This program divides the pixel being processed by a low pass filter. This has the general effect of highlighting the edges of grouped contrasted DN values in the image, and therefore, feature changes will be emphasized. Predominant changes will be emphasized with larger boxcars, and smaller changes can be searched for using smaller boxcars.

This is done by convolving an NxM boxcar through the data, where N and M are odd integers. In other words, the average DN value of the boxcar is divided into its middle pixel DN value. Please refer to the following diagram:

Thus, P(output) = P(input) / lowPassFilter, or P(s,l) = P(s,l) / average(s,l,N,M) where P is the target pixel, s and l are the sample and line position in the boxcar, N and M are the size of the boxcar (in this case 3x3) and average(s,l,N,M) is the average of the NxM centered box.

For details on the box filtering see the following references:

      M. J. McDonnell, Box-Filtering Techniques, Computer Graphics
      and Image Processing, Vol. 17, 1981, pages 65-70

      E. M. Eliason and L. A. Soderblom, An Array Processing System
      for Lunar Geochemical and Geophysical Data,
      Porc. Lunar Sci. Conf. (8th) , 1977, pages 1163-1170.