echo 0 0 0 -0.000000 -0.052336 0.998630 |rtrace -h -I+ sky_irr.oct
result : 8.998834e+02 W/m2

Equals the direct irradiance which was "put" into the model (the sky is ignored, as in your example).

cheers

Jan

···

On 08.08.17 09:17, Joe wrote:

Hello,

when I use gendaylit to create a sun my images are too bright.

The command used is

gendaylit -ang 90 0 -W 900 200

void light solar
0
3 7.576e+006 7.576e+006 7.576e+006

solar source sun
0
4 -0.000000 -0.052336 0.998630 0.533000

This will put the sun to the zenith with reasonable values
for direct normal and diffuse irradiance in W/m^2.

Correct so far?

I removed the skyfunc, just the sun source is active.

I put together a simple scene:
- plastic cylinders as coordinate origin,
- large brown ground plate
- dielectric spheres (I just love those :slight_smile:

Here is an image with renderings of different sun settings:

Imgur: The magic of the Internet

When I use the sun as stated above the image on the right
side is rendered. Much too bright.

Reducing the parameters of void light solar:

middle image (7.576e+005 7.576e+005 7.576e+005 )
left image (7.576e+004 7.576e+004 7.576e+004)

Viewed with default parameters of rvu.

Do you have an idea what I am doing wrong
or might have missed?

Kind regards,
Joe

_______________________________________________
Radiance-general mailing list
[email protected]
https://www.radiance-online.org/mailman/listinfo/radiance-general

--
Dr.-Ing. Jan Wienold
Ecole Polytechnique Fédérale de Lausanne (EPFL)
EPFL ENAC IA LIPID

LE 1 111 (Office)
Phone +41 21 69 30849

Hi,

I am quite new to Radiance and have not used the exposure setting so far.
So far I had expected to get a "well illuminated" scene with some shadows
cast.
I have to read a bit on how exposure works.

Does it work like an ambient light to sort of adjust the overall brightness
of the image?

From what I understand so far exposure only affects
the view of the rendered image, not the actual physical properties
like illuminance, radiance and irradiance.
And if rtrace is used to "view" the image instead of rvu the
illumination would be consistent.

Is that correct?

Kind regards,
Joe

···

Am 08.08.2017 10:01 schrieb Jan Wienold:

Hi Joe,

why do you think it is too bright? Is the luminance to high? I guess
you just have to adapt your exposure setting for the image.

The physics for gendaylit and your parameters is correct:

1. test of luminance of the sun (ray in sun direction):

gendaylit -ang 90 0 -W 900 200 >sky.rad
oconv -f sky.rad >sky.oct

echo 0 0 0 -0.000000 -0.052336 0.998630 |rtrace -h sky.oct | rcalc -e
'$1=47.4*$1+120*$2+11.6*$3'
result: 1.356104e+09 cd/m2

This value looks reasonable - the max is about 1.6 e9 cd/m2

2. test of irradiance (sensor also looks towards sun):

gendaylit -ang 90 0 -O 1 -W 900 200 >sky_irr.rad
oconv -f sky_irr.rad >sky_irr.oct
echo 0 0 0 -0.000000 -0.052336 0.998630 |rtrace -h -I+ sky_irr.oct
result : 8.998834e+02 W/m2

Equals the direct irradiance which was "put" into the model (the sky
is ignored, as in your example).

cheers

Jan

On 08.08.17 09:17, Joe wrote:

Hello,

when I use gendaylit to create a sun my images are too bright.

The command used is

gendaylit -ang 90 0 -W 900 200

void light solar
0
3 7.576e+006 7.576e+006 7.576e+006

solar source sun
0
4 -0.000000 -0.052336 0.998630 0.533000

This will put the sun to the zenith with reasonable values
for direct normal and diffuse irradiance in W/m^2.

Correct so far?

I removed the skyfunc, just the sun source is active.

I put together a simple scene:
- plastic cylinders as coordinate origin,
- large brown ground plate
- dielectric spheres (I just love those :slight_smile:

Here is an image with renderings of different sun settings:

Imgur: The magic of the Internet

When I use the sun as stated above the image on the right
side is rendered. Much too bright.

Reducing the parameters of void light solar:

middle image (7.576e+005 7.576e+005 7.576e+005 )
left image (7.576e+004 7.576e+004 7.576e+004)

Viewed with default parameters of rvu.

Do you have an idea what I am doing wrong
or might have missed?

Kind regards,
Joe

_______________________________________________
Radiance-general mailing list
[email protected]
https://www.radiance-online.org/mailman/listinfo/radiance-general

Yes; exposure affects the appearance of the image but not the underlying
photo/radiometric values, or pixel data if you will. A Radiance image is
this amazing collection of spot meter data, basically. I’m not sure I
understand your last statement there, about the “illumination [being]
consistent”.

You can take a Radiance luminance image and run it through pcond to
'tonemap' the image and create all kinds of valuable effects. The simplest
is the '-h' option, which will map the pixel luminance according to human
vision capability. Thus, creating an image that conveys the appearance of
the scene as viewed by a human observer in that scene.

The main takeaway is that the raw Radiance images are physically accurate,
yet you are viewing them on dynamic range-limited devices, so they will be
hard-pressed to “look right”.

- Rob

···

On 8/8/17, 5:28 AM, "Joe" <[email protected]> wrote:

From what I understand so far exposure only affects
the view of the rendered image, not the actual physical properties
like illuminance, radiance and irradiance.
And if rtrace is used to "view" the image instead of rvu the
illumination would be consistent.

Is that correct?

Kind regards,
Joe

Thanks for your responses!
It works now as expected.

For future reference: I also found a good explanation here, page 9.
https://www.radiance-online.org/archived/radsite/radiance/refer/rc97tut.pdf/at_download/file

Does pcond -h something like automatically calculate the maximum luminance
and adjust the exposure? Or will it also apply the different perception of red, green and blue?

Is there a parameter for rvu to automatically set the exposure using the maximum Luminance
calculated?

···

Am 08.08.2017 17:03 schrieb Guglielmetti, Robert:

Yes; exposure affects the appearance of the image but not the underlying
photo/radiometric values, or pixel data if you will. A Radiance image is
this amazing collection of spot meter data, basically. I’m not sure I
understand your last statement there, about the “illumination [being]
consistent”.

You can take a Radiance luminance image and run it through pcond to
'tonemap' the image and create all kinds of valuable effects. The simplest
is the '-h' option, which will map the pixel luminance according to human
vision capability. Thus, creating an image that conveys the appearance of
the scene as viewed by a human observer in that scene.

The main takeaway is that the raw Radiance images are physically accurate,
yet you are viewing them on dynamic range-limited devices, so they will be
hard-pressed to “look right”.

- Rob

On 8/8/17, 5:28 AM, "Joe" <[email protected]> wrote:

From what I understand so far exposure only affects
the view of the rendered image, not the actual physical properties
like illuminance, radiance and irradiance.
And if rtrace is used to "view" the image instead of rvu the
illumination would be consistent.

Is that correct?

Kind regards,
Joe

_______________________________________________
Radiance-general mailing list
[email protected]
https://www.radiance-online.org/mailman/listinfo/radiance-general

Does pcond -h something like automatically calculate the maximum
luminance
and adjust the exposure? Or will it also apply the different perception
of red, green and blue?

Something like that, yeah. When low-luminance scenes are detected (by
inspecting the pixels’ values), the loss of color acuity is modeled.
Disability glare and veiling reflections are modeled as well. The exact
goings-on are detailed in Chapter 7 (Roadway Lighting) of the Rendering
with Radiance book, and the -h option to pcond is simply a combination of
luminance-based tweaks to the pixels, discussed in that chapter. The
output images lose all validity at the pixel value level, but they make
for compelling images on low dynamic range displays.

You will definitely want to look at Rob Shakespeare’s work (he’s
presenting an update at this years’ Radiance Workshop) with using Radiance
images and tone mapping to simulate scenes as perceived by aging eyes.
It;s awesome stuff, and a great illustration of the value of Radiance
images and the tools used to create and analyze them.

Is there a parameter for rvu to automatically set the exposure using the
maximum Luminance
calculated?

In rvu you can hit ‘e’ and the pick a point near where you know max
luminance to be and it will scale the exposure accordingly, albeit
linearly. You can also use ‘pextrem’ at the command line to get the
maximum and minimum pixel values (and x,y locations) of an input image.
Another utility you may be interested in is ‘phisto’, which will create a
histogram of a series of input images and supply a value you can use to
expose the input images uniformly (sorta).

- Rob

···

On 8/8/17, 11:50 PM, "Joe" <[email protected]> wrote: