Probably the most widely used mapping method today is the mean value
mapping technique. The idea is to map the average radiance to 0.5
input value, and then clip the values larger than 1 to 1:
where 
is the average radiance value, and n is set to 1
if n>1.
According to the above equation, the value 
will be
mapped to 1 and all values larger than will be
clipped to 1. Obviously the information in the range 
is lost, although there can be some interesting
details. Another problem is the case when few very high radiance
values increase the average too much, making the final image too
dark. There is also a problem that arises from the fact that the
global illumination solution is linear in source radiance
[ArKi90], that means results for any two light source strengths
are directly proportional. Therefore the mean value mapping will
produce the same images for various light sources' strengths. This
problem cannot be overcome unless absolute units are known. Using
fictitious units makes it impossible to know whether the scene is
supposed to be well lit or dark. Another drawback of the mean value
mapping technique arises when the average scene reflectance is very
low or very high. Imagine an image representing a heap of coal. If the
average value (which is low) is mapped to 0.5 the whole image will be
too bright. On the other hand, an image of snow covered mountains will
be too dark. In spite of all these drawbacks this is still the most
widely used mapping method today. Most of today's rendering software
can not produce absolute unit raw images, and most scenes are not very
dark or very bright. These two facts make it possible to use mean
value mapping successfully for most renderings. Of course, when an
appropriate lighting atmosphere or the correct objects' colors are
important, some other mapping should be used. Result images rendered
using the mean value mapping method are shown in results chapter,
color plates 1a, 5a, 9a, 9b, 11a, 14a, and 14b.