The Levels Histogram

The Levels Histogram

You might have noticed the odd appearance of the histogram
for an unadjusted gradient. If you were to try this
setup on your own, depending on the size of the layer to
which you applied Ramp, you might see a histogram that is
flat along the top with spikes protruding at regular intervals.


The histogram is exactly 256 pixels wide; it is effectively
a bar chart made up of 256 single pixel bars, each corresponding
to one of the 256 possible levels of luminance in
an 8 bpc image (these levels are displayed below the histogram,
above the Output controls). In the case of a pure
gradient, the histogram is fl at because luminance is evenly
distributed from black to white; if spikes occur in that case,
it's merely an indication that the image height in pixels is
not an exact multiple of 256.
In any case, it’s more useful to look at real-world examples,
because the histogram is useful for mapping image data
that isn’t plainly evident on its own. Its basic function is
to help you assess whether any color changes are liable to
help or harm the image. There is in fact no one typical or ideal
histogram—they can vary as much as the images
themselves, as seen back in Figure 5.8.
Despite that fact, there’s a simple rule of thumb for a basic
contrast adjustment. Find the top and bottom end of the
RGB histogram—the highest and lowest points where there
is any data whatsoever—and bracket them with the Input
Black and Input White carets. To “bracket” them means to
adjust these controls inward so each sits just outside its corresponding
end of the histogram (Figure 5.11). The result
stretches values closer to the top or bottom of the dynamic
range, as you can easily see by applying a second Levels
effect and studying its histogram.black level in the histogram) and then blow out the whites
(moving Input White below the highest white value). Don’t
go too far, or subsequent adjustments will not bring back
that detail—unless you work in 32 bpc HDR mode (Chapter
11). Occasionally a stylized look will call for crushed
contrast, but generally speaking, this is bad form.
Black and white are not at all equivalent in terms of how
your eye sees them. Blown-out whites are ugly and can be
a dead giveaway of an overexposed digital scene, but your
eye is much more sensitive to subtle gradations of low black
levels. These low, rich blacks account for much of what
makes fi lm look like fi lm, and they can contain a surprising
amount of detail, none of which, unfortunately, would be
apparent on the printed page.
The occasions on which you would optimize an image by
raising Output Black or lowering Output White controls
are rare, as this lowers dynamic range and the overall
contrast. However, there are many uses in compositing
for lowered contrast, to soften overlay effects (say, fog and
clouds), high-contrast mattes, and so on. More on that
later in this chapter and throughout the rest of the book.