...Approach
Support for this work was provided by NSF grant IRI-9306454, ONR grant N00014-93-I-0221, Sverige-Amerika Stiftelsen and the Fulbright Commission
.
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  
...Jägersand
Also Chalmers, 412 96 Göteborg, Sweden
.
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  
...ranges
This is not true of most natural scenes [5]. In fact a scale study of the fractal dimension of an image has been used to determine whether it contains man-made objects.
.
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  
...image.
What we consider to be the underlying event for our distribution is really not that important. One way of thinking of p(x) is as the distribution function for the probability that a light quantum hits position x given exactly one light quantum hitting some position in the image.
.
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  
...them
Note that the x-axis is reversed to show resolution lengths, as in the information distribution plots, rather than spatial frequency.
.
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  .
  

jag@cs.rochester.edu