The ISO definition of dynamic range - new tutorial article

This "room" is not ideally chartered to receive this matter, but it seems to have as relevant a portfolio as is currently available
************
We often encounter discussions of the dynamic range of a digital camera and how it may be influenced by various design parameters. Often these discussions are hindered by failure to articulate a specific definition of "dynamic range".

The ISO standard for such matters revolves around essentially this definition:

The dynamic range of a digital camera is defined as the ratio between (in a single "shot") the largest uniquely-representable luminance (the "saturation" luminance) and the luminance for which the signal to noise ratio (SNR) of the image is 1.0 (or actually, the luminance at which the SNR would be 1.0, based on a noise level determined at a fixed small luminance, a finesse made necessary by the fact that an actual SNR of 1.0 is "problematical").

I have recently posted to my technical information site, The Pumpkin, a new tutorial article, "The ISO Definition of the Dynamic Range of a Digital Still Camera", available here:

http://doug.kerr.home.att.net/pumpkin/index.htm#ISO-DR

It describes in detail the ISO definition and the tricky way it is defined.

Note that this standard revolves around examination of a digital "image", and this cannot be directly applied to the matter of the dynamic range of a camera from the perspective of its raw output, a matter that introduces a number of thorny issues.

Doug,

Thisc topic is so important. We've already been suckered for the last 10 years with marketing bluster and nonsense about the increased dynamic range of scanners. The put in a 16 BIT A to D convertor, label the $70 to $800 scanner as 48 BIT and come up with claimed O.D. of 4.1!

This is obviously either ignorant delusiion or else willful scamming of the less astute artist or photographer.

So, Doug, the claims of Canon and Nikon are wonderful, but are they really valid in practice?

Asher

Thanks very much for providing what looks to be a most interesting article on a clearly important subject. At least now I have the choice of dying somewhat less ignorantly.

Doug Kerr said:

Note that this standard revolves around examination of a digital "image", and this cannot be directly applied to the matter of the dynamic range of a camera from the perspective of its raw output, a matter that introduces a number of thorny issues.

Click to expand...

Maybe in this regard the DR definition is no worse than the definition of depth of field which relies on how an image is printed and viewed as well as on the optical performance of the lens and camera.

- Alan

Hi, Alan,

Alan T. Price said:

Maybe in this regard the DR definition is no worse than the definition of depth of field which relies on how an image is printed and viewed as well as on the optical performance of the lens and camera.

Click to expand...

Indeed.

Best regards,

Doug

Asher Kelman said:

Doug,

Thisc topic is so important. We've already been suckered for the last 10 years with marketing bluster and nonsense about the increased dynamic range of scanners. The put in a 16 BIT A to D convertor, label the $70 to $800 scanner as 48 BIT and come up with claimed O.D. of 4.1!

This is obviously either ignorant delusiion or else willful scamming of the less astute artist or photographer.

So, Doug, the claims of Canon and Nikon are wonderful, but are they really valid in practice?

Click to expand...

The DR of individual pixels hasn't improved much in the last few years, except perhaps within the models of certain brands.

Except for the Pentax K10D, which, according to someone who I tend to trust, breaks this barrier (at 4550:1), the maximum signal to read noise level in a pixel hasn't surpassed about 3000:1 for Canon or Nikon (and Canon has been there since at least the 1Dmk2, and I believe the Fuji DSLRs have been there for a few years, too). The definition Doug gives is the ratio of maximum signal to the signal at which a 1:1 ratio occurs, and that level is a little bit above the read noise "floor", because of the contribution of shot noise. The more photons collectable and unclipped, while maintaining this read noise floor at 0.033% of saturation, the closer the 1:1 SNR-limited DR gets to actually being 3000:1. This is just about the pixel, though, or a fixed numbers of pixels as an image. The dynamic range of an image, however, has to be related to the number of pixels, as well as to their individual characteristics. A 1MP camera with a DR of 3000:1 at the pixel level is not going to have more DR than one that has a pixel DR of 1500:1 and 22 MP. That's preposterous, yet that is a commonly held belief. Perhaps the misleading metaphor of "noise floor" is partly to blame; the misconception is that there is nothing below it, and it doesn't matter how many pixels you have, there is nothing visible under the floor. The idea that nothing is visible under the floor, though, is wrong. The noise floor is just a semantic boundary, above which, signal is greater, and below which, noise is greater. Below the threshold of visual resolution, the noise floor "drops" for practical image-level purposes, as averages of local pixels are experienced, instead of single ones.

Hi, John,

John Sheehy said:

The definition Doug gives is the ratio of maximum signal to the signal at which a 1:1 ratio occurs . .

Click to expand...

Not quit. The conceptual basis of the ISO definition is what you state, but there are difficulties in measuring SNR's as "poor" as 1:1 (the net value would be negative a substantial percentage of the time, and if since we are working in luminance-equivalent terms, that is impossible, so the measured noise level understates the random phenomenon.).

This dilemma is finessed by the actual ISO definition, which is the ratio of the saturation luminance to the luminance-equivalent noise (RMS) at a luminance of 1/100 saturation. (That numerical value is used as a "proxy" for the signal level that would have an SNR of 1:1 if that were possible!)

I don;lt mean to suggest that this invalidates what you said - just tying to be sure I am quoted accurately.

Le crayon rouge ne dort jamais.

Best regards,

Doug

High dynamic range (HDR) images enable photographers to record a greater range of tonal detail than a given camera could capture in a single photo. This opens up a whole new set of lighting possibilities which one might have previously avoided—for purely technical reasons. The new "merge to HDR" feature of Photoshop CS2 allows the photographer to combine a series of bracketed exposures into a single image which encompasses the tonal detail of the entire series. There is no free lunch however; trying to broaden the tonal range will inevitably come at the expense of decreased contrast in some tones. Learning to use the merge to HDR feature in Photoshop CS2 can help you make the most of your dynamic range under tricky lighting—while still balancing this trade-off with contrast.