[opendtv] Re: News: High Dynamic Range imaging

  • From: Craig Birkmaier <craig@xxxxxxxxx>
  • To: opendtv@xxxxxxxxxxxxx
  • Date: Fri, 4 Aug 2006 09:39:36 -0400

At 9:32 AM -0400 8/3/06, Tom Barry wrote:
>Interesting article, and something I don't think I've ever properly
>considered before.  Can anyone explain why you couldn't solve the
>problem simply by extra pixel bit depth, maybe in a log scale?  I'm not
>sure why you need to take multiple pictures.

Mark and others have already provided some helpful feedback to your 
questions, so i will only expound on a few points.

One of the magical aspects of film negatives is that they do not have 
linear capture characteristics. The knee and toe that i mentioned 
yesterday, are in fact a very important aspect of film response. Over 
the mid range values film tends to be linear, as is the case for 
electronic sensors. But in the blacks and whites at the ends of the 
linear range film has a logarithmic response - essentially an 
S-curve. This additional range can provide more image detail in the 
blacks and whites (based on the actual exposure), and it can be used 
during printing or digital scanning to bring out the details desired.

Logarithmic coding at 10-14 bits is commonplace today for the files 
produced by film scanners. And we are now seeing digital imaging 
systems that capture RAW image data from the sensors (before gamma 
correction) at extended bit depths (10-12 bits); this added 
information can then be used in post production to extract the 
desired image details during color grading and correction.

The reason to take multiple pictures is the linear response of the 
analog image sensors. By taking two or three images with different 
exposures we are trying to emulate the logarithmic response of film 
negative, providing the additional range of blacks and whites that 
are typically lost with a single exposure.

>Also, assume you instead are Hollywood and currently starting from film
>cameras, getting some extra dynamic range.  Is there still extra range
>by the time you transfer it to some electronic medium?

This is certainly possible, but is limited by the weakest link in the 
chain. If the image is subjected to 8 bit processing then the range 
of possible values is limited. But the mapping of values from the 
original HDR image is typically much better than a camera with linear 
response, acquiring the images with 8-bit depth.

>If so, why?  Do
>telecine machines just have better range than cameras?  Can that extra
>range transfer all the way to, say, DVD, or HDTV displays?

Telecine machines are now capable of delivering 12-14 bit response, 
even higher in some cases. This is a major reason that most Hollywood 
projects are still captured with film, which is then digitized for 
post production and eventual release, either as film prints or 
digital presentation.

>And does film have better range in theater film projectors as opposed to
>current digital theaters?

Theater film projectors have their own liabilities. As Mark stated, 
the quality of release prints is not the same as the original 
negatives, or a high quality print. In addition, film gate jitter 
tends to act as a low pass filter, removing some of the high 
frequency detail that may be present in the release print. Thus 
digital presentation may eventually prove to offer superior 
presentation characteristics, especially as the dynamic range of the 
projectors improves.

You can UNSUBSCRIBE from the OpenDTV list in two ways:

- Using the UNSUBSCRIBE command in your user configuration settings at 

- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word 
unsubscribe in the subject line.

Other related posts: