Specifications
CHAPTER 2 OPERATION AND TIMING
2-48
COPYRIGHT
©
1999 CANON INC. CANON SUPER G3 FAX BOARD-F1 REV.0 MAY 1999 PRINTED IN JAPAN (IMPRIME AU JAPON)
Figure 2-407 Conceptual Diagram of Arithmetic Coding
When the prediction is correct, the present data will have the same position on the numerical line
as the previously coded data, characterizing the arithmetic coding method by the absence of the need
for an additional bit used to express the position. This fact helps increase the rate of compression if
correct predictions continue (i.e., not resulting in an increased volume of coding).
If wrong predictions continue, on the other hand, additional bits will be needed to express the
positions of wrong predictions, causing the volume of coding to increase, while at the same time
decreasing the rate of compression.
The learning table continues to learn and correct table parameters while coding takes place so as
to make correct predictions and increase the rate of compression without increasing the volume of
coding.
Reference:
*
1
The actual rate of probability varies depending on the applicable status, since it is affected
by the share of LSZ in the area A.
*
2
The actual area is 8000 to 10000 (hexadecimal).
If the prediction is correct, the area A will be as follows:” A1 = 10000 (hexadecimal) -
LSZ, A2 = A1 - LSZ, A3 = A2 - LSZ.
If the prediction is wrong, the area A will be equal to LSZ.
*
3
Actually, it is 8000 (hexadecimal).
*
4
Actually, the hexadecimal value is shifted to the left and doubled so that it will be 8000 or
higher (hexadecimal).
Position (binary)
Area A1
LPS
MPS
LPS
MPS
LPS
MPS
LPS
MPS
Area A2
Area A3
0.00111.000
0.111
0.110
0.101
0.100
0.011
0.010
0.001
0
0.001010
0.001001
0.00100010
0.0010
Pixel in question for
coding
Position of code of
each pixel by arithmetic coding
1st
Correct
2nd
Correct
3rd
Correct
4th
Correct
5th
Correct
6th
Correct
7th
Correct
8th
Correct
Stan-
dardized
area