In the classical mean value approach the simple scale factor
is applied to all values. The next step is to clip all
values larger than 1 to 1 and those less than 0 (usually there are no
such values) to 0. This approach clips the values far from the
average. Actually, all values that are more than twice as large as
are clipped to 1, which means that some interesting
details, may not be displayed. By varying the contrast and the
aperture this problem can be successfully solved. Let us define the
contrast as the ratio between the largest and the smallest luminance
value in a float image. The contrast window on the log scale always
has the same size for a given contrast and this is one more reason for
using a logarithmic histogram. The original contrast of a float image
can be anything from as low as 10 or less (although such low-contrast
images are rare) to as high as 1000 or more. Using a larger contrast
interval, less pixels will be clipped, but when they are mapped to the
input values, a low-contrast image will be generated. On the other
hand, the use of a small contrast clipping interval will produce an
image which is considered as a high-contrast image. (See results
chapter for visualisation of this confusing fact). In figure
4.1 various contrast windows and aperture settings are
displayed, with the logarithmic histogram of the image shown in
results section, color plates 3 and 4.
Figure 4.1: Various contrast and aperture settings