Idaho Airships, Inc.

Gamma is the rate of change in the luminosity of a pixel relative to the change in that pixel’s input value. At a gamma of 1, a perfectly black pixel will be perfectly black and a fully energized pixel will be perfectly white, with all intermediate steps produced in equal increments. This is the way that most digital camera sensors capture images, and it is termed a “linear” format.

Photoshop Curves Dialogue Box demonstrating a “linear” situation, in which the white point is white, the black point is black, and all intermediate steps in luminosity increase at regular degrees.

Photoshop Curves Dialogue Box demonstrates a “nonlinear” situation, in which the white point is white, the black point is black, but intermediate steps in luminosity increase at irregular, exponentially-determined degrees.

Human visual perception is not linear, so digital captures must be adjusted to appear natural to us and to allow an efficient tonal range presentation using limited bit depth. This adjustment is nonlinear, and as such has a variable rate of change throughout the range of the continuum from black to white. That amount of change is determined by how much a certain point on the line/Curve is adjusted. In the graphic above, the center luminosity point (gray value of 128, the “Input”) is reduced to a value of 64 (the “Output”). Without our directly engaging the math, this adjustment can be described as an exponent (“Output” = “Input” ^ exponent), which for the purposes of digital imaging has been conventionally termed “Gamma.”

When you adjust a Curve, one portion of the resulting Curve will have a steeper slope and one will have a more gradual slope than before the adjustment. Please note that in the adjustment demonstrated above there is much more length of the adjustment line above (in the higher luminosity segment) the centerpoint than below. Obviously, there will be more difference between the steps of luminance above the middle point than below. In other words, the amount of luminance change between luminance 20 and 21 is very small while the amount of luminance change between 230 and 231 is relatively much greater.

In fact, the Curve above is a negative curve (it reduces the midpoint luminance), but is the type of Curve administration most commonly used on properly exposed digital photographs (this is a single-point adjustment, while many advanced practitioners use multi-point Curves adjustments that exceed this exploration of gamma).

The Gamma applied to linear image captures so that they will appear natural to humans and aerial photographers (!) are positive, and described by exponent values such as 1.8 or 2.2. In the case of positive gamma adjustment, the area of lower luminance will have much larger changes in luminance than the area of higher luminance-exactly the opposite of what is demonstrated above.

Apple hardware typically uses a Gamma of 1.8, at which pixel luminance graduations in the dark portions of the continuum have less difference than with a Gamma of 1, but there is still some potential for obvious stepping, or “posterization.” PC hardware typically uses a Gamma of 2.2, for which the rate of change is lessened so that graduations in the dark portions of the continuum have less difference than they do at Gamma 1.8. The differences are palpable.

Therefore, the PC Gamma selection is more hospitable to image output, while the Apple Gamma selection is otherwise more perceptually accommodating. (This is a point of great contention and my phrasing, while succinct, will frustrate practitioners with a deep knowledge of this matter. Please accept this description as a generality).

When you “expose to the right” during digital imaging captures, you are in fact facilitating the retention of as much data as possible during the linear-to-nonlinear conversion. An excellent treatise on Gamma and “exposing to the right” is offered online by Adobe.

If you’d like to extend your familiarity with Gamma, and perhaps manipulate/optimize your system Gamma, please visit Norman Koren’s Gamma and Black Level page. Another good resource is this page by Charles Poynton, which also explores the NTSC address of Gamma.

Incidentally, if you noticed that the background Histogram in the examples is a bell-shape, and wondered why this might be the case for a Linear Gradient, it’s because I drew the Gradient with “Dither” selected. Why? Gradients drawn with Dither, formatted at 72dpi and compressed with .jpg (for monitor display) posterize noticeably less than those without dithering.

L