Encoding Method - Canon FAX L1000 Service Manual

3.2.3 Encoding Method

In the JBIG encoding used in the Single Progression Sequential Bi-level Image Compression
Method, uses in the encoder shown below to encode to the original the results of comparison
of the line currently being processed and the previous line, as well as the predicted value of an
image pixel (white or black) used in a model template.
The study table used in the prediction makes the next prediction more accurate by learning
and correcting the study table every time the model template moves to the adjoining pixel. It
is characteristic of this method that if the prediction is accurate the amount of encoding will
not increase, and if the prediction is off the amount of encoding increases, so the increase in
prediction accuracy of this study table is very important.
An outline of the encoding procedure is shown below.
Comparison of current
and previous lines
Image data
TYPICAL
input
PREDICTION
Prediction of pixel
to be encoded
TEMPLATE
Figure 3-42 Encoder and Flow of JBIG Encoding
a) In the pattern prediction section, the line currently being processed and the current line are
compared, and judged to match or not match. A flag showing whether or not the lines
match (1 bit, 0: match, 1: don't match) is attached to the head of each line according to this
judgment. When the lines match, only this flag is encoded in the arithmetic encoding
section as a suspected pixel, the pixel of the line being currently processed is not encoded.
When the lines do not match, the pixel of the line currently being processed is encoded in
the arithmetic encoding section based on the results of a comparison of the value of the
actual pixel and the pixel (white or black) which is predicted using the model template and
the study table.
When the lines are judged to match, the line currently being processed is said
to be "typical". When the lines are judged to not match, the line is said to be
"not typical". When the very first line of an image is predicted, the
NOTE
background color is used as the previous line.
ENCODER
Output of match/no match
No-match path
Pixel to be encoded
(For comparison against prediction)
MODEL 10-bit pixel
pattern
Prediction of pixel using
10-bit pixel pattern
STUDY
TABLE
Refer to the next status and pixel prediction
based on the prediction results (Learning)
Next prediction based on
prediction results
PROBABILITY
ESTIMATION
TABLE
3-93
FAX-L1000 Chapter 3: Technical Reference
ARITHMETIC
ENCODING
Prediction and status reference
and predition results
Relevant status area reference
Output of image data
encoded by JBIG