[Radiance-general] (no subject)

[Greg Ward]

I tried sending this yesterday, but used the wrong sending address and 
had it rejected...

During my hiatus from physically-based rendering in 1999-2001, I was 
employed as a software engineer at Shutterfly, Inc.  I worked 
intimately with their digital photographic printers, which went 
directly from 24-bit RGB images to photo paper via LED and fluorescent 
line-exposure units.  From this experience, I can assure you that the 
best printers cannot reproduce a black darker than 1%, which doesn't 
include the 3% reflection you get off the surface coating.  In other 
words, a glossy print, angled so that the specular reflection mirrors 
something black, produces a maximum contrast no better than 100:1.  In 
an uncontrolled environment where the glossy reflection is not black, 
the maximum contrast can degrade terribly, to as little as 2:1 or 
worse.  (That's why people looking at high contrast prints always angle 
them to minimize reflections.)

Although it is theoretically and physically possible to produce prints 
with better than 100:1 contrast, the surface reflection described will 
undermine this contrast to where there is little point in improving it 
further.  This is most likely why no one has bothered.  Viewers in a 
typical ambient environment simply cannot appreciate additional depth 
in the shadows.  As Mark pointed out, there is nowhere to go on the top 
end, as the best photo paper has a maximum reflectance around 90%.  So 
much for reflection prints.

However, it is possible to get excellent dynamic range out of film 
transparencies.  A well-made transparency, viewed through a slide 
projector or light table, can have a dynamic range of 1000:1 or better. 
  What's more, you can actually SEE the detail in the shadows and 
highlights with a transparency, unlike a reflection print where you're 
constantly fighting with the background.  Unfortunately, I haven't 
found any reasonably priced digital film recorders or services that do 
a good job of taking higher resolution pixel data and converting it to 
a wide range of densities, and this is largely a problem with the D/A 
converters they employ.  There is no reason I know why they couldn't do 
better, and there are probably Hollywood post-production labs that 
acheive much better image depth, but they are quite expensive and not 
widely used at this time.  In the consumer market, there is nothing.

Around the corner -- in the next year or two -- you can expect to see 
high dynamic range display devices enter the commercial market.  I am 
working with a Canadian company, Sunnybrook Technologies, that has made 
great strides in this direction.  For more information, you can check 
out this year's Siggraph paper on the topic:

Seetzen, Helge, W. Heidrich, W. Stuezlinger, G. Ward,  L. Whitehead, M. 
Trentacoste, A. Ghosh, A. Vorozcovs,  "High Dynamic Range  Display 
Systems," ACM Trans. Graph. (special  issue SIGGRAPH 2004), August 
2004.  <http://www.anyhere.com/gward/papers/Siggraph04.pdf>

Ultimately, I think HDR displays are the best way to appreciate 
lighting simulations.  Until then, we're kind of stuck with low 
dynamic-range, tone-mapped prints and displays.

If anyone knows of a digital path to high-contrast transparencies, I'd 
love to hear about it.

-Greg

P.S.  Regarding the OpenEXR format.  It is not a cheap rip-off of 
Radiance's RGBE format by any stretch of the imagination.  It is a more 
accurate format that is valuable for post-production and special 
effects image processing, and a very well thought-out library.  For a 
comparison of various HDR image formats, check out my web page on the 
subject.

	http://www.anyhere.com/gward/hdrenc/

Erik Reinhard, Paul Debevec, Sumant Pattanaik, and I are currently 
working on a book for Morgan Kaufmann Publishers tentatively titled, 
"High Dynamic Range Imaging."  We expect it to be out by next summer.