Luminance and Contrast

Definitions

Ideal Luminance:

This means the intensity of a pixel expressed on a scale from 0 to 1, where 0 is the lowest intensity the visual display hardware can produce, and 1 is the highest. It is often convenient to work with ideal luminances because they allow you to express stimulus characteristics in a hardware-independent way. However you would need to convert them to physical luminances (perhaps using the function IdealToPhysicalLuminance()) before reporting them in a publication.

In Shady, the carrier content of a Stimulus is generally expressed on this ideal 0-to-1 scale: this includes its signalAmplitude, the pixel values in its floating-point texture array (if any), its noiseAmplitude and its backgroundColor. The only exception is when texture content is expressed as an 8-bit integer array (values ranging from 0 to 255) - for example when it is loaded directly from an image file.

Physical Luminance:

This means the intensity of a stimulus patch in real physical units such as candela / m^2. Given accurate gamma-correction, we assume that physical luminance is proportional to ideal luminance, with an additive offset. The offset is a consequence of the fact that a screen’s minimum intensity is never truly zero (either due to reflected ambient light, or the imperfect black level of most current screen technologies). An OLED screen in an otherwise perfectly dark room may achieve close to 0 - this is currently the only setup in which you can ask how much more black the screen could be and the answer is none. None more black.

Ideal Contrast Ratio:

By “ideal contrast ratio”, we mean “a contrast ratio computed from ideal luminance values”, independent of whether this is computed by the RMS method or the Michelson method. Since ideal luminances go down to zero, ideal contrast ratios can reach 1.0, unlike physical contrast ratios.

It is sometimes convenient to work with an ideal contrast ratio because it can be computed without having to perform actual photometer measurements, and can provide an approximation to the corresponding true physical contrast ratio. But you would need to convert it to a physical contrast ratio (perhaps with the function IdealToPhysicalContrastRatio() ) before reporting it in a publication.

Physical Contrast Ratio:

By “physical contrast ratio”, we mean “a contrast ratio computed from physical luminance values”, independent of whether this is computed by the RMS method or the Michelson method. Since the physical luminance at the location of a “black” pixel is never actually zero, a physical contrast ratio will never reach 1.0 (though it may come very close, depending on the screen technology and ambient light control).

RMS Contrast Ratio:

This is a method of computing a contrast ratio according to \(\frac{\sqrt{\frac{1}{N}\sum_x\sum_y (L(x,y) - L_{\mu})^2}}{L_{\mu}}\), or in plain text:

average_over_x_and_y(  (luminance(x,y) - background_luminance) ** 2  ) ** 0.5
----------------------------------------------------------------------------
                        background_luminance

If luminance \(L(x,y)\) and background_luminance \(L_{\mu}\) are both ideal luminances, then the result is an ideal contrast ratio. If they are both physical luminances, then the result is a physical contrast ratio.

Michelson Contrast Ratio:

This is a method of computing a contrast ratio according to \(\frac{L_{\max} - L_{\min}}{L_{\max} + L_{\min}}\), or in plain text:

L_max - L_min
-------------
L_max + L_min

If L_max and L_min are both ideal luminances, then the result is an ideal contrast ratio. If they are both physical luminances, then the result is a physical contrast ratio.

The Michelson method is best suited to simple periodic stimuli whose luminance varies equally above and below the background luminance. For visual noise or natural stimuli you would probably want to use an RMS contrast ratio instead.

Normalized Contrast:

This may be a somewhat misleading name, but we use it anyway, as do other psychophysical software packages. To mitigate confusion, we are careful always to include the word “ratio” when talking about RMS and Michelson contrast ratios, and to omit it here when talking about normalized contrast, which is not a ratio.

By “normalized contrast”, as in the property Stimulus.normalizedContrast, we mean a straightforward scaling factor that acts as a multiplier on deviations of luminance from the background, regardless of absolute luminance level:

luminance = background_luminance + signal * normalized_contrast

So, for example, if our background_luminance is 0.5, in ideal units, and our signal is 0.5 * sin(x), a normalized_contrast of 1.0 would allow this sine-wave signal to span the full luminance range of the screen (note that the background_luminance could not be anything other than 0.5, otherwise the signal would go out of range). A normalized_contrast of 0.2 would mean that the same signal spans one fifth of the full intensity range of the screen (so now you could set the ideal background_luminance to anything from 0.1 to 0.9).

Contrast-ratio computation functions (ideal or physical)