Why Digital Imaging Doesn't Work


It appears that I wrote this way back in the year 2000. Long before HDR and tone-mapping. Well, thankfully, what I wanted is coming to pass. It's March 2010 right now, and I haven't been following photography very closely, but if High Dynamic Range CCD's for digital cameras aren't available yet, they will be soon. Photographers are starting to get the tools and flexibility that audio engineers have had for many years. It's about time.


At this point [2000], I've been struggling for over 2 years to get decent images in my digital darkroom. I admit there is a big learning curve, and I'm still on it. Along the way, I've noticed a few things that strike me as odd. Most of them have direct parallels in my previous obsession, audio engineering. I'll try to explain my observations here and give a few suggestions that might make the world of Digital Imaging a bit easier to bear.

Looking Good?

The first thing that struck me as odd was when I was reading Giorgianni and Madden's book, "Digital Color Management: Encoding Solutions". This book (and consequently the entire goal of the Kodak corporation) has what I believe to be a fundamental flaw. It insists that the most important aspect of an imaging system is that the results look "good". My question is "Good to who?" The "average observer"? Oh, please. I really don't want my hard work to only look good to the average observer! Instead the tools should give me the flexibility to make something look however I want it to look. No imaging system I've ever used has this seemingly "magic" capability. Audio systems I've used do. The digital imaging industry hasn't figured this out yet for some reason.

Here's an example. Why isn't there a low-contrast unbiased transparency film? In the audio industry we have low-noise, near-flat response microphones. Why can't we have the same thing in the imaging industry? Scan the transparency into the computer and turn it into whatever you like. Because it is low-contrast, everything is there for you to work with (modest CCD scanners aren't taxed). Because it is unbiased (color-wise), the color will be predictable and easy to mold into whatever you want it to be. Sure, the image on the slide will look like crap to the "average observer", but that's not what the slide is for! In the audio example, that flat-response microphone sounds horrible to the untrained ear. The experienced engineer can turn it into something that sounds wonderful.

This idea would kill several birds with one stone. Color Negative film and all its associated problems would be eliminated. This new kind of film would require little color correction and would be able to capture a scene dynamic range that is equivalent to that of color negative film. Would it be possible to make such a film? I'm not a chemist, so I don't have any idea. My understanding is that the behavior of current transparency films was engineered simply to make sure they "Look Good".

Gamma Compression

Another example is Gamma compression. I'm not Timo Autiokari, but his fundamental point that "Gamma 1 (linear) is better" is correct and supported by Giorgianni and Madden themselves:

If a greater number of bits had been available, the metric might have been linear, rather than non-linear, with exposure.
- Page 281, Digital Color Management: Encoding Solutions

A similar situation exists in audio. For lower-quality audio systems, like the telephone, Gamma compression is used extensively. For higher-quality audio systems, like CD-players, there is no Gamma compression. Admittedly, linear digital imaging systems require even more memory and disk space for manipulating images, so there are practical considerations at this point in time. I'm hoping this will become less of an issue in the near future.


I hate to talk like a marketing-type, but these are the kinds of "revolutions" that are long overdue in the imaging industry. I'm sure they will come to pass, it is only a matter of time. In the meantime, we continue to suffer.

<- Back to my Photography page.

Copyright ©2000-2010, Ted Felix Disclaimer