In order to better understand what a histogram is, lets first take a look at
tables and charts in general using tuxedo rentals as an example.
Table 1. A table showing rentals by color
Figure 1. The same data plotted as a spreadsheet bar chart
Figure 2. With so many bars, they get jammed together
Figure 3. A Photoshop histogram. Mid tone is in the center
of the X axis
Table 1 shows tuxedo rentals by color in tabular form. This information
can also be shown graphically. Figure 1 shows the same data plotted as a
vertical bar chart. The advantage of the bar chart in Figure 1 over Table
1 is the bar chart tells us very quickly which colors have the most rentals, the
least rentals and in between.
Assume that we are a much larger renter of tuxedos. In fact, we offer our
clients over 250 shades and styles of black, gray, off-white and white. If
we created a chart of all these shades of black, gray and white, it would be a
very wide chart indeed if we tried to label each color. If instead, we
plotted this information in a chart without the labels with the darker shades on
the left and the lighter shades on the right, we would get something like Figure
2. Even though Figure 2 is not as precise as Table 1, it does give us an
idea that most people prefer darker tuxedos over the medium gray tuxedos.
It also appears that light gray tuxedos are also popular. In a chart like
Figure 2, the creator of the chart assumes that people know what the X and Y
axis mean. In this case, the X axis is shades of black, gray and white
with black on the left and white on the right. The Y axis is quantity.
Now, lets switch gears and think about digital images. On the
Information page, we learned that light is composed of color and
lightness. If we were able to chart this information we would be able
to see where our information is clustering. Well, the good news is this
information is charted for us in Photoshop and this chart is called the
histogram. A sample histogram is shown in Figure 3.
A histogram for digital images is also a vertical bar chart, and like Figure 2
above, the X and Y axis are not labeled because it is assumed we know what the X
and Y axis represent. First, lets discuss what the histogram is charting.
A histogram plots the number of pixels along the tonal scale. The X axis
is the digital tonal scale with solid black on the left, solid white on the
right, mid tone in the middle and all the shades of luminance in-between.
The Y axis is relative quantity. The higher we got up the Y axis, the
greater the number of pixels in a given tone. Because the tonal scale
ranges from 0 to 255, the individual vertical bars that show the relative number
of pixels in a given tone are jammed together so tightly that it forms one
solid, curving mass, as seen in Figure 3.
Examining Figure 3, we can tell the following. There is no to very little
solid blacks and solid whites. The number of pixels darker than mid tone
is greater than the number of pixels brighter than mid tone. The number of
pixels in the mid tones is far less than either the number of darker pixels or
the number of lighter pixels. Does this mean the image was improperly
exposed? No. We would not be able to tell if this were an improper
exposure unless we saw the image that produced it and we knew what the
photographer was trying to accomplish. From an exposure perspective, what
we can tell about this histogram is that neither highlights nor shadows were
clipped. So we would assume we should be able to see detail in most of the
image. In addition, because the curve formed by the mass of histogram bars
has no gaps, we would expect to see smooth gradations between tones in the
Figure 4 is an image with its actual histogram. From the histogram we can
tell there are no solid whites, most of the image is darker than mid tone and
there is some slight clipping of blacks. But when we look at the image,
the clipped blacks are in the deep shadows of the trees, so this is not a
Figure 4A. The image ...
Figure 4B. ... and its histogram
When an image is over or under exposed, its histogram will show a concentration
of pixels to the left if under exposed and to the right when over exposed.
When the concentration is bunched up against the edge of the histogram, we know
that significant detail has been lost in the image. This loss of detail is
referred to as clipping. Figure 5 below shows the same image with major
loss of detail. This is reflected in the image's histogram.
Figure 5. Over/under exposure ...
... appears as clipping in a histogram
Gaps is the term used to describe the situation where an image's tonal range is
missing small sections. Gaps appear in a histogram as small vertical
slits, as seen on the left in Figure 6. Gaps are not necessarily bad.
However, if there are enough gaps and they are wide enough, the image will show
signs of banding, especially in broad areas such as sky and water. An
example of banding can be seen on the right in Figure 6. The green arrow
in the example marks the boundary where the transition between top and bottom is
not as smooth as could be.
Because of the difference in a printer's and a monitor's
ability to display tonal gradations, it is possible for banding to appear in
a print even if not visible on a monitor. This is one reason why
prints should be closely examined.
Figure 6. Gaps
If we look at the Channel drop down box in Figure 7, we will see RGB. This
means the histogram is plotting the tonal information of the composite color
channels. In Photoshop CS, we can use the histogram panel menu (see
yellow arrow in Figure 7) and select the All Channels View. In this view,
we will see four histograms for a RGB image. The top histogram is the
composite histogram. The histograms below the composite are for the
individual color channels.
As we work on an image, we will want to refresh the histogram. This is
done by clicking the Refresh icon
This icon is circled in Figure 7.