JPH0591343A - Binary data encoding method and binary data decoding method - Google Patents

Abstract

PURPOSE:To provide the binary data encoding method and the binary code decoding method which efficiently compress a specific document picture or the like of a FAX or the like. CONSTITUTION:The appearance probability of symbols is estimated by a symbol appearance probability estimating means 1 in accordance with the symbol pattern of a symbol sequence to be encoded. A code space reduction means 6 reduces the area of the code space, where the symbol sequence to be encoded is mapped, with the reduction rate where partiality peculiar to the application appearing in position coordinates of the code space is taken into consideration. A decoding operation means 2 maps the symbol pattern in the numeric line [0,1] section of the reduced code space area in accordance with the appearance probability of symbols estimated by the means 1 and its value.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION The present invention relates to a facsimile machine (F
AX), etc., a binary data encoding method using an arithmetic encoding method for compressing a binary data series read from a document image including characters, figures, etc., or decoding the compressed data, and decoding thereof. It relates to the method of conversion.

[0002]

2. Description of the Related Art Conventionally, as a technique in such a field,
For example, some documents were described in the following documents. Reference; IBM J. RES. DEVELOP,32
[6] (1988-11) (US) R.A. B. Arps etc.
"A Multi-purpose Buyer SII Chip Fo
Adaptive Data Compression Observer
E-level Images (A multi-purpose VLSI chip fo
r adaptive data compression of bilevel image) ”
P. 775-795 The above-mentioned document describes a binary data sequence using an arithmetic coding method.
The technique of the binary data encoding method to compress is described.
It In this binary data encoding method, the following encoding is performed.
The binary data series is encoded by an algorithm.

For example, a length N generated from a binary information source.
Symbol sequence S _N = s ₁ s ₂ ... s _N (s _i = 0,1)
think of. Of the two symbols of the information source, the appearance probability of the inferior symbol (the symbol having a lower symbol appearance probability) LS is P _L , and the appearance probability of the superior symbol MS is P _M. The code is C and the code space length is A.

First, initialization is performed according to the following equation (1). C (λ) = 0 (λ; empty sequence) A (λ) = 1 (1) Then, the symbol sequence S _N is encoded by the following processing.

Currently, up to the i-1th symbol are processed, and for the symbol sequence S _i-1 = s ₁ s ₂ ... s _i-1 ,
If C (S _i-1 ) and A (S _i-1 ) are given, then C (S _i ), A after encoding the next symbol s _i
(S _i ) is determined by the following equations (2) and (3).

[0006]

[Equation 1]

The value of A is left-shifted so as to satisfy 0.75≤A≤1.5. At the same time,
The C register (code register) must also be left-shifted by the number of times the A register (code space length register) is shifted. This is called reproduction normalization (implementing floating-point arithmetic with a finite length register). After that, the value of the C register is shifted out, and the overflowed value becomes the code output.

[0008]

However, the conventional method has the following problems. In the binary data coding method using the arithmetic coding method, the coded symbol sequence is mapped to the coordinate values of the number line [0, 1) section. In the conventional method, all possible binary symbol sequences are mapped onto this number line [0,1) interval. That is, in the conventional method, the code space is assigned to all possible binary symbol sequences. However, in many cases not all possible symbol patterns are used.

In the conventional method, since the code space is allocated to the unused symbol patterns, there is waste in the code space, the coding limit is lowered, and the compression rate is lowered. There is. As a result, the memory capacity for storing the document image data is increased, and the transmission time is extended for transmission, which is difficult to solve.

The present invention provides a binary data coding method and a decoding method using the same, which solves the problem that the coding limit is lowered and the compression rate is lowered, as a problem that the above-mentioned prior art has. To do.

[0011]

In the arithmetic coding method, a symbol series to be coded is mapped to coordinate values of a number line [0, 1) section. In the conventional binary data encoding method, all possible binary symbol sequences are mapped on this number line [0,1) section, but in many cases,
Not all possible symbol patterns are used.

Therefore, in the binary data encoding method of the first invention, the symbol appearance probability estimating means estimates the appearance probability of the symbol from the symbol pattern of the symbol sequence to be encoded, and the code space reducing means performs the estimation. The area of the code space is reduced at a reduction rate in consideration of the application-specific bias that appears in the position coordinates of the code space to which the symbol sequence to be encoded is mapped. Thereafter, the code calculation unit maps the symbol pattern to the number line [0, 1) section of the reduced code space region according to the symbol appearance probability and its value estimated by the symbol appearance probability estimation unit. is doing.

In the binary data decoding method of the second invention, the symbol appearance probability is estimated from the data encoded by the binary data encoding method of the first invention, and then the symbol appearance appears. The binary data series is decoded by performing the inverse operation of the code operation means according to the probability and its value.

[0014]

According to the first aspect of the invention, since the binary data encoding method is configured as described above, the symbol appearance probability estimating means estimates the symbol appearance probability from the symbol pattern of the symbol sequence to be encoded. And sends it to the code calculation means. The code space reducing means reduces the area of the code space to which the symbol sequence to be encoded is mapped. The code calculation unit maps the symbol pattern to the number line [0, 1) section of the code space region reduced by the code space reduction unit according to the estimated appearance probability of the symbol and its value.

As described above, if the code space assigned to the symbol series such as characters and figures is reduced in advance and the code space is assigned to the symbol pattern by the code calculation means, the code space length (number line section) is reduced. The reduction ratio is smaller than that of the conventional method, and the number of digits after the decimal point of the code is smaller than that of the conventional method. Therefore, the code length becomes shorter,
The encoding limit is increased and the compression rate can be improved. Therefore, the above problem can be solved.

According to the second invention, the data coded by the binary data coding method of the first invention is input,
After estimating the symbol appearance probability, the inverse operation of the code operation means of the first invention is performed to decode the binary data series.
This enables accurate decoding of the compressed data.

[0017]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIGS. 2A to 2C are diagrams for explaining the principle of an embodiment of the present invention. In the binary data encoding method of the present embodiment, when a code space is assigned to a symbol pattern appearing in a specific application such as a FAX document (character / figure), the position coordinates of the code space are:
The bias peculiar to the application is shown.

For example, as shown in FIG. 2A, it is assumed that the symbol "1" is arranged in the lower part of the code space and the symbol "0" is arranged in the upper part of the code space. When a symbol string which is a basic constituent element of a character / graphic as shown in FIG. 2 (b) is coded, the code space allocated to the first '1' symbol of the symbol string is allocated to the symbol string. In many cases, the code space to be used is a diagonal position as shown in FIG.

Utilizing this property, the code space assigned to the symbol series such as characters and figures is reduced in advance, so that the coding performance is improved and the compression rate is improved as compared with the conventional case. Furthermore, in the binary data decoding method of this embodiment,
By performing the inverse calculation process of such an encoding method, the binary data series is accurately decoded.

That is, in the binary data encoding method of this embodiment, instead of the conventional code arithmetic expressions such as the following expressions (4) and (5),

[0021]

[Equation 2]

The following code arithmetic expressions (6) and (7) are used.

[0023]

[Equation 3]

In the code arithmetic expressions (6) and (7), the rate of reduction of the code space length A is smaller than that of the conventional methods (4) and (5), and the number of digits after the decimal point of the code is the conventional method. And the code length becomes shorter.

Further, the code constituted by the code arithmetic expression of this embodiment is uniquely decoded by the binary data decoding method of this embodiment. This is guaranteed from the reduction condition of the code space assigned to the symbol sequence described later.

The binary data encoding method in the first embodiment and the binary data decoding method in the second embodiment will be described in detail below.

First Embodiment FIG. 1 is a functional block diagram of an encoder for carrying out a binary data encoding method showing a first embodiment of the present invention. The symbols used in the following description are the same as those used in the description of the prior art.

This encoder is an individual circuit using an integrated circuit or the like, or a computer or a digital signal
It is configured by program control of a processor (DSP) or the like, and has a symbol appearance probability estimating means 1 for estimating a symbol appearance probability from a symbol input. On the output side of the symbol appearance probability estimating means 1, the code calculating means 2 is provided.
And A value calculation means (code space length calculation means) 3 are connected. The code calculation unit 2 has a function of performing a predetermined code calculation according to the symbol appearance probability estimated by the symbol appearance probability estimation unit 1 and its value. The A value calculation means 3 has a function of calculating the size of the code space according to the value of the symbol to be encoded. Sign calculation means 2
A C register (code register) 4 for outputting a code is connected to the output side of the
A code space reduction unit 6 and a comparison calculation unit 7 are connected via an A register (code space length register) 5. The code space reduction unit 6 has a function of reducing the code space, and the comparison calculation unit 7 has a function of determining whether or not the value of the A register 5 is smaller than a comparison value (for example, 0.5). is doing.

The encoder also includes a "0" symbol counter (SC0) 8 for counting the number of "0" symbols that appear after the "1" symbol based on the symbol input.
And a "1" symbol counter (SC1) 9 that counts the number of "1" symbols that appear after the "0" symbol. The output sides of the comparison calculation means 7, the “0” symbol counter (SC0) 8 and the “1” symbol counter (SC1) 9 are connected to the control means 10. The control means 10 has a function of controlling each of the functional means 1 to 9 via a plurality of control lines 10a according to a sequence of an encoding processing flow described later.

Next, the general operation of FIG. 1 will be described. The symbol sequence to be encoded is a symbol appearance probability estimation unit 1, a code calculation unit 2, a “0” symbol counter (SC
0) 8 and the '1' symbol counter (SC1) 9 are input, the symbol appearance probability estimation means 1 uses the binary symbol s _i (1 ≦ i ≦ N, which is to be encoded).
N: number of symbols) symbols s ₁ s ₂ that appeared before
The probability P (0) that the symbol s _i is a “0” symbol and the probability P (1) that it is “1” are obtained from s _i−1 . Such an estimation result is sent to the A value calculation means 3 and the sign calculation means 2. The A value calculation means 3 calculates the size A (s _i ) of the code space according to the following expression (8) according to the value of the symbol s _i to be encoded. A (S _i ) = P (1) (when s _i = 1) A (S _i ) = A (S _i−1 ) −P (1) / 2 (when s _i = 1) ... ( 8) The calculated code space size A (S _i ) is temporarily stored in the A register 5 and then fed back to the A value calculation means 3 and also to the code space reduction means 6 and the comparison calculation means 7. Sent.

In the code space reducing means 6, the symbol "1" is given.
For the size of the code space assigned to the first "1" symbol in a continuous "1" symbol run, the last "1"
The size of the code space allocated up to the symbol is 1/2
The code space is reduced so that, and the reduction result is temporarily stored in the A register 5.

The sign calculation unit 2 which has input the estimation result of the symbol appearance probability estimation unit 1 performs the sign calculation of the following expression (9) according to the value s _i of the symbol. C (S _i ) = C (S _i-1 ) (when s _i = 1) C (S _i ) = C (S _i-1 ) + P (1) / 2 (when s _i = 0) (9) This operation result is temporarily stored in the C register 4, then sequentially updated and finally output as encoded data.

When the value of the A register 5 is sent to the comparison calculation means 7, the comparison calculation means 7 changes the value of the code space length A to A.
It is determined whether ≦ 0.5, and the determination result is sent to the control means 10. In the control means 10, the comparison calculation means 7,
By inputting the values of the '0' symbol counter (SC0) 8 and the '1' symbol counter (SC1) 9, each functional means is controlled according to a sequence of a predetermined encoding processing flow,
The encoded data is output from the C register 4.

FIGS. 3 (a) and 3 (b) are encoding processing flows for explaining the binary data encoding method of the present embodiment using the encoder of FIG. 1, see this figure. Meanwhile, the encoding processing procedures (1) to (4) will be described.

(1) Step 1 (Processes 101 to 103) In order to perform the initialization process, the symbol number i is set to 1 in the process 101 of FIG. 3A, and the code space length A is set in the process 102.
Is set to 1, and the code C is set to 0. Further, in process 103,
The values of the “0” symbol counter (SC0) 8 and the “1” symbol counter (SC1) 9 are set to 0.

(2) Step 2 (Processing 104) In order to obtain the symbol appearance probability, the symbol appearance probability estimating means 1 has a probability P (0) that the symbol Si becomes 0 and a probability P (1) that the symbol Si becomes 1 in the process 104. ).

(3) Step 3 (Processes 105 to 121) In this step 3, the code calculating means 2 calculates the code C and the A value calculating means 3 calculates the code space length A. That is, in process 105, it is determined whether or not the symbol Si is 0. When it is 0, the process proceeds to process 106, and when it is not 0, the process proceeds to process 113.

In process 106, the "0" symbol counter (SC0) 8 is incremented by 1, and in process 107 it is determined whether the "1" symbol counter (SC1) 9 is 1. SC0 = 1
In the case of, the code space reducing means 6 causes the processing 108,
At 109, code space reduction processing is performed. Next, after renormalization processing is performed in processing 110, arithmetic processing 111 is performed on the first “0” symbol following the “1” symbol, and arithmetic processing 112 is performed on the other “0” symbols. 3B, the renormalization process 118 of FIG. 3B is performed, and in process 119, it is determined whether Si = 0.

On the other hand, if Si = 0 is not found in the process 105, a '1' symbol counter (SC
1) +1 is added to 9 and the probability P (1) of becoming 1 in processing 114 is stored in the A register 5, and then renormalization processing 115 is performed. Then, in process 116, it is determined whether or not SC1 = 1. If yes, in process 117, the contents of the A register 5 are shifted, and the process proceeds to the determination process 119 of FIG.

When Si = 0 in the judgment processing 119, SC1 is cleared to 0 in processing 120, and SC0 is cleared to 0 in processing 121 when Si = 0 is not satisfied.

(4) Step 4 (Processes 122 and 123) After clearing SC0 and SC1 to 0 in processes 120 and 121, the symbol number i is incremented by 1 in process 122, and process 123 is performed.
Then, it is determined whether the number of symbols N is smaller than the symbol number i. When N <i, the symbol encoding process is terminated. If the symbol encoding process is not terminated, the symbol is passed through the connector. Then, the process returns to the process 104 of FIG.

Table 1 shows the results of comparative evaluation of the binary data coding method of this embodiment and the conventional coding method by computer simulation. The document image used for this evaluation is CCITT (International Telegraph and Telephone Consultative Committee) test chart No. 1-No. It is 6.

[0043]

[Table 1]

It can be seen from Table 1 that the compression rate of the encoding method of this embodiment is considerably higher than that of the conventional method.
Therefore, if the encoding method of the present embodiment is used, a document image such as a FAX document (character / figure) can be compressed with high efficiency, thereby reducing the memory capacity in storing the document image data and transmitting time in transmission. Can be shortened. Second Embodiment FIG. 4 is a functional block diagram of a decoder for carrying out a binary data decoding method showing a second embodiment of the present invention, in which elements common to those in FIG. Are assigned common reference numerals.

As in the case of FIG. 1, this decoder is configured by an individual circuit using an integrated circuit or the like, or by program control using a computer, DSP or the like, and is the reverse of the encoder in FIG. The calculation is performed to decode the encoded data.

That is, this decoder has a symbol appearance probability estimating means 1, an A value calculating means 3, a C register 4, an A register 5, a code space reducing means 6, a comparison operating means 7, and 0 which are the same as those in FIG. 'Symbol counter (SC0) 8 and' 1 '
In addition to having the symbol counter (SC1) 9, a decoding calculation means 12 is provided instead of the code calculation means 2.
Further, control means 20 for controlling each of these functional means
Has a function of controlling each functional means via a plurality of control lines 20a according to a sequence of a decoding processing flow described later.

The decoding calculation means 12 has a function of performing the decoding calculation of the following expression (10) according to the value of the symbol s _i . C (the case of _{s i = 0) (S i} -1) = C (S i) (s i = the case of _{1) C (S i-1} ) = C (S i) -P (1) / 2 · .. (10) FIGS. 5 (a) and 5 (b) are a decoding process flow for explaining the decoding method of the present embodiment using the decoder of FIG. 4, with reference to this figure. , Decoding procedure (1) ~
(4) will be described.

(1) Step 1 (Processes 201 to 204) In order to perform the initialization process, the code C is shifted in in the process 201 of FIG. 5A, and the symbol number i is set to 1 in the process 202.
In process 203, the code space length A is set to 1, respectively. Further, in process 204, the '0' symbol counter (SC
0) 8 and the '1' symbol counter (SC1) 9 are set to 0 respectively.

(2) Step 2 (Process 205) In order to obtain the symbol appearance probability, the symbol appearance probability estimation means 1 uses the code input, and in the process 205, the probability P (0) that the symbol Si becomes “0”, The probability P (1) of "1" is obtained.

(3) Step 3 (Processes 206 to 224) In process 206, the decoded symbol is judged, and if it is judged that the decoded symbol is "1", a series of coding operations from process 207 are carried out. .. That is, in process 207, SC
1 is incremented by 1, P (1) is stored in the A register 5 in process 208, and renormalization process 209 is performed. Next, in process 210, it is determined whether or not SC1 = 1. If yes, process 211 shifts the contents of the A register 5 and process 212
Then, "1" is decoded as the symbol Si, and the process proceeds to the process 222 of FIG. 5B via the connector.

On the other hand, in process 206, the decoded symbol is "0".
If it is determined that, the decoding calculation means 12 and the like perform a series of decoding operations from the processing 213. That is, in processing 213, SC0 is incremented by 1, and in processing 214, it is determined whether SC0 = 1 or not. 217 is performed. Next, regarding the first “0” symbol following the “1” symbol, the arithmetic processing 21
Perform 8. For the other '0' symbols, arithmetic processing 219 is performed.

After the calculation processing of the processing 218 or 219,
The renormalization process 220 is performed, and the renormalization process 220 is performed, as shown in FIG.
In processing 221 of (b), '0' is decoded as the symbol Si, processing 222 determines whether Si = 0 or not. If YES, processing SC 223 clears SC1 to 0, and if NO, processing 224. Clear SC0 to 0 with.

(4) Step 4 (Process 205, 206) In process 225, the symbol number i is incremented by 1, and in process 226 C
It is determined whether or not = 0, and if yes, the decoding process ends, and if the decoding process has not ended, the process returns to the process 205 of FIG. 5A via the connector.

As described above, in the decoding method of this embodiment, the coded data can be accurately decoded into the symbol series by performing the inverse operation of the coding method of the first embodiment.

The present invention is not limited to the above embodiment,
For example, the encoding processing flow of FIG. 3 and the decoding processing flow of FIG. 5 are changed to another processing procedure, and further, the encoding processing and the decoding processing thereof are performed. Various modifications are possible, such as changing the configuration of the above to another configuration.

[0056]

As described in detail above, according to the first aspect of the present invention, when a code space is assigned to a symbol pattern that appears in a specific application, the position coordinate of the code space shows a bias peculiar to the application. By utilizing the property, the code space assigned to the symbol series such as characters and figures is previously reduced by the code space reducing means, so that the symbol series can be encoded with the encoding limit exceeding the Markov entropy limit, and the document Images can be compressed with high efficiency.
Therefore, by using the encoding method of the present invention, it is possible to reduce the memory capacity for storing document image data and the like and shorten the transmission time for transmission.

According to the second invention, the coded data coded according to the first invention is accurately converted into a binary data series by performing an operation process or the like which is the reverse of the coding method according to the first invention. Can be decrypted.

[Brief description of drawings]

FIG. 1 is a functional block diagram of an encoder for implementing a binary data encoding method showing a first embodiment of the present invention.

FIG. 2 is a diagram illustrating the principle of an embodiment of the present invention.

FIG. 3 is a diagram showing an encoding processing flow using the encoder of FIG. 1.

FIG. 4 is a functional block diagram of a decoder for implementing the binary data decoding method according to the second embodiment of the present invention.

5 is a diagram showing a decoding processing flow using the decoder of FIG.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 symbol appearance probability estimating means 2 code calculating means 3 A value calculating means 6 code space reducing means 7 comparison calculating means 10 and 20 control means 12 decoding calculating means

Claims (2)

[Claims] 1. A binary data encoding method for encoding a symbol sequence, which is a binary data sequence, by an arithmetic encoding method, wherein a symbol appearance probability estimating means causes a symbol to appear from a symbol pattern of the symbol sequence to be encoded. The probability is estimated, and the code space reducing means reduces the area of the code space at a reduction rate in consideration of the application-specific bias appearing in the position coordinates of the code space to which the symbol sequence to be encoded is mapped, The calculation means maps the symbol pattern to the number line [0, 1) section of the reduced code space region according to the appearance probability of the symbol estimated by the symbol appearance probability estimation means and its value. And a binary data encoding method. 2. The symbol appearance probability is estimated from the data encoded by the binary data encoding method according to claim 1, and then the inverse of the code operation means is performed according to the symbol appearance probability and its value. A binary data decoding method, characterized by performing an operation to decode the binary data series.