Specifications
COPYRIGHT
©
1999 CANON INC. CANON SUPER G3 FAX BOARD-F1 REV.0 MAY 1999 PRINTED IN JAPAN (IMPRIME AU JAPON)
2-47
CHAPTER 2 OPERATION AND TIMING
e. When the level of prediction has been found to be correct or wrong in reference to the actual
pixel, model template, and rate of prediction table, the arithmetic coding block codes the
result of prediction so that coded image data is generated and transmitted.
f. Unlike the conventional MH or MR coding method, coding (arithmetic) by the arithmetic
coding block does not use a conversion table. It rather uses LSZ (area, indicating an estimate
of making a wrong prediction), and codes data by expressing the progress of making correct
prediction in terms of a numerical line (between 0 and 1.0). Codes expressed on this
numerical line, as shown in Figure 2-407, will be under MPS if the prediction is correct and
under LPS if the prediction is wrong.
This way of arithmetic coding, further, holds the concept of an “area” (A). This area (A)*
2
represents the area (A) of MPS if the prediction is correct for each pixel; it represents the area
of LPS if the prediction is wrong.
When this area (A) grows smaller than a specific area*
3
, the initial bit (except the code ‘0’
indicating the position on a numerical line) is moved to the left as a coded image data unit for
output.
At this time, an area (A) smaller than the specific area in question cannot express a position
finer than it presently can, requiring enlargement*
4
for better expression of a position. This
process is called “standardization,” and the area (A) is set to a size larger than the specific area
in question*
3
.
The following is a concept of arithmetic coding:
Here the following is assumed:
· The rate of making correct predictions is 50%, and the rate of making wrong predictions
is 50%.*
1
· The area of correct predictions is MPS, and the area of wrong predictions is LPS.