The following images show the Cornell University's famous "Cornell Box" 
reference scene for 3D rendering...

(1) as a photograph taken from the "real thing" at Cornell University, 
using sophisticated calibrated equipment, at a wavelength of 600nm; the 
original is an OpenEXR image; I applied linear brightness adjustment for 
more contrast and converted to PNG for convenience, which shouldn't hurt 
its fitness for serving as a reference image, but if in doubt you can 
compare with the original 
(http://www.graphics.cornell.edu/online/box/box_exr.zip).

(2) as a POV-Ray 3.7 render using geometry, surface reflectance and 
light source data as published by the Cornell University 
(http://www.graphics.cornell.edu/online/box/data.html) for a wavelength 
of 600nm, using "assumed_gamma 2.2"; light intensity was adjusted in an 
attempt to obtain a similar overall brightness as the photograph; notice 
how the side walls still appear much too dark while some other areas are 
already too bright.

(3) ditto but with gamma-correction applied to the color values; notice 
how this fixes the relative brightness of the surfaces, but otherwise 
still leaves the image with a much stronger contrast than with the "real 
thing".

(4) using the original color values again, but without "assumed_gamma". 
The brightness of the light source was adjusted by exactly the same 
factor that the photograph had been brightened; no other parameters were 
tweaked. Notice how the "real thing" is simulated pretty faithfully.

Post a reply to this message

Attachments:
Download '600nm.png' (1639 KB) Download 'cornell_gamma22_raw.png' (95 KB) Download 'cornell_gamma22_adjusted.png' (112 KB) Download 'cornell_gamma10.png' (89 KB)

Preview of image '600nm.png'

Preview of image 'cornell_gamma22_raw.png'

Preview of image 'cornell_gamma22_adjusted.png'

Preview of image 'cornell_gamma10.png'

clipka <ano### [at] anonymousorg> wrote:
> The following images show the Cornell University's famous "Cornell Box"
> reference scene for 3D rendering...
>
> (1) as a photograph taken from the "real thing" at Cornell University,
> using sophisticated calibrated equipment, at a wavelength of 600nm; the
> original is an OpenEXR image; I applied linear brightness adjustment for
> more contrast and converted to PNG for convenience, which shouldn't hurt
> its fitness for serving as a reference image...

I hope I'm not muddying the issue (or doing something stupid at my end), but
your posted Cornell .png photograph is definitely the worst of the lot, when
viewed on-line in the latest Firefox browser (v3.6.13): low contrast and darker
than any of the POV-Ray images. Perhaps it needs tweaking in some further way,
to make it a good comparison image? (I'm *guessing* that it was meant to compare
more favorably with the gamma 1.0 image, or vice versa.) Here are some simple
screen shots, assembled in Photoshop (which didn't materially affect how they
appear in Firefox); just a simple JPEG.

Ken

Post a reply to this message

Attachments:
Download 'images_compared.jpg' (124 KB)

Preview of image 'images_compared.jpg'

Am 21.12.2010 20:31, schrieb Kenneth:
> clipka<ano### [at] anonymousorg>  wrote:
>> The following images show the Cornell University's famous "Cornell Box"
>> reference scene for 3D rendering...
>>
>> (1) as a photograph taken from the "real thing" at Cornell University,
>> using sophisticated calibrated equipment, at a wavelength of 600nm; the
>> original is an OpenEXR image; I applied linear brightness adjustment for
>> more contrast and converted to PNG for convenience, which shouldn't hurt
>> its fitness for serving as a reference image...
>
> I hope I'm not muddying the issue (or doing something stupid at my end), but
> your posted Cornell .png photograph is definitely the worst of the lot, when
> viewed on-line in the latest Firefox browser (v3.6.13): low contrast and darker
> than any of the POV-Ray images. Perhaps it needs tweaking in some further way,
> to make it a good comparison image? (I'm *guessing* that it was meant to compare
> more favorably with the gamma 1.0 image, or vice versa.)

Its dark appearance in FF may be due to the fact that the images are 
presented on a white background, and the "real thing" has much more 
black border around it; especially with the preview this messes a lot 
with the image's appearance.

As for having low contrast, that's the point I'm making: The photograph 
/is/ a physically accurate image, so if there's any tweaking to do to 
make it look like the POV-Ray renders, then the tweaking needs to be 
done on the POV-Ray side - this comparatively low contrast is physically 
/correct/.

Post a reply to this message

"clipka" <ano### [at] anonymousorg> schreef in bericht 
news:4d110528$1@news.povray.org...

> Its dark appearance in FF may be due to the fact that the images are 
> presented on a white background, and the "real thing" has much more black 
> border around it; especially with the preview this messes a lot with the 
> image's appearance.

Idem for me as for Kenneth, also in Firefox. It is not the white background. 
Even the ceiling light in POV-Ray is pure white, compared to muddy grey in 
the photograph.

Thomas

Post a reply to this message

On 12/22/2010 09:58 AM, Thomas de Groot wrote:
> Idem for me as for Kenneth, also in Firefox. It is not the white background. 
> Even the ceiling light in POV-Ray is pure white, compared to muddy grey in 
> the photograph.

  There indeed seems to be something wrong with the original photograph,
as posted. One would indeed assume the ceiling light should be pure
white in the image (I'm assuming in the original HDRI image it's "whiter
than white").

  Exactly what kind of adjustment was made to the original data in order
to get the posted "photograph"?

Post a reply to this message

clipka <ano### [at] anonymousorg> wrote:

> (2) as a POV-Ray 3.7 render using geometry, surface reflectance and
> light source data as published by the Cornell University
> (http://www.graphics.cornell.edu/online/box/data.html)...

There's a rather strange typo in that link (in the original post, that is!) When
viewed *there* (in Firefox), it begins with http://www://www.graphics...  But as
you can see above, it suddenly appears correct in this reply! (I did no editing
to it here; that's the way it now shows up.)  WTF??? I wouldn't be surprised if
it reverts to the error again, when I post this. No way to know!

I think my computer is caught in a quantum superposition state-----

Anyway, here's the correct link (hopefully, *it* will show up without error)...

http://www.graphics.cornell.edu/online/box/data.html

Ken

Post a reply to this message

"Kenneth" <kdw### [at] earthlinknet> wrote:
> clipka <ano### [at] anonymousorg> wrote:

> ...I wouldn't be surprised if
> it reverts to the error again, when I post this...

And so it did! That's just...ODD. (You'll have to take my word for it
that the link appeared *correct* during the writing of my reply.)

Post a reply to this message

Warp <war### [at] tagpovrayorg> wrote:
> On 12/22/2010 09:58 AM, Thomas de Groot wrote:
> > Idem for me as for Kenneth, also in Firefox. It is not the white background.
> > Even the ceiling light in POV-Ray is pure white, compared to muddy grey in
> > the photograph.
>
>   There indeed seems to be something wrong with the original photograph,
> as posted. One would indeed assume the ceiling light should be pure
> white in the image (I'm assuming in the original HDRI image it's "whiter
> than white").
>
>   Exactly what kind of adjustment was made to the original data in order
> to get the posted "photograph"?

The original was taken with a CCD, so the captured response is linear. Taking
the 600nm OpenEXR version and converting it to a 2.2 gamma gives a realistic
image to my eyes. If the original was a photograph, then you'd have a non-linear
tone curve to deal with (exponential, but with rolloff on the top and bottom)

Cheers,
Edouard.

Post a reply to this message

Am 22.12.2010 20:55, schrieb Kenneth:
> "Kenneth"<kdw### [at] earthlinknet>  wrote:
>> clipka<ano### [at] anonymousorg>  wrote:
>
>> ...I wouldn't be surprised if
>> it reverts to the error again, when I post this...
>
> And so it did! That's just...ODD. (You'll have to take my word for it
> that the link appeared *correct* during the writing of my reply.)

And it does indeed appear correct here.

Post a reply to this message

Am 22.12.2010 22:06, schrieb Edouard:
> Warp<war### [at] tagpovrayorg>  wrote:
>> On 12/22/2010 09:58 AM, Thomas de Groot wrote:
>>> Idem for me as for Kenneth, also in Firefox. It is not the white background.
>>> Even the ceiling light in POV-Ray is pure white, compared to muddy grey in
>>> the photograph.
>>
>>    There indeed seems to be something wrong with the original photograph,
>> as posted. One would indeed assume the ceiling light should be pure
>> white in the image (I'm assuming in the original HDRI image it's "whiter
>> than white").
>>
>>    Exactly what kind of adjustment was made to the original data in order
>> to get the posted "photograph"?

The original image was much dimmer (half the physical brightness as the 
PNG I posted). I guess the dim overhead light is due to reaching the 
CCD's dynamic range limit, and the image was linearly adjusted in 
brightness to represent absolute light intensity in some nice SI unit.

> The original was taken with a CCD, so the captured response is linear. Taking
> the 600nm OpenEXR version and converting it to a 2.2 gamma gives a realistic
> image to my eyes. If the original was a photograph, then you'd have a non-linear
> tone curve to deal with (exponential, but with rolloff on the top and bottom)

The original is an OpenEXR, and therefore by definition linear. A proper 
image viewer should do the gamma adjustment for your display 
automatically, so you shouldn't do any manual gamma correction on the 
image, otherwise you'll just screw it up.

Post a reply to this message