Page 40 - Digital Analysis of Remotely Sensed Imagery

Overview 13

is known as row or line, and 10 as column, position, or pixel. This
convention of representation is not universally adhered to, so it can
vary with the image processing system. Of particular note is that the
first row and last row are counted in determining the number of pixels/
columns of an image. Also, the first row or column can start from 0 as
well as from 1.
As with all raster data, the coordinates of pixels in an image
are not explicitly stored in the computer except for a few strategic
ones (i.e. the four corner pixels). Instead, all pixels are recorded
sequentially by column first and by row next as a long list. Their
geographic location is implicitly defined by their relative position in
the list or their distance from the origin (i.e., the first pixel). This
relative position can be converted into a pair of absolute coordinates
expressed as row and column from this distance as well as the physical
dimension (e.g., number of rows by number of columns) of the image.
These coordinates may be further converted into the metric expression
by multiplying them by the spatial resolution of the image.

1.4.4 Histogram
A histogram is a diagram displaying the frequency distribution of
pixels in an image with respect to their DNs (Fig. 1.5). It can be
presented either graphically or numerically. A graphic histogram
contains two axes. The horizontal axis is reserved for the pixel’s DN.
It is an integer with an increment of 1 or other larger integers specified
by the analyst. Thus, the histogram is not smooth but discrete. The
vertical axis represents the frequency, in either relative terms
(percentage) or absolute terms (actual number of pixels). A graphic
histogram is an effective means of visualizing the quality of a single
spectral band directly. For instance, a broad histogram curve signifies
a reasonable contrast while its position relative to the horizontal axis
is indicative of the overall tone of the band (Fig. 1.5a). A position
toward the left suggests that the image tends to have an overall dark
tone, a phenomenon equivalent to underexposure in an analog aerial
photograph (Fig. 1.5b). On the other hand, a position toward the right
shows that the image has a bright tone throughout, with an appearance
similar to an overexposed aerial photograph. Unlike a graphic
histogram, a numeric histogram displays the exact number of pixels at
every given DN level. In order to reduce the number of DN levels, a
few DNs may be amalgamated. In this case, the frequency refers to
the combined pixels over the indicated range of DNs. Both forms of
histogram are essential in contrast manipulation of spectral bands. A
preview of a graphic histogram enables the analyst to prescribe the
kind of enhancement method most appropriate for the image. A
numeric histogram provides important clues in deciding critical
thresholds needed in performing certain kinds of image contrast
stretching.