Specifications

COPYRIGHT © 1999 CANON INC. CANON SUPER G3 FAX BOARD-F1 REV.0 MAY 1999 PRINTED IN JAPAN (IMPRIME AU JAPON)
2-41
CHAPTER 2 OPERATION AND TIMING
3. Coding an Image
In the single progression sequential bi-level image compression method, a coding device is used
to code an image based on the result of comparison between the current line and the immediately
preceding line and on the result of prediction about the level of the pixel in question using a model
template (to find out whether it will be white or black).
The learning table used for prediction is corrected each time the model template moves to the
next pixel, thereby improving the rate of correct prediction. In this method of coding, a correct
prediction will not increase the size of coding. In this respect, it is highly important that the learning
table improve itself as it processes data for correct predictions.
Figure 2-402 is a diagram of the coding procedure used.
Figure 2-402 Coder and the Flow of JBIG Coding
1. In the typical prediction block, the current line and the immediately preceding line are compared
to find out if the latter matches the former. According to the result of this process, a flag (1 bit; 0
for match, 1 for non-match) is attached to the head of each line.
If a match was identified, the flag (only) is coded by the arithmetic coding block as a pseudo
image, without coding the pixels of the line being processed at the time. If a non-match was
identified, the pixels of the line being processed at the time are coded by the arithmetic coding
block using the model template and the learning table and based on the result of compression
between the prediction (white or black) of the pixels in question and the actual levels.
Image data input
Compares current line and
immediately preceding line
Predicting for
pixel in question
Coder
Typical prediction Arithmetic coding
Flag output (match/non-match)
Actual pixel in question
(for comparison against prediction)
If non-match
Pixel pattern (10 bits)Model template
Prediction for pixel in question
in relation to image pattern (10 bits)
Prediction for next in relation to
result of prediction
Learning table
Refers to prediction and
status
Result of prediction
Status for next based on result of
prediction and Reference to predic-
tion (learning)
Reference of area for
status in question
Rate of probability
table
Image data outlet
(coded by JBIG)