Neighborhood Parallel Access Bit Map Memory

Neighborhood Parallel Access Bit Map Memory

      Neighborhood image processing requires computation with
neighboring pixel data to produce a modified value of the central
pixel.  A 3 by 3 neighborhood is typically used, as shown in Fig. 1,
with the neighborhood defined by pixels labelled "A - H" and the
central pixel labelled "*."

      Classic random access to this neighborhood is accomplished by
multiple reads to the image memory containing this neighborhood.
Classic sequential access to this neighborhood is accomplished by
multiple raster delay line auxiliary storage, creating the
neighborhood at the end of a pipeline.

      The method described below avoids the time-consuming multiple
reads of the random-access technique and the time consuming pipeline
latency of the sequential technique.  The entire neighborhood is
accessed in one read using the address of the central pixel.  This
requires that the neighboring pixel data be resident in their
separate but simultaneously operating memories.  The maximum size of
the neighborhood determines the number of separate memories.  An
addressing and data steering scheme is required to avoid the need for
redundant data storage capacity.

      Fig. 2 illustrates the architecture of the disclosed image
memory.  A neighborhood 10 (or only a central pixel generally) is
written into individual RAMs 20, which could be segments of a single
RAM.  The neighborhood is then selected and addressed by RAM and MUX
(multiplex er) address generation ROMs 30-37 which are driven by the
central pixel image horizontal and vertical address.  The
neighborhood associated with the same (or other) c...