Databases Reference
In-Depth Information
B
A
a
b
c
d
e
f
C
X
g
h
i
F I GU R E 7 . 10
Pixels used to determine the value of a lower-level pixel.
black pixel causes a severe loss of detail, and randomly replacing with a black or white pixel
introduces a considerable amount of noise into the image [
92
].
Instead of simply taking the average of every 2
2 block, the JBIG specification provides
a table-based method for resolution reduction. The table is indexed by the neighboring pixels
shown in Figure
7.10
, in which the circles represent the lower-resolution layer pixels and the
squares represent the higher-resolution layer pixels.
Each pixel contributes a bit to the index. The table is formed by computing the expression
×
4
e
+
2
(
b
+
d
+
f
+
h
)
+
(
a
+
c
+
g
+
i
)
−
3
(
B
+
C
)
−
A
If the value of this expression is greater than 4.5, pixel
X
is tentatively declared to be 1. The
table has certain exceptions to this rule to reduce the amount of edge smearing, generally
encountered in a filtering operation. There are also exceptions that preserve periodic patterns
and dither patterns.
Redundancy Removal
There are two kinds of redundancy removal available in the JBIG standard. These are known as
typical prediction
and
deterministic prediction
. Typical prediction exploits the fact that many
bi-level images have large regions that are the same color. This being the case, knowledge of
the low-resolution pixel value is sufficient to determine the value of the high-resolution pixel
for the progressive modes. Deterministic prediction operates on the concept that given the fact
that the low-resolution images were constructed from the high-resolution images, knowledge
of the values of the low-resolution pixels is sufficient to determine the high-resolution pixels
exactly under certain conditions.
Typical Prediction
Figure
7.11
indicates the relationship between various high- and low-
resolution pixels. In the figure the squares correspond to the high-resolution pixels and the
circles correspond to low-resolution pixels. Notice that there is one low-resolution pixel
corresponding to each of the four
high-resolution pixels.
Many of the pixels in a bi-level document occur in regions of constant color, and many of
the pixels in a document image occur in regions of white pixels. For these regions, pixels in
both high-resolution and low-resolution images have the same color. Therefore, if the low-
resolution pixels are available, the high-resolution pixels need not be transmitted. The typical
prediction is a labeling of the high-resolution pixels that do not need to be transmitted. The
(
2
×
2
)
