Databases Reference
In-Depth Information
It Will Soon Be June 4
That's Ruby's Birthday!
F I GU R E 7 . 18
The mask layer.
This area not
coded or sent.
F I GU R E 7 . 19
The foreground layer.
textual data in the stripe. Two-layer stripes (2LS) only contain two layers, with the third set to
a constant value. This kind of stripe would be useful when encoding a stripe with multicolored
text and no images, or a stripe with images and bi-level text or line drawings. The third kind
of stripe is a one-layer stripe (1LS), which would be used when a stripe contains only bi-level
text or line art or only continuous tone images.
Once the document has been partitioned it can be compressed. Notice that the types of
data we have after partitioning are continuous tone images, bi-level information, and multi-
level regions. We already have efficient standards for compressing these types of data. For
the mask layer containing bi-level information, the recommendation suggests that one of sev-
eral approaches can be used, including modified Huffman or modified READ (as described in
recomendation T.4), MMR (as described in recommendation T.6), or JBIG (recommendation
T.82). The encoder includes information in the datastream about which algorithm has been
used. For the continuous tone images and the multilevel regions contained in the foreground
and background layers, the recommendation suggests the use of the JPEG standard (recom-
mendation T.81) or the JBIG standard. The header for each slice contains information about
which algorithm is used for compression-3.
7.9 Summary
In this chapter, we have examined a number of ways to compress images. All of these ap-
proaches exploit the fact that pixels in an image are generally highly correlated with their
neighbors. This correlation can be used to predict the actual value of the current pixel. The
prediction error can then be encoded and transmitted. Where the correlation is especially high,
as in the case of bi-level images, long stretches of pixels can be encoded together using their
