Browse Prior Art Database

Improved Error Diffusion On Imaging Systems

IP.com Disclosure Number: IPCOM000100776D
Original Publication Date: 1990-Jun-01
Included in the Prior Art Database: 2005-Mar-16
Document File: 5 page(s) / 205K

Publishing Venue

IBM

Related People

Todd, SJ: AUTHOR

Abstract

Techniques are described to improve error diffusion on imaging systems either by roughing smooth areas or by extending Peano Curves. These techniques are applicable to outputting color or black and white continuous tone images on limited capability devices, such as the IBM 5080 and PS2 graphics adapter. They are applicable to imaging systems such as IAX (Image Access Executive) and rendered output of solids of WINSOM (Winchester Solid Modeller) and CAEDS (Computer Aided Engineering & Design System). Roughing smooth areas for improved error diffusion eliminates the regime caused when the required input is equal to available output over a significant area. Extending Peano curves for improved error diffusion eliminates the more sophisticated pattern regimes. ROUGHING SMOOTH AREAS

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 28% of the total text.

Improved Error Diffusion On Imaging Systems

       Techniques are described to improve error diffusion on
imaging systems either by roughing smooth areas or by extending Peano
Curves.  These techniques are applicable to outputting color or black
and white continuous tone images on limited capability devices, such
as the IBM 5080 and PS2 graphics adapter.  They are applicable to
imaging systems such as IAX (Image Access Executive) and rendered
output of solids of WINSOM (Winchester Solid Modeller) and CAEDS
(Computer Aided Engineering & Design System).  Roughing smooth areas
for improved error diffusion eliminates the regime caused when the
required input is equal to available output over a significant area.
Extending Peano curves for improved error diffusion eliminates the
more sophisticated pattern regimes. ROUGHING SMOOTH AREAS

      Two solutions are described.  One adds a small amount of random
noise to the image before the output level is chosen.  The other
counts the number of consecutive occurrences of a single output, and
uses the next smaller output when this count exceeds a threshold.
The second solution is preferred.

      The core of regular error diffusion works in the following
order on each pixel:
-    given input value i to be encoded
-    given error e
-    compute diffusion value d = i + e
-    find the nearest available output value o
-    output the value o
-    carry error e = d - o to the next step

      Where i is equal to an available output o over a significant
area, the error e is carried over, and the output o is used for each
point in the area.  Thus the area is smooth in the output.
Smoothness is generally desirable. However, when the smooth area is
surrounded by areas that have been patterned by a mix of different
values, the smoothness stands out, and emphasizes the roughness.  If
noticeable banding of the overall picture results, smooth areas
should be prevented.

      Random Noise The following variant of the algorithm is used:
-    given input value i to be encoded
-    given diffused error e
-    compute diffusion value d= i + e
-    add noise dx = d + rand
-    find the nearest available output value ox to dx
-    output the value ox
-    carry error e = d - ox to the next step

      This is effectively mixing the technique of error diffusion
with another standard technique of dithering.  The scale of the
random noise should be sufficient that a small proportion of ox
values (about 10 percent) differ from the value o that would be used
with no randomization.  High frequency noise gives better results
than white noise when pure dithering is used.  The use of high
frequency noise does not noticeably improve results when used as an
adjunct to error diffusion.  The error carrying is based on the
diffusion value d before noise is added and prevents the noise having
a cumulative effect on the average grey level.

      Smooth Are...