Zones

So, we have established a reference value for our Zone System in the form of the light scattered from an 18% diffuse gray card. We have defined this reference light intensity as Value V. We have established a unit of measure for our scale; namely, the stop (a doubling or halving of light energy). We have learned that practical scenes can have specular elements or direct light sources that exceed Value V by 3 stops or more. We have also learned that the L-channel of the LAB color space has been designed around human vision; it is an accessible proxy for how people will view prints, digitally rendered images, and scenes.

For the accurate rendering of a scene in the form of a print, it would seem natural to establish the goal of printing any tonal element of the scene that metered as Value V at the same Value on the print. The same would seem to be true for any other Value. However, in attempting to achieve this desideratum, we encounter the obstacle that scenes can have a broader range of light than a print can express. This was as true in Ansel Adams’ day as it is now; and Adams proposed a solution to the problem. DSLRs behave a little differently than film; but the sun isn’t that much different from what Adams had to deal with. He seemed to get it about right; and so can we now.

The Zone System is a mapping between a Value and a Zone. The basic reference Value is Value V, the luminance of an 18% diffuse gray in the scene illumination. The basic reference Zone is also Zone V, which is the “normal” exposure setting. For example, suppose we meter a scene and read a Value of 1/30 at f/8. A Zone V, “normal” exposure is to select 1/30 at f/8. The notion is called “placing”. We say that we place Value V (the reading of an 18% diffuse gray) on Zone V. Once this placement is made, the rest of the luminance values fall on the other Zones. This gives us one degree of freedom. We can pick one Value to map to one Zone. The rest of the mapping falls into place with that one selection.

Here is an example of how this works in practice. The following shots were taken with a D700 using a Nikon 50mm f/1.8D lens with spot metering set for the center of the image. To avoid any vignetting, they are cropped in to about 1/4 of their original size in pixels. The images were taken in open shadow of the side panelling of my home. The first image is based on the built-in metering of the camera as applied to the panelling. It was shot at ISO 200, f/8, 1/200s. This is the Zone V exposure in this sequence. The other exposures are as follows:

 

Zone IV 1/200 at f/11 Zone VI 1/200 at f/5.6
Zone III 1/200 at f/16 Zone VII 1/200 at f/4
Zone II 1/200 at f/22 Zone VIII 1/100 at f/4
Zone I 1/400 at f/22 Zone IX 1/50 at f/4
Zone 0 1/800 at f/22 Zone X 1/25 at f/4

 

The following image was taken under the same lighting conditions but shows a Lastolite 18% gray card. The exposure was spot metered off the card, and turned out to be ISO 200, 1/60s at f/8, which is 1 2/3 stops higher than the Zone V exposure in the sequence above. In other words, Value V is ISO 200, 1/60s at f/8. In the sequence as shown, Value V has been mapped onto, roughly, onto a Zone III exposure. I say “roughly” because the offset is not quite 2 stops. This also suggests that the Zone VII exposure is the closest to “normal”.

18% gray card Value

18% gray card Value

These images were shot as 14-bit RAW files. They were brought into Photoshop CS6 through Adobe Camera Raw where the only processing was to correct for white balance using the gray card as a reference. The RAW files were converted as 16-bit AdobeRGB images. In Photoshop, each image was translated from AdobeRGB into LAB. The a and b channels were set to 0 values and the images were cropped to about 1/4 their original area. The images are screen shots from within PS showing a 101×101 pixel sample of the L-channel value near the center of the image. I confirmed the accuracy of the L-channel readings by applying a 1000 pixel radius Gaussian blur to the image in order to get a good average reading. Here is a graph showing the results.

L-Value vs Zone

L-Value vs Zone

I have shown curves like this before in posts on this site. And in spite of what you may have read about the behavior of DSLRs, this curve is extremely “film-like”. In fact, since there is a relationship between the traditional measure of density for negatives and prints, on the one hand, and Luminosity (aka L-channel value), on the other, it is quite possible to show a curve giving the density for this set of DLSR exposures in a very traditional manner. But that’s not for right now.

Looking back at the images for the various Zone exposures given above, the average L-Values for wall in the Zone VII exposure are around 86 while those in the final exposure with metering off the gray card are 92. This is consistent with the 1/3-stop offset between the two exposure settings.

As well, let’s distinguish between the texture within a panel of the wall, and the detail that distinguishes panels. The variation in L-Value within a panel is on the order of a few points; e.g., 90-94. Let us refer to small variations of a few tenths of a stop or so as texture. The variation in L-Value within a panel and the shadows of the panels themselves is about 1-stop; e.g. 93-73. Let us call local variations of the order of 1-stop as detail. I picked this example because of the relatively low range of light in the subject and the distinctions between texture within the panels and detail distinguishing the panels.

With that in mind, look back at the sequence of Zone exposures. The Zone X exposure has the barest hint of detail in a pure white background. There is no visible texture. That there is any visible detail at all is a credit to the D700. In the Zone IX exposure, detail is visible and texture is becoming noticeable. In the Zone VIII exposure, both detail and texture are apparent. Typically, this Zone is where you would want to place textured whites like a snowy surface or highlights on white skin. By Zone VII, texture and detail are both apparent without any compression. The overall sense of this Zone is a very light gray; it would be a suitable exposure for very light skin, or snow with lighting from the side (although that might give a “muddy” impression depending on the rest of the image.) Zone VI is the standard exposure for white skin or a light blue sky. Zone V is middle gray: a middle blue northern sky, weathered wood, dark skin in sunlight. Zone IV is shadows on white skin in sunlight, dark foliage, dark stone, light shadows in landscapes. Zone III is a strong dark material and yet retains texture and detail. Zone II clearly shows both texture and detail but with a very dark quality. Traditionally (with film), Zone I would show minimal detail and Zone 0 would show none; however, it should be apparent that the D700 with 14-bit RAW capture retains both texture and detail even in Zone 0.

There is a reason for this difference. In the film world, the lower noise limit had little to do with effective ISO. To represent deep darks on a negative, it was not exposed, leaving it as transparent as possible. Still, no film base was completely transparent; and superimposed on the film base was some “fog” or light noise. These effects do not occur in a digital sensor. The ultimate limits are the combination of digital noise in a pixel and the quantization limit of the analog to digital converter. For a 14-bit RAW capture, the quantization limit is, by definition, 14 stops down from the maximum value. At base ISO, the digital noise in the sensor will usually be of the order of a few quantization steps at the lowest sensor values. This allows for much more detail to be retained in the lowest exposure values with a modern pro-grade DSLR than was possible with film.

I have purposely made this presentation highly derivative of what Adams himself wrote in The Print (p. 48-60) just to show this distinction between the film-based examples he gives there and what a modern camera is capable of. Note also that the “Info” panel values shown in the exposure sequence underestimate the texture retained in the lower exposures because they are based on an 8-bit display setting. The PSD files are actually 16-bit; so there is even more available data in the textural variations than you might at first imagine.

Let’s expand on this point. Near the center of the L-channel versus Zone curve, 1-stop (1 Zone) corresponds to a 20 point change in L-Value. Below is a graph showing the standard deviation of the L-channel for each Zone exposure. [If the phrase “standard deviation” means nothing to you, think of it as the span up and down from the average L-channel value that about 2/3rds of the pixels fall within. It is a measure of the range of light in the exposure. Three standard deviations up and down from the average is about 99% of the full range.]

The curve is clearly peaked with a maximum of about 13 points (just over 1/2 stop) for Zone V. You can think of this as measuring the local tonal contrast. As we move up or down in exposure Zone from Zone V, the tonal contrast is increasingly compressed. The curve is roughly symmetric about Zone V with Zones II and VIII (3 stops up and down from the middle) being about equal at 4 points; and this is roughly 1/3 the variation in the mid-tone exposure. However, it is apparent from the data, which is consistent with a subjective view of the image gallery, that there is more detail and texture at Zone 0 than in Zone X. This curve is confirming what our eyes have told us.

L-channel standard deviation versus Zone

L-channel standard deviation versus Zone

But the amount of detail, measured statistically from the L-channel value, begins to fall off even in Zones IV and VI. Of course, this is straight out of the camera without any enhancement from sharpening or tonal contrast filtering. Still, we are seeing a 13% range in the Zone V exposure and only 10% in Zones IV and VI. By Zones II and VIII, the range is down to 4% of maximum.

The images I’ve posted here are, perforce, JPGs, which have only an 8-bit representation. The smallest step is 1 part in 256 or about 0.5%. Zones 0 and IX have ranges of only 1%, so these will be severely clipped in the gallery here. Zone X has a standard deviation of about the minimum expressible range in a JPG image, so any detail that the 14-bit RAW NEF or PSD file might have had will be lost when rendered in that manner. Since JPG format, and its internal sRGB tone curve, are designed for digital monitors, whatever the 16-bit range in my PSD file might be, most monitors would also have issue presenting it anyway.

The same is true for many printers and print services, which are limited to 8-bit files. For example, if you are reading these pages because you are obsessed with B&W (as am I), and you have an Epson printer, you might now wonder whether you would get better detail and texture rendering in the extreme Zones by using the Advanced B&W mode, which is limited to 8-bit resolution, or the AccuPhoto HD2 mode, which has 16-bit resolution. Stay tuned for the results in later pages.

We have noted that the Zone V exposure for my wall was “incorrect” in the sense that Value V from an 18% gray card came out almost 2 stops too low. What happens if we take the Zone V exposure and use Photoshop to “re-expose” it properly? What does that mean? Well, take a look at the following screen shot:

Re-exposing by 2 stops

Re-exposing by 2 stops

In the shot, I’ve taken the Zone V exposure and applied a curves layer to the L-channel that pushes the L-Value of 55 up by 2 stops, or 40 points, to 15. [Note that the “15” refers to density the way I have my UI set up, so that’s the same as 100 – 15 = 85 as an L-Value.] The result is an average value of L=92 near the middle of the image and a standard deviation of 6.78, which is perfectly consistent with the standard deviation of the Zone VII exposure from the camera. In other words, I have been able to rework the data from the original exposure to render a result that appears to have been shot with a more sensitive setting.

Here is a screen shot of the two files side-by-side. One is the Zone VII exposure; the other is the Zone V exposure with a correcting curve. Which is which?

Zone V -corrected & Zone VII

Zone V -corrected & Zone VII

There are many different way to achieve an effect like this in programs like Photoshop, Lightroom, Capture One, Capture NX2, Aperture, and so on. In my humble opinion, they form the basis for for dealing with (drum roll please) the full range subject; that is, images that include diffuse surfaces from nearly 0% to 100% and specular highlights or direct lighting.

Previous Page … ••• Next Page …