JPH04100322A - Data compression and decoding system - Google Patents

Abstract

PURPOSE:To reduce the redundancy between character strings and to obtain a high compression rate by representing an index representing a registration position as the entire dictionary with an index in numbers arranged in the order of appearance of each division dictionary. CONSTITUTION:A registration dictionary is split into some division dictionaries DD1-DD5, and an index representing a registration position of the entire dictionary is indicated by an index in a split dictionary group when the division dictionaries DD1-DD5 are arranged in the order of high appearance rate. For example, when a character string of an index T in the division dictionary DD4 is registered, a coded index is indicated and coded by representing the index as a sum 25 between an index 7 in the registered division dictionary DD4 and a total sum (n1+n2+n3=18) of the character strings in preceding division dictionaries DD1-DD3 to the registered division dictionary DD4. Thus, the relation of subordinate between coded character strings is fetched at index coding, the redundancy between character strings is reduced and a high compression rate is realized.

Description

[Detailed Description of the Invention] [Summary] Regarding a data compression and decompression method using LZW encoding as an improvement of incremental decomposition encoding, which is a type of universal encoding, the present invention aims to improve the compression rate by reducing redundancy between character strings. For the purpose of increasing the number of subsequences, the registered dictionary is divided into several parts, and when registering a new subsequence in the dictionary, specify and register the subsequence divided according to the dependency relationship with the already registered subsequence. At the same time, the registration position of the dictionary as a whole is represented by a reference number indicating the registration position within the divided dictionary group, and this reference number is encoded. Furthermore, at the time of restoration, divided dictionaries that are classified according to their subordination with already registered subsequences are designated and registered, and the code specified by the reference number indicating the registration position within the divided dictionary group is restored.

[Industrial Application Field] The present invention relates to an image data compression method using an LZW code, which is known as an improved incremental decomposition type of universal code.

In recent years, computers have come to handle various types of data such as character codes, vector information, and images.
The amount of data handled is also rapidly increasing. When handling large amounts of data, it is desirable to reduce storage capacity and speed up transmission by eliminating redundant parts of the data and compressing the amount of data.

Universal encoding has been proposed as a method that can compress various types of data using one method.

Here, the field of the present invention is not limited to character code compression.
Although it can be applied to a variety of data, in the following we will follow the nomenclature used in information theory, and refer to a single word unit of data as a character, and a string of data that is connected by multiple warps. .

A representative method of universal encoding is the Jibo Lempel (2i-Lempel) code (for details, see
For example, Munakata "2iv-Lempel's data compression method"
.

Information processing, Ma01.26. No. 1.1985).

2iv-Lempel code has two types: ■ Universal type and ■ Incremental parsin type.
Two algorithms have been proposed: g).

Furthermore, as an improvement of the universal algorithm, LZ
There are SS codes (T, C, Be1l, ' BeNe
r OPM/L Text Compression'
, IEEE Trans, on CommonVo
l, C0M-34, NO, 12, Dec, 1986
reference).

Moreover, as an improvement of the incremental decomposition type algorithm, L Z
There is a W (Lempel-2iy-Welch) code (T, A Welch, 'A Technique f
or High-Performance DataC
emptession', Computer, Ju
ne 1984).

Among these encoding methods, the LZW code has come to be used for file compression in storage devices because of its ability to perform high-speed processing and its simple algorithm.

[Prior Art] FIG. 7 shows an encoding processing flow using a conventional LZW code, and FIG. 8 shows a decoding processing flow.

First, LZW encoding processing has a rewritable dictionary,
Divide the input character string into different character strings (substrings) and register these character strings in the dictionary with reference numbers in the order in which they appear, and also register the currently input character string in the dictionary and find the longest string. - It is represented and encoded by a reference number in a scattered character string.

FIG. 9 shows an explanatory diagram of LZW encoding, FIG. 11 shows an explanatory diagram of LZW decoding, and FIG. 10 shows an example of decoding and a dictionary structure created during decoding.

In addition, in Figure 9.10.11, to simplify the explanation, a
An example of compressing and restoring data consisting of a combination of three characters bc is taken up.

In the LZW encoding process shown in FIG. 9, first, in step 81 (hereinafter "step" will be omitted), a character string consisting of one character for each character is registered in the dictionary as an initial value, and then encoding is started. Also, the encoding in Sl is the first input character K
Search the dictionary to find the reference number ω, and use this as the initial character string.

Next, the next character K of the input data is read in step 82, and after checking whether all character input has been completed in step S3, the process proceeds to step S4, where the character K read in step 82 is added to the initial character string ω determined in step S1. A search is made to see if the character string (ωK) to which the letter K is added is found in the dictionary.

If the character string (ωK) is not in the dictionary in S4, proceed to S6 and use the code word code (
ω), add a new reference number to the character string (ωK), register it in the dictionary, replace the input character K in S2 with the reference number ω, increment the dictionary address n, and return to S2. Read the next character K.

On the other hand, if the character string (ωK) is in the dictionary in S4, the character string (ωK) is replaced with the reference number ω in S5, the process returns to S2, and the maximum growth is continued until the character string (ωK) cannot be found in the dictionary in S4. Continue searching.

LZW encoding will be specifically explained as follows with reference to FIGS. 9 and 10.

First, the input data 1npl in FIG. 9 is read from left to right.

When the first character a is input, there is no matching character string in the dictionary other than a, so 0UTPUT C0DE 1 (reference number ω) is output as a code word. Then, reference number 4 is added to the expanded character string ab and it is registered in the dictionary. The actual dictionary registration is on the right side of Figure 10 (ALT-ERNATE TA
BLE) is registered as a character string 1b.

Then the second b becomes the beginning of the string. Since there is no matching character other than b in the dictionary, the reference number 2 is output as a code word, and at the same time, the expanded string ba is also not in the dictionary, so the string ba is represented by 2a and the reference number 5 is added. Register in the dictionary. The third a becomes the beginning of the next string. This process continues in the same manner.

The decoding process shown in FIG. 8 performs the reverse operation of the encoding process shown in FIG. 7.

In the LZW decoding shown in FIG. 8, decoding is started after a character string consisting of one character for every character is registered in the dictionary as an initial value in the same way as during encoding.

First, read the first code (reference number) with Sl, and
Since 0DE is set as 0LDcode and the first code corresponds to one of the reference numbers of one character already registered in the dictionary, the character code (K) matching the input code C0DE is searched and the character K is output.

In addition, the output characters are written to FINchi for later exception handling.
Set it to r.

Next, go to 82, read the next code, and set it to C0DE.INc
Set as ode. In S3, it is checked whether there is a new code, that is, whether the code input has ended, and the process proceeds to S4, where it is checked whether the code C0DE inputted in S3 is defined (registered) in the dictionary.

Normally, the input code word has been registered in the dictionary in the previous processing, so the process advances to S5 and the character string code (ωK) corresponding to the code C0DH is read from the dictionary, and the character K is temporarily stacked in S6. , the reference number C0DE(ω) is changed to a new code C0DE, and the process returns to S5 again. S6
Repeat the steps recursively until the reference number ω reaches one character K, and finally proceed to 87 and change the stacked character to L in S6.
Pop up and output in IFO (Last in FastOul) format. At the same time, at 87, a new reference number is added to a character string in which the previously used code ω and the first character K of the character string restored this time are expressed as a set (ω, K), and the result is registered in the dictionary.

The LZW decoding process will be specifically explained as follows with reference to FIG.

First, in Figure 11, the first input code (1! IPUT C0D
E) is 1, - for the letters a, b, c already the reference number 1. 2. Since it is registered in the dictionary as 3 as shown in Figure 10, the code 1 is registered by referring to the dictionary.
The character string a with the reference number that matches is replaced and output.

Similarly, the next code 2 is replaced with a letter S and output. At this time, a new reference number 4 is added to the character string (1b), which is a combination of the previously processed code 1 and the first character decoded, and is registered in the dictionary.

The third numeral 4 replaces the character string 1b found by searching the dictionary with the character string ab and outputs the character string ab. At the same time, a new reference number 5 is added to a character string 2a (=ba), which is a combination of the previously processed code 2 and the first character a of the currently decoded character string, and is registered in the dictionary.

This process is repeated in the same manner.

The LZW decoding shown in FIG. 11 involves the following exception handling.

This exception handling occurs in the decoding of the sixth input code 8. Code 8 is not defined in the dictionary at the time of decoding and cannot be decoded. In this case, exceptional processing is performed in which a character string 5b is obtained by adding the first character ``s'' of the previously decoded character string ba to the previously processed code 5, and is further replaced with 2 a b = bab and output. Then, after outputting the string, the previous code word 5
The reference number 8 is added to the character string 5b obtained by adding the first character of the character string just decoded to the character string 5b, and the result is registered in the dictionary.

This exception handling is carried out at 84 and S in the decoding process flow in FIG.
Finally, in step 87, the character string is output and the new character string is registered in the dictionary with a reference number added.
It will be held at 7.

In addition, in the LZW decoding shown in Figures 8 and 11, we have explained the case where the dictionary is created in real time while decoding the code on the decoding side, but the dictionary created during encoding is used as a copy on the decoding side as it is. It may be encoded by doing this. In this case, exception handling on the decoding side becomes unnecessary.

[Problems to be Solved by the Invention] As described above, in data compression using the conventional LZW code, the index (reference number) in the dictionary in which already encoded subsequences used for encoding are registered is As shown in FIG. 12, the character strings in the dictionary are represented by numbers indicating the order in which they were registered.

Therefore, the dictionary index used as a codeword only indicates the number in which a character string is registered, and the correlation between the indexes is small, and each index is used almost randomly as a codeword. The index has to be expressed using a fixed number of bits, and the bit length depends on the dictionary size, which poses a problem in that the index cannot be encoded efficiently.

The present invention has been made in view of such conventional problems, and an object of the present invention is to provide a data compression and restoration method that can reduce redundancy between character strings and increase compression ratio.

[Means to solve the problem] FIG. 1 is a diagram explaining the principle of the present invention.

As shown in FIG. 1(a), the present invention divides encoded data into different subsequences, adds a different reference number to each subsequence, and registers it in a dictionary TD, and input data is divided into different subsequences. The target is a data compression method that encodes by specifying the reference number of the most matching subsequence among the subsequences of the dictionary TD, and a data restoration method that restores a character string from the code word obtained by the data compression method. do.

First, in the data compression method of the present invention, as shown in FIG. At the same time, the registration position of the entire dictionary is set to the divided dictionary groups DD1 to DD, which have a higher probability of appearance than the divided dictionary to which the registered encoded partial sequence belongs, and the total number n of subsequences in the group. The entire registration position is designated by a number added to the reference number in the divided dictionary DDi to which the encoded subsequence belongs, and this number is configured to be encoded.

On the other hand, in the data decoding method of the present invention, when registering a new decoded subsequence into a dictionary, a divided dictionary DDi classified according to the dependency relationship with already decoded subsequences is specified and registered, and Divided dictionary DD to which the decoded subsequence belongs
Divided dictionary groups DD, ~DD14. with higher appearance probability than i.
The code is configured to decode the code specified by the total number of subsequences in the subsequence plus the reference number in the divided dictionary DDi to which the decoded subsequence belongs.

[Operations] Data compression and decompression according to the LZW code with such a configuration provides the following effects.

That is, in conventional LZW codes, when an input character string is divided into different character strings and encoded, each character string currently being encoded is encoded as appearing independently of the previously encoded character string. It takes the form of

This method has no problem in encoding memoryless information sources. However, many data such as actual sentences can be considered as memory information sources. However, in the conventional LZW code, the history of occurrences of character strings cannot be fully utilized, and even after data compression, redundancy remains in the dependencies when character strings appear.

Figure 1(a) shows the entire dictionary by conventional LZW encoding.
In this case, the root of the dictionary search tree (root = index O) is empty, and in the LZW code, the history of strings that previously appeared for the string being encoded is shown. indicates that it has not been considered.

Therefore, the present invention focuses on the fact that the redundancy between character strings can be reduced and the compression rate can be increased by incorporating the dependency relationships between encoded character strings at the time of index encoding. In this way, the registered dictionary is divided into several divided dictionaries DD.
, ~DD, and the index indicating the registration position of the dictionary as a whole is expressed as an index in the divided dictionary group when each divided dictionary DD, ~DD is arranged in order of probability of appearance. Aim to improve compression ratio.

For example, if a character string with index 7 indicated by a black circle is registered in the divided dictionary D D a of the divided dictionaries DD to DD5 arranged in the order of appearance probabilities as shown in the figure, the encoded index will be the registered divided The index in the registered divided dictionary DD, for example, the index 7 of the divided dictionary DD4, is added to the sum of character strings in the divided dictionaries DD, ~DD3 up to one before the dictionary DD4, nl+n2+n3=18, and is expressed as a value 25 and encoded. do.

[Embodiment] FIG. 2 is a block diagram showing an embodiment of the present invention.

In FIG. 2, 10 is a CPUr, which performs data compression by encoding character strings according to the present invention according to the LZW code, and restores character strings from code words obtained by data compression.

A variable length encoder/decoder 12 is provided for the CPU10, and is specifically realized by encoding software based on a coding algorithm and decoding software based on a decoding algorithm. For encoding and decoding by the variable length encoder/decoder 12, a stack memory 14 for converting the order of restored characters is used.

Furthermore, for the CPU 10, a total dictionary (TD) 16, a divided dictionary (DD) 18, and a memory 20 for storing divided dictionary elements in the order of their appearance are provided. The general dictionary 16 has the conventional LZ
Dictionary registration similar to W encoding and decoding is performed. The divided dictionary 18 is a dictionary unique to the present invention, and is divided into N parts in advance. Taking encoding as an example, when registering a new encoded character string in the dictionary, the subordination relationship with already registered character strings is determined. The divided dictionary DDi separated by is specified, and an index (address) indicating the registration position in the corresponding overall dictionary 16 is registered. When specifying an arbitrary divided dictionary in the divided dictionary 18 and registering a character string address, the order in which the divided dictionary elements appear is stored in the appearance order storage memory 20. The order of appearance of the divided dictionaries stored in the appearance order storage memory 20 is used to create a reference number indicating the position of the character string registered in the divided dictionary 18 in the entire dictionary, that is, an index to be encoded. be done.

Next, the principle of the present invention will be explained using encoding processing as an example.

First, in the encoding process, the reference dictionary is divided into several subsets, that is, divided dictionaries DDi (
i=1 to N). When registering a new character string, the divided dictionary in which the new character string is to be registered is specified and registered based on the dependency relationship with the already registered character strings.

Note that the registered data of the divided dictionary is the address of the character string registered in the entire dictionary.

For example, as shown in FIG. 3, five divided dictionaries DD, ~
DD, and the character string to be newly registered is classified into the divided dictionary DD4 due to its dependency with the already registered character strings, and is the seventh registration in the divided dictionary DD4. , character strings are registered in the divided dictionary (whole dictionary address) as indicated by index 7 indicated by a black circle in the divided dictionary DD4.

Once such divided dictionary registration is completed, the divided dictionaries are arranged in the order in which character strings belonging to the time-divided dictionaries DD, to DD are likely to appear. In the case of Fig. 3, the divided dictionaries DD, ~DD, are arranged in the order of ease of appearance, with the divided dictionary DD+ being the most likely to appear, and the divided dictionary DD being the least likely to appear.
The divided dictionary DD4, which has been registered this time, is located in the fourth most likely order of appearance.

If the divided dictionaries DD, ~DD are arranged in the order in which character strings belonging to the divided dictionaries are likely to appear, the registration position of the entire dictionary for index 7 indicated by the black circle of the registered divided dictionary DD4 is determined by index i. Find it as.

For example, if the character string of interest being encoded belongs to the divided dictionary DDx and is the character string registered in the divided dictionary DDA, the index i indicating the registration position in the dictionary as a whole is The index i indicating the registered position of can be expressed by the following equation.

i= Σ n (D + kj ・
・ ・ (1−)P (11>P (11) where n (x) is the number of nodes in the divided dictionary X p (x) is the appearance probability of the divided dictionary Taking the registration of a character string as an example, calculating the index i that indicates the registration position in the entire dictionary is as follows.First, Σ
n (j) is 1 from the divided dictionary DD4 in which character registration was performed.
Since it represents the total number of nodes of character strings registered in each of the previous divided dictionaries DD1 to DD, Σn (j) = nl + n2 + n3 = 6 + 8 + 4 = 18
becomes. In addition, D is a divided dictionary in which character strings are registered.
The index indicating the registration order in D4 is 7. Therefore, the index i of the entire dictionary can be determined as 1=18+7=25.

In the conventional LZW encoding method, since the order registered in the dictionary is used as an index, the correlation between indexes becomes small. On the other hand, in the present invention, since the index is calculated using the above-mentioned equation (1), it is possible to allocate an index with a smaller number than in the conventional method. The index given by the above formula (1) will have the smallest value for character strings that are likely to appear, so instead of encoding the index as a fixed length code, encode it so that it is shorter than the bit length of the index. By doing this, a high compression ratio can be obtained.

Here, the index i given by the above equation (1) is the order of the cumulative appearance probability of the divided dictionary, and for this cumulative appearance probability, use either of the following ■ or ■ to find the order in which the divided dictionary is likely to appear. .

■Order of appearance probabilities of divided dictionaries; If the number of indexes of each divided dictionary is n (j), the appearance probability of the j-th divided dictionary DDj is as follows: Probability of appearance = n
(j)/n... Since it is approximated by (2),
The order of the appearance probabilities of divided dictionaries is expressed in the order of the number of indexes n (D) in each divided dictionary. However, n in the above formula (2) is n = Σn (D However, N is the number of divided dictionaries is the total number of indexes in the entire dictionary defined by .

■Order of transition probabilities in divided dictionaries: Encode the target character string by determining the probability p (Klj) of transition from the last character K of the previous character string to the character string group j of the divided dictionary to which the target character string to be encoded this time belongs. Alternatively, the divided dictionaries are obtained while being updated each time they are decoded, and the divided dictionaries are expressed in the order of the size of p(Klj).

By using the transition probabilities of this divided dictionary, redundancy between character strings can be efficiently reduced.

Next, the encoding process and decoding process according to the present invention will be explained in detail with reference to the operational flow diagrams of FIGS. 4 and 5.

First, the encoding process of the present invention will be explained with reference to the operational flow diagram of FIG. 4 as follows. Note that the underlined portions in each step in FIG. 4 are different from the conventional encoding process shown in FIG. 7.

In FIG. 4, initialization processing is first performed in Sl. This initialization process has the following contents.

(2) Initialize the entire dictionary TD to include the first character. Specifically, since addresses n=0 to 255 are assigned to the first character for which initial registration has been performed,
Let the starting address n of the entire dictionary TD be n=256.

■The first address n(i) of the divided dictionary is all n(i)−
Set it as 〇. However, i=0 to the number of divisions.

■Input the first letter K and use this as the initial letter string ω.

■The divided dictionary number f (K) to which the input character K belongs is set to f (K
)=d.

After the above initialization processing of Sl is completed, the process advances to 82 and the next input character K is read. In S3, it is checked whether reading of all input characters has been completed. If not, the process proceeds to S4, and a character string (ωK ) is found in the overall dictionary TD. If the character string (ωK) is in the overall dictionary TD, the process advances to S5 and the character string (ωK) in the overall dictionary TD is
) is stored as the next initial character string ω, the process returns to S2, and the next input character K is read.

By repeating steps 82 to S5, if the character string (ωK) is not found in the overall dictionary TD in S4, the process advances to S6.

In S6, the following processes (1) to (2) are performed.

■Calculate the index i based on the above formula (1),
It is output as code (g (ω)) by variable length encoding.

That is, since the divided dictionary number in which the character string (ωK) is to be registered is d, the appearance probability of all divided dictionaries p (
x), calculate the total number of subsequences of the divided dictionary that have a higher occurrence probability than the d-th occurrence probability p (x) that is currently being registered, and add the subsequences in the divided dictionary with the divided dictionary number d to this total number. An index i is calculated by adding the index value indicating the registration order of , and this is encoded and output as g(ω). Further, the appearance order h(x) of the divided dictionary obtained when calculating this index i is stored in the storage memory 20 in the order of appearance as shown in FIG.

■Register the character string (ωK) at address n of the overall dictionary TD.

■Address n of the overall dictionary TD where the character string (ωK) is registered
The address n of the divided dictionary DD with the divided dictionary number d (d)
Register.

■Replace the last character K of the character string (ωK) with a new initial character string ω.

■ Divided dictionary number f (K) to which the letter K belongs to f (K)
=d.

(2) Increment the address n of the entire dictionary TD by one.

(2) Increment the address n (d) of the divided dictionary DD by one.

After completing the processing in S6 shown in (1) to (4) above, the process returns to S2 again to read the next input character K.

When it is determined in S3 that reading of all input characters has been completed by repeating the above processing, the process advances to S7 and
6, the index i is calculated, and the calculated index i is variable-length coded as code (g (ω
)) is output and the series of processing ends.

FIG. 5 is an operational flow diagram of the decoding process according to the present invention, and the underlined parts of each step are different from the conventional process shown in FIG.

In FIG. 5, first, the following initialization processing shown in (1) to (4) is performed in Sl.

(2) Initialize the entire dictionary TD to include the first character. That is, since the address n=Q~255 of the entire dictionary TD is assigned for the first character, the address n=Q~255 of the entire dictionary TD
Let the start address n of the file be n=256.

■The first address n(i) of the divided dictionary DD is all n(i
) = set to O. However, i indicates O to the number of divisions N.

■Read the first code C0DE.

(2) Find the index ω of the entire dictionary TD using g-'(code-' (CODE)).

However, "code" is a table for decoding a variable length code into a fixed length code.

g −1 is the index of the whole dictionary TD divided into dictionary D
Table to convert to index of D.

■ The index ω of the entire dictionary TD obtained in ■ is ω=
Place it in OLDω.

■ Find the first character K from the overall dictionary index ω.

■Output the first character K found 3. ■Place the first letter K as 1.

Next, proceed to 82 to read the next input symbol code C0DE, and in the same way as ■ in Sl-, the whole dictionary index C0
Determine DE and place this index C0DE in INω.

Next, proceed to 83, check whether reading of all input codes has been completed, proceed to S4, and add C0D to the overall dictionary TD.
Check if E is defined. Normally, the input code word has been registered in the dictionary in the previous processing, so the process proceeds to S5 and reads the corresponding character string (ωK) from the general dictionary using the general dictionary index corresponding to C0DH, and the character K is read out in S6. It is temporarily stuck, the reference number ω is set as a new C01)E, and the process returns to S5. This S5. The procedure in S6 is recursively repeated until the entire dictionary index ω or one character K is reached, and finally the process proceeds to 87 and the stacked characters in S6 are LIFO (La+t In Fact 0u
t) Pop up and output in format.

The process of S7 has the following contents.

■Pop up and store the stack until the stack is empty, then output a certain character.

(2) The number d of the next divided dictionary is determined from the restored character KJ- and 嘗<0, and (2) the first character of the restored character is 1.

(2) Store the set (OLDω, K) at address n of the entire dictionary TD.

(2) Store the address n of the entire dictionary TD at the address (j, n (d)) of the divided dictionary index conversion table.

■Increments Atoshi/Sn in the entire dictionary by one.

(2) Increment the address n (d) of the divided dictionary by one.

■Set the overall dictionary index INω as 0Ll)ω.

Note that S5 performs exception handling similar to the conventional example shown in FIG. ■Output the character FINcha+, which has already been placed in 1=.

■ OLDω set in ■ of S1 is set as C0DE.

(2) Set the overall dictionary index obtained from the set of TD(ωKl) as INω.

Next, the conversion process from the entire dictionary index to the divided dictionary index, which is performed using the table g-1 at 81 and S2 in FIG. 5, is performed by the following steps 1 to 2.

■The number j indicating the order of the divided dictionaries arranged in order of appearance frequency is j
=0.

■While the divided dictionary index X is positive, the next loop is executed.

■Increment by one as j=j+1.

■ Find x=x−n (h (D ). However, n(h
(j)) is the number of strings (number of nodes) in the j-th divided dictionary
It is.

■When the divided dictionary index X becomes a negative value,
The j=j-1st index, that is, the index X calculated in the previous divided dictionary, becomes the value of the divided dictionary index.

(2) A pair (j, x) of the dictionary number j of the divided dictionary obtained in (2) and the index X in the j-th divided dictionary is output.

For example, if the overall dictionary index = 25 is given in Fig. 2, after setting j = 0, set j = j + 1 = 1 to obtain x = 25 - nl = 25 - 6 = 19, then set j =
2, find x=19-8=11, then set j=3 to find x=11-4=7, and then set j=4 to find x=7-
Find 8=-1. Since x=-1 at j=4, which is a negative value, x=7 at the previous j=3 is determined as the divided dictionary index. That is, index 7 indicated by a black circle in the divided dictionary DD4 is found as a converted value.

In the encoding and decoding operation graphs of FIGS. 4 and 5 explained above, the encoding code and decoding code of the index indicating the position of the entire dictionary are as small as the index.
Variable length encoding, such as Huffman encoding, which is a known technique for assigning short codes, can be used. Furthermore, in order to maintain universal properties so that various data can be compressed, for example, an Elias (Elia+) code that is artificially shortened and assigns a short code length to the number may be used.

The Elias code is composed of a γ code and 6 codes as shown in FIG. The γ code of the Elias code is a binary number with a prefix indicating a significant digit added thereto. δ
The code is the prefix of the γ code further represented by the γ code.1. With the erase code, the number of binary digits can be determined from the prefix, so even if the code word is bit-packed, it can be unambiguously decoded. can.

"Effects of the Invention" As explained above, according to the present invention, dictionary registration can be divided into several parts, and the registration position of the two dictionaries can be shown as a whole. By representing each character string by an index in a number arranged in the order of appearance in each divided dictionary, redundancy between character strings can be reduced, and a higher compression rate can be obtained than with the conventional 15ZW code.

[Brief explanation of drawings]

FIG. 1 is a diagram illustrating the principle of the present invention; FIG. 2 is a diagram illustrating the configuration of an embodiment; FIG. 3 is a detailed diagram illustrating the present invention. Fig. 4 is an encoding process flow diagram of the present invention; Fig. 5 is a decoding process flow diagram of the present invention; Fig. 6 is an explanatory diagram of the Elias code used in the present invention; Fig. 7 is a conventional LZW
Encoding process flowchart: FIG. 8 is a conventional LZW decoding process flow diagram; FIG. 9 is an explanatory diagram of ZW encoding; FIG. 10 is an explanatory diagram of a dictionary configuration example; FIG. 11 is an explanatory diagram of LZW decoding. In the figure, 10: CPU 12: Variable length encoder/decoder 14, stack J6 for converting the order of restored characters/total dictionary (TD) 18: divided dictionary (DD)

Claims (2)

[Claims] (1) Divide the encoded data into different subsequences, add a different reference number to each subsequence, and register it in a dictionary, and match the input data with the maximum length among the subsequences in the dictionary. In a data compression method that specifies and encodes a subsequence using its reference number, when registering a new encoded subsequence in a dictionary, a divided dictionary is created that is classified according to the dependency relationship with already registered subsequences. In addition to specifying and registering, the registration position of the entire dictionary is set to the total number of subsequences in the divided dictionary group that have a higher probability of occurrence than the divided dictionary to which the registered encoded subsequence belongs. A data compression method characterized by specifying a number with a reference number added and encoding the number. (2) Divide the decoded data into different subsequences, add a different reference number to each subsequence, and register it in the dictionary.
In a data restoration method in which the input restored data is decoded by specifying the reference number of the maximum length matching subsequence among the subsequences in the dictionary, when registering a new decoded subsequence in the dictionary, , specifies and registers a divided dictionary that is classified according to the dependency relationship with the already decoded subsequence, and also registers the registration position of the entire dictionary in a divided dictionary group that has a higher probability of appearance than the divided dictionary to which the decoded subsequence belongs. A data restoration method characterized by decoding a code specified by a number obtained by adding a reference number in the divided dictionary to which the decoded subsequence belongs to the total number of subsequences.