JP3115066B2 - Dictionary search method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a dictionary search method in data compression, and in particular, divides an already encoded character string into different partial character strings, registers the partial character strings in a dictionary, and stores input character strings in a dictionary. The present invention relates to a dictionary search method in data compression that searches a dictionary for a partial character string that matches the longest string and a string, and specifies and codes the number of the longest matching character string.

In recent years, various types of data such as character codes, vector information, and images have been handled by computers, and the amount of data handled has been rapidly increasing. When dealing with a large amount of data, by compressing the amount of data by omitting redundant portions in the data, it becomes possible to reduce the storage capacity or to transmit data at high speed. Universal coding has been proposed as a method that can compress various data in one system.

[0003]

2. Description of the Related Art This universal code is an information preserving type data compression method. Since statistical properties of an information source are not assumed in advance at the time of data compression, various types (character codes, object codes, etc.) can be used for data. Can be applied. Document images have similarities in character outlines and character intervals, and halftone images have similar dot periodicity and halftone dot shape. Redundancy having this similarity can be reduced by the universal code, and effective compression can be performed. In the following, following the name used in information theory, one word unit of data will be called a character, and data connected with an arbitrary word will be called a character string.

[0004] As a typical method of the universal code,
There is a Ziv-Lempel code. For example, Munakata "Ziv-Lempel Data Compression Method", Information Processing, Vol. 26, No.
See 1,1985. In this Ziv-Lempel code, two algorithms of a universal type and an incremental decomposition type (Incremental parsing) have been proposed.As a practical method using the universal type algorithm, an LZSS code (T.
C. Bell, "Better OMP / LText Compression", IEEE Tran
s. on Commun., Vol. COM-34, No. 12, Dec. 1986), and a practical method using an incremental decomposition type algorithm is an LZW (Lempel-Ziv-Welch) code ( TA
Welch, "A Technique for High-Performance Data Co
mpression ", Computer, June 1984) Among these codes, the LZW code is used for file compression of a storage device because of its high-speed processing and the simplicity of the algorithm.

[0005] LZW encoding In LZW encoding, a rewritable dictionary is provided,
Divides the input character string into different character strings, assigns numbers to the character strings in the order in which they appear, and registers them in the dictionary, and stores only the dictionary number of the longest matching character string registered in the dictionary with the currently input character string. And encode it.

FIG. 8 is an explanatory diagram of LZW encoding, FIG. 9 is an explanatory diagram of a dictionary configuration, and FIG. 10 is a flowchart of LZW encoding processing. For the sake of simplicity, it is assumed that a character string composed of three characters a, b, and c is subjected to LZW encoding and data compression. In advance, a character string (a,
b, c) is given a registration number and is initially registered in the dictionary, and the number N of registrations in the dictionary is set to the number M of character types (M → N).・
・ Step 101

In such a state, the first character K is inputted, the registration number of the character is set as the reference number ω, and this is referred to as the initial character string (p
refix string) (step 102). Next, the next character K of the input data is read (step 103),
Step 103 is applied to the word string ω obtained in step 102.
A search is performed to determine whether or not the character string (ωK) to which the character K read in (1) is added in the current dictionary (step 104).

If the character string (ωK) exists in the dictionary, the character string (ωK) is replaced with ω (step 105), and thereafter, it is determined whether the input data has been completed (step 10).
6) If the data is not completed, the process returns to step 103 and the subsequent processing is repeated, and the search for the maximum matching length character string is continued until the character string (ωK) cannot be searched from the dictionary. On the other hand, if the input data is completed in step 106,
The reference number ω is output as the code word code (ω) (step 107), and the encoding process ends.

If the search for the longest matching character string continues and the character string (ωK) does not exist in the dictionary in step 104, the reference number ω is output as a code word code (ω), and the character string (ω ωK) is registered in the dictionary address N, the character K read in step 103 is replaced with the reference number ω, and the dictionary address N is incremented by 1 (step 108). Next, it is determined in step 106 whether the input data has been completed, and the subsequent processing is repeated according to the determination result.

[0010] The LZW encoding will be specifically described below with reference to FIGS. 8 and 9. That is, FIG.
Is read one character at a time from left to right.
When the first character a is read, there is no matching character string other than a in the dictionary, so the registration number “1” of a (reference number ω
= 1) as a codeword (code (ω)). Then, the extended character string ab is assigned a registration number 4 and registered in the dictionary. The actual registration is in the form of a character string "1b". Subsequently, the second character b becomes the head of the input character string. Since there is no matching character string other than b in the dictionary, the registration number (reference number) 2 of b is output as a code word, and the expanded character string ba is actually assigned a registration number 5 in the form of 2a. Register in the dictionary.

As described above, the third character a becomes the head of the input character string. Since the first character a exists in the dictionary, a character string “1b” obtained by adding the next character b to the registration number 1 of the character
Check if exists. Since the character string "1b" exists,
A character string “4” in which the following character c is added to the registration number 4 of the character string
c "is present. Since the character string “4c” does not exist, the registration number “4” of the longest matching character string “1b” is output as a code word, and the expanded character string “4c” is registered number 6
Then, the encoding and dictionary registration are repeated in the same manner to execute the LZW encoding process for all input characters.

FIG. 11 is a flowchart of the LZW decoding process. In the decoding process, the reverse operation of the encoding is performed. That is, at the time of decoding, as in the case of encoding, a character string (a, b, c) consisting of one character for every character is assigned a registration number and initially registered in the dictionary, and the number N of registered dictionary is changed to the number of character types. Let M be (M → N). ..Step 201 Then, the first code CODE is read and the code OLDc is read.
ode. Further, since the first code corresponds to one of the registration numbers of one character already registered in the dictionary, the character K indicated by the input code CODE (= registration number) is output. The output character K is set as char for later exception processing. ..Step 202 above

Thereafter, the next code CODE is read and NEWc is read.
ode (step 203) and code CO
It is checked whether DE (= registration number) is defined (registered) in the dictionary (step 204). Normally, input code C
Since the ODE (= registration number) has been registered in the dictionary in the processing up to the previous time, "NO" is obtained in step 204. Next, the registered character string of the dictionary indicated by the code CODE (= registration number) is ( ωK). That is, it is determined whether the registered character string of the dictionary indicated by the code CODE is a combined character string of the reference number ω and the character K, such as (ωK) (step 2).
05).

If the character string is a combination of the reference number ω and the character K, the character K is temporarily stacked and the code word of the reference number ω is
code (ω) (actually code (ω) = ω) is set as a new CODE, and the number of characters C is incremented by 1 (step 20).
6) Return to step 205. Thereafter, step 205,
Step 206 is recursively repeated until the registered character string indicated by CODE reaches one character.

In step 205, if the character string indicated by the CODE is one character, that is, if the registered character string in the dictionary indicated by the code CODE is (K), K is output, and then the stack is executed. LIFO (Last i
n Pop up and output in Fast Out) format. A character string (OLDcode, OLDcode, OLDcode) obtained by adding the first character K of the character string decoded this time to the code OLDcode used in the previous decoding.
K) is registered in the dictionary as a registration number N, and then N is incremented by 1 (N + 1 → N). Further, the first character K of the decoded character string is set to char, and NEWcode is set to OLDcode.・
・ Step 207

Thereafter, it is determined whether code input has been completed (step 208). If not completed, the process returns to step 203 to read the next code and repeat the decoding process. by the way,
In the encoding process, since the encoding of a certain character string and the dictionary registration of the character string in which the next leading character is added to the character string are performed at the same time, in the next encoding process, Codewords can be used. However, in the decoding process, since the character string obtained by adding the head character string of the currently decoded character string to the character string decoded immediately before is registered in the dictionary, dictionary registration is delayed by one time as compared with the encoding process. For this reason, if the code word of the character string coded immediately before is used in the encoding process, the code word may not be registered (defined) in the decoding process. This is step 2
In 04, no code is defined, and "Y
ES ”.

For example, at the time of encoding, as shown in FIG. 12, an OLD code is output for a character string "a... Z", and the character string "a. Is output by NEWcode, and the character string "a ... zab" is registered in the dictionary. Now,
When the decoding side reads the code word NEWcode, the decoding cannot be performed because the code word is not registered in the dictionary on the decoding side. However, comparing NEWcode and OLDcode, there is the following relationship: NEWcode character string = OLDcode character string + OLDcode character string first character (char). For this reason, if "NO" in step 204, the set char is stacked and OLDc
The character string in which ode is regarded as CODE and OLDcode is added with char is set as NEWcode (step 209), and the processing from step 205 onward is performed using CODE thereafter.

The decoding process will be described below in detail with reference to FIG. The first input code is “1”, and one character a, b, c is already registered with the registration number 1,
Since the dictionary is registered as 2, 3 (similar to FIG. 9),
By referring to the dictionary, it is replaced with the character string a of the registration number that matches the code “1” and output. Next, the code "2" is similarly replaced with the character b and output. At this time, “1b”, which is a combination of the previously processed code “1” and the first character “b” decoded this time, is registered in a new registration number 4 in the dictionary.

The third code "4" is obtained by searching the dictionary.
The character string “ab” is output by replacing “1b” with “ab”. At the same time, the character string “2a (= a) is obtained by combining the code“ 2 ”processed last time and the first character a decoded this time.
b) is registered in a dictionary with a new registration number 5. Hereinafter, similarly, the decoding process is repeated. Step 209 in FIG.
Exception processing occurs in the decoding of the sixth input code “8”. The code “8” is not defined in the dictionary at the time of decoding and cannot be decoded. In this case, a character string "5b" is obtained by adding the first character b of the character string "ba" decoded last time to the code "5" processed last time, and further replaced with "2ab" and "bab". Is output. Then, a registration number "8" is added to the character string "5b" obtained by adding the character b of the character string decoded this time to the previous code word "5", and the dictionary is registered in the dictionary.

As described above, the universal code is a compression method in which, even if the property of the object to be encoded is unknown, the encoding is performed while learning it. Is displayed, the registration number is transmitted as encoded data (codeword). However, when encoding is performed in accordance with the flowchart of FIG. 10, when searching a dictionary for one character string, the entire dictionary must be searched in the worst case, and there is a problem that the encoding process takes time. Therefore, conventionally, the processing speed is increased by using an external hashing method (open hashing or chaining) for dictionary search (for example, see Ohmsha, edited by Information Processing Society of Japan, Information Processing Handbook).

Considering a set S composed of external hash character strings, a high-speed search can be performed if the address of the storage position of the character string x in the set S can be directly calculated from the character string x. The hash method realizes this. If addresses from 0 to (m-1) are added to the storage location (hash table),
In the hash method, one function h: S → [0, 1, 2,... (M−1)] is determined, and the address of the character string x of S is obtained by h (x). The function h is a hash function, and the value h (x) is a hash function of x.
It is called an address. The hash method is usually used when the size of a character string set S is much larger than the number of addresses m. Therefore, no matter how the hash function h is selected, different character strings x ₁ ,
x ₂ with respect to _{h (x 1) = h (} x 2) and the case may occur made.
This is called a collision, and an external hash method is used as one of the measures against the collision. In the external hash method, as shown in FIG. 14, a linked list (name
next) An LST is prepared, and x where h (x) = i is stored sequentially from the head of the linked list. Note that each list having the same hash address is called a bucket.

FIG. 15 is a hash table (dictionary) when the external hash method is used for creating a dictionary of the LZW code and searching the dictionary.
In a hash address i specified by a certain character string x, an address (next address) for storing a character K (extension extension) following the character string x and a character other than the character K following the character string x ) And the storage address (first address) of the character further following the character K. The first address corresponds to the index dictionary in FIG. 14, and the next address is the linked list (n
ame next).

FIGS. 16A and 16B are explanatory diagrams of a dictionary structure using the external hash method. FIG. 16A shows a dictionary using the conventional LZW encoding, FIG. 16B shows a dictionary using the external hash method, and FIG. It is a tree structure diagram, in the order shown in FIG.
This is a case where an input character string consisting of three characters a, b, and c occurs. In the address i of FIG. 16B, the first column, the next column,
An extension column is associated with the extension column, and stores data in the structure shown in FIG. That is, a character K linked to the character string x indicating the address i is written in the extension field of the address i, an address storing characters other than the character K linked to the character string x is written in the next field, The column stores a storage address (first address) of a character further linked to the character K. For example, if attention is paid to the character b of the address 4, the address 4 is designated by the first address of the character (a character string composed of one character) of the address 1 and the e of the address 4
The character b connected to the character string a is written in the xtension field, the address 10 for storing another character a connected to the character string a is written in the next field, and the character c further following the character b is written in the first field. Address 6 is written.

Initially, the extensions of addresses 1, 2, and 3
In the column, all the one-character strings a, b, and c are initially registered, and the other columns are “empty (= 0)”, and thereafter, code processing by the external hash method described later is performed, and FIG. A dictionary (FIG. 16B) is created in a tree structure shown in c). In (c), the numbers enclosed by squares are addresses. Thus, for example, referring to the character a at the address 1, the character a at the address 4 is linked to the character b at the first direction, the character b is further linked to the character c at the address 6 at the first direction,
Further, the character a of the address 10 is linked to the character a,
It is shown that the character a at the address 10 is sequentially connected to the character a at the addresses 11 and 12. Address 2
Focusing on the character b, the character b of the address 5 is connected to the character b in the first direction, and thereafter, the characters b and a of the addresses 8 and 9 are sequentially connected. Further focusing on the character c at the address 3, the character c has the address 7
Is connected in the first direction.

Encoding Process Using External Hash Method FIG. 17 is a flowchart of LZW encoding process using the external hash method. In this encoding process, an address of a character string obtained by adding one character to the character string of the reference number i by the external hash method is subtracted as a hash address (index). The linked list stores first and next addresses for storing a character to be added to the character string of reference number i, and checks whether the character and the input character K match, and if they do not match, sequentially processes the linked list. As a result, all the one-character additional character strings that have appeared so far can be searched. If the added character string does not exist in the bucket, 0 is finally obtained from the linked address of the list, and it can be known that the corresponding character string is not registered.

A character string consisting of one character for all characters in advance
(a, b, c,...) are initially registered in the extension fields of dictionary addresses 1 to M (M is the number of character types), and the leading address n of the dictionary is set to the number of character types M + 1 (M + 1 → n). Further, an address (a reference number of the character) at which the first character K is inputted and the character is stored is defined as i, and this is represented by a prefix string (prefix st).
ring). Further, the contents first [1, NMAX] of the first field of all addresses, the contents next [1, NMAX] of the next field, and the contents of the extension fields of the addresses M + 1 to NMAX are all initialized to 0. ..Step 301

In this state, the next character K is inputted (step 302), i is substituted into ω (i → ω, the reference number of the character string up to immediately before K is ω), and j = 0. (Step 303). Also, the candidate character of the current address i
The data first (i) indicating the address for storing the candidate character to be connected to ext (i) in the first direction is set to i (step 30).
4). If the candidate character ext (i) at the current address i does not include a character connected in the first direction, first (i) = 0 and i = 0
Becomes

Next, it is determined whether or not i = 0, in other words, it is checked whether or not there is a candidate character to be connected in the first direction (step 305).
The reference number (address) ω saved in step 3 is the code word code
(Ω) (step 306).

Thereafter, i = n, n is incremented by 1 (n + 1 → n), and the character K input in step 302 is written in the extension field of the address i.
(K → ext (i)). That is, the subsequent character K is registered in the dictionary (step 307). Next, it is checked whether j = 0 (step 308). If j = 0, i → first
(ω) (step 309). As a result, i (address for storing the current character K) is written in the first column of the address for storing the character input immediately before K (= the address indicated by the reference number ω to the character input immediately before K). Will be.

Thereafter, the address of the character K input in step 302 is set as i (step 310), and it is checked whether the data is completed (step 311).
After setting ω, ω is output as a code word code (ω) (step 312), and the encoding process is terminated. If the data is not completed, the process returns to step 302 and the subsequent processes are repeated.

On the other hand, if i ≠ 0 in step 305, in other words, if there is a candidate character to be connected in the first direction, the character (the character ext (i) written in the extension field of the address i) is It is checked whether it matches the character K input in step 302 (step 313). If they match, the process jumps to step 311. If the data ends, i
After setting ω, ω is output as a code word code (ω) (step 312), and the encoding process is terminated. If the data is not completed, the process returns to step 302, and the next character is input. The subsequent search processing for the longest matching character string is repeated.

In step 313, if the candidate character to be connected in the first direction does not match the character K input in step 302, i is substituted for j and the address i
The address data next (i) written in the “next” column is set as a new i (step 314), and the process returns to step 305. If there are no characters connected in the next direction, the address i
0 is written in the next column of, and i = 0.

Thereafter, if i ≠ 0, the flow shifts to step 313 to repeat the same longest matching character string search process.
When the matching character is no longer present, i = 0 in step 305, the reference number (address) ω stored in step 303 is output as the code word code (ω), and the above-described processing is repeated. Incidentally, if i = 0 is determined in step 305 immediately after the processing of step 314, step 3
At 08, j ≠ 0, and i → next (ω) is set (step 315). As a result, i (address for storing the current character K) is written in the next column of the address indicated by the reference number ω up to the character input immediately before K.

In summary, when a new character K is input, candidate characters to be connected to the previous character string are obtained in the first direction. If found, the same characters are obtained in the first direction. Examine, if found, firs again
In the direction t, the same process is repeated to output the code word code (ω) of the longest matching character string with ω being the reference number i when no more characters are found. First, next, extension, etc. are registered. When the character string shown at the top of FIG. 8 is coded and output according to the above flow chart, the character string is registered in the dictionary as shown at the bottom, and FIG. 18, FIG. 19, and FIG.
, The dictionary registration amount increases. FIG. 18A shows a state after the initialization.

FIG. 21 is a block diagram of a conventional dictionary search circuit using the external hash method. The MPU (microprocessor unit) 1 reads the input character K and stores it in the register 2a of the match checking unit 2, and also stores the candidate character K 'and the first address fω and next connected to it from the dictionary memory 3.
The address nω is read and latched in the registers 4a, 4b, 4c of the reading unit 4, respectively. The comparison circuit 2b of the match checking unit 2 performs a comparison check whether the input character K matches the candidate character K 'latched in the register 4a. If they do not match, the controller 5 controls the multiplexer (MP
X) The next latched in the register 4c by 4d
The address nω is selected. Thereby, MPU1 ne
A dictionary search is performed at the xt address nω, and a new candidate character K ′ is searched.
And the first address fω and the next address nω connected to it
Are read, and the registers 4a and 4
b and 4c to perform a comparison test.

On the other hand, in the comparison circuit 2b, the input character K
And the candidate K ', the controller 5 causes the multiplexer 4d to select the first address fω latched in the register 4b. This gives M
PU1 performs a dictionary search using the first address fω, and finds a new candidate character K ′ and the first address fω and the next
The address nω is read out and latched in the registers 4a, 4b, 4c of the reading unit 4, and the next input character K is read out and stored in the register 2a.

Thereafter, the above processing is performed, and no match is obtained in the comparison circuit 2b.
In other words, search should be performed in the connection detection unit 6
When it is confirmed that the first address fω and the next address nω are no longer present, the search for the longest matching character string is completed. At this point, the dictionary search is stopped, and the search for the longest matching character string is performed for the next input character. Do.

[0038]

As described above, in the LZW encoding process using the external hash method, the address of the candidate character K 'to be connected to the end of a certain character string is specified, and the candidate character K' is added to the address. And first address and next
Since the address is stored, there is an advantage that the dictionary search can be performed at a higher speed than the conventional LZW coding that does not use the external hash method. However, in the dictionary search by the external hash method, only one candidate character K 'and one set of the first address and the next address can be read for one dictionary access. Is a problem.

Accordingly, it is an object of the present invention to provide a dictionary search method capable of performing a high-speed dictionary search by the external hash method. Another object of the present invention is to perform a dictionary search by reading a plurality of candidate characters by a single dictionary search and collating the plurality of candidate characters with one input character at a time in a dictionary search of LZW encoding by an external hash method. The purpose of the present invention is to provide a dictionary search method that can perform the search at high speed.

Still another object of the present invention is to read out a plurality of addresses together with a plurality of candidate characters by a single dictionary search, and compare and match the plurality of candidate characters with one input character (coincidence with the first candidate, A plurality of candidate characters to be referred next are immediately read out from the predetermined address based on the second candidate or not matched with any of the candidates, and compared and compared to provide a dictionary search method capable of high-speed dictionary search. It is to be.

[0041]

FIG. 1 illustrates the principle of the present invention.
FIG. 11 is a plurality of candidate sentences connected to the searched character string
Search for multiple data elements so that characters can be searched
Part encoded and stored at the address specified by the
A dictionary memory for storing minute character strings;
Or read candidate characters, addresses, etc. from the dictionary memory.
MPU to output or register new character strings in dictionary memory
(Processor), 13 read out simultaneously from dictionary memory
A register section 14 for storing a plurality of pieces of data.
Comparison matching that matches characters and multiple candidate characters
, 15 based on the result of the comparison.
Address selection section (multiplexer MP)
X). The plurality of data elements include, for example, (1) a first character (ext₁) And (2)
Character string number up to one character (ω₁) And (3) the searched sentence
The second character that is connected to the character string and is different from the first character
(Ext _Two) And (4) the number of the character string up to the second character (ω_Two)
And (5) first and second characters connected to the searched character string.
Is the storage address of another third character (next_Two) And (6) First sentence
Storage address of the fourth character connected to the character (first₁) And (7)
The storage address of the fifth character connected to the second character (first_Two)
And (8) how many of the first and second characters are stored
It has a flag (flag) shown.

[0042]

A plurality of data elements are stored in an address of the dictionary memory 11 specified by the searched character string so that a plurality of candidate characters connected to the searched character string can be searched, and a dictionary is created. When searching for a character string, the MPU 12 reads the next input character following the searched character string,
A plurality of candidate characters to be linked to the searched character string and data elements including address data designating the position of the next candidate character are collectively read out from the dictionary memory 11 and registered in the register unit 1.
3 is stored. The comparison / matching unit 14 compares a plurality of candidate characters with one input character to perform matching and, if there is a matching candidate character, includes the next plurality of candidate characters based on the address data. The data element is read from the dictionary memory and stored in the register unit 13. Thereafter, the next input character is compared with the next plurality of candidate characters, and the longest match search process is continued. As described above, a plurality of candidate characters are read out by one dictionary search, and the plurality of candidate characters are collated with one input character at a time to perform a dictionary search, so that a dictionary search can be performed at high speed. . Also, a plurality of addresses are read together with a plurality of candidate characters by a single dictionary search, and the result of comparison and comparison between the plurality of candidate characters and one input character (matching with the first candidate, matching with the second candidate, matching with any of the candidates) A plurality of candidate characters to be referred next can be immediately read out from the predetermined address based on the non-coincidence, and compared and collated to perform a dictionary search at high speed.

Further, the data element includes a first character (ext ₁ ) connected to the searched character string and a second character (ext ₂ ) connected to the search character and different from the first character. A storage address (next ₂ ) of a third character different from the first and second characters linked to the search character string, a storage address of a fourth character linked to the first character (first ₁ ), Fifth concatenation of two characters
It has at least a character storage address (first ₂ ) and a flag (flag) indicating how many of the first and second characters are stored. (1) One input character and a plurality of candidate characters When matching the first and second characters, the input character and the first and second characters are compared.
If the second character is different, the next data element is read out based on the address (next ₂ ), and a plurality of candidate characters indicated by the data element are matched with the input character, and (2) the input character and the second If one character matches, the address (firs
The data element for the next input character is read based on t ₁ ), and a plurality of candidate characters specified by the data element are matched with the next input character, and (3) the input character matches the second character In this case, the data element for the next input character is read based on the address (first ₂ ), and a plurality of candidate characters indicated by the data element are matched and matched with the next input character to continue the longest match search processing. I do. With this configuration, a plurality of addresses are read together with a plurality of candidate characters by a single dictionary search, and the result of comparison and comparison between the plurality of candidate characters and one input character (matching the first candidate, matching the second candidate, A plurality of candidate characters to be referred to next are immediately read out from the predetermined address based on the result of the comparison, and the comparison and comparison can be continuously performed, and the dictionary search can be performed at high speed.

[0044]

FIG. 2 is an explanatory view of the structure of data stored at one address of a dictionary memory according to the present invention. The address i (= ω ₁ ) specified by a certain character string x includes: (1) the first character (ext ₁ ) connected to the last character of the character string x;
(2) a character string number up to the first character (ω ₁ ), (3) a second character (ext ₂ ) which is a character connected to the last character and is different from the first character, (4) A character string number up to the second character (ω ₂ ), (5) a storage address (next ₂ ) of a third character different from the first and second characters connected to the last character, and (6) The storage address (first ₁ ) of the fourth character linked to one character,
(7) The storage address of the fifth character linked to the second character (first
₂ ) and (8) flags indicating how many of the first and second characters are stored are stored, and a dictionary is created.

FIG. 3 is a diagram for explaining the contents of a dictionary memory according to the present invention. FIG. 3 (a) is a diagram for explaining encoding, and FIG. 3 (b) is a dictionary according to the present invention. first
Column ₁ , first ₂ column, next ₂ column, ext ₁ column, ext ₂ column, ω ₁ column, ω ₂
A column is provided. In the order shown in the upper part of FIG.
When an input character string consisting of three characters b and c is generated, a code word is output as shown in the middle part by an encoding process described later, and the character string is registered in a dictionary as shown in the lower part. In this dictionary registration, a character string is registered at each address of the dictionary memory in the data structure of FIG. 2, and its contents are as shown in FIG. 3 (b).
The tree structure shown in (c) is obtained.

For example, the first character a (ext₁)
, The first character a is first ₁Address 4 in direction
(First₁Address) ext₁Field, ext_TwoCharacter stored in field
Etc., and the character string up to the first character a (one character string a)
Reference number (= 1) is ω₁Field indicates that
Is done.

In the ext ₁ column of the fourth address designated in the first ₁ column of the address 1, the character a in the ext ₁ column of the address 1 is stored.
The first character b to be linked to (the last character of one character string a) is written, and the ext ₂ column of the fourth address contains the ext _{1 of} the address _1.
A second character a to be connected to the character a in the column is written, and the first character b is assigned an address 6 (first ₁ address) in the first ₁ direction.
The characters stored in the ext ₁ column are concatenated, and the second character a is firs
The characters stored in the ext ₁ column of address 11 (first ₂ address) are connected in the t ₂ direction, and the reference number (= 4) of the character string (2 character strings ab) up to the first character b is stored in the ω ₁ column. The reference number (= 1) of the stored character string up to the second character a (two character strings aa)
0) are indicated as being stored in omega ₂ column. Since the next ₂ column of the fourth address is “empty (= 0)”, there is no character string connected to the character a (ext ₁ ) of the address 1 other than the first and second characters b and a. You can see that.

[0048] The ext ₁ column sixth address specified by the first ₁ column address 4, the ext ₁ column address 4 characters b (2
First character c for connecting to the last character) strings ab is written, first the reference number of the string to the character c (3 string abc) (= 6) is stored in the omega ₁ column Is shown. Since the contents of the flag column of the sixth address are 1-0, it is understood that there is no character to be connected to the first character b (ext ₁ ) of the address 4 other than the character c. Further, since the first ₁ column sixth address is "empty (= 0)", 3 strings abc
It can be seen that there is no character linked to.

The eleventh designated in the first ₂ column of address 4
In the ext ₁ column of the address, the character a in the ext ₂ column of the address 4
The first letter a is written for coupling to (2 characters last character string aa), said first character address in the first ₁ direction a 12
The characters stored in the ext ₁ column of (first ₁ address) are linked, and the reference number (= 11) of the character string (3 character strings aaa) up to the first character a is stored in the ω ₁ column. Is shown. Since the contents of the flag column of the sixth address are 1-0, it can be seen that there is no other character to be connected to the second character a (ext ₂ ) of the address 4.

The first specified in the first ₁ column of the address 11
The ext ₁ column 2 address, first letter a coupling to ext ₁ column of character a the address 11 (the last character of the three-character string aaa)
Is written, and the character string up to the first character a (4 character strings aa
aa) the reference number (= 12) is shown to be stored in the omega ₁ column. The contents of the flag column of the twelfth address are
Since it is 1-0, it is understood that the first character a (ext ₁ ) of the address 11 has no other characters to be linked. Also, the first
Since the first ₁ column of the two addresses is "empty (= 0)", it can be seen that there are no characters linked to the four character string aaa. Similarly, a character string linked to the character b at address 2 and the character c at address 3 is registered in the dictionary.

FIGS. 4 and 5 are flowcharts of the encoding process (dictionary search and dictionary registration) according to the present invention. In advance, character codes (a, b, c,...) Are initially registered in the ext ₁ column (ext ₁ [1, M]) of addresses 1 to M of the dictionary memory (M is the number of character types) and ω ₁ The address (reference number) corresponding to the character code is initially registered in the field (ω ₁ [1, M]), and further, the flag field (flag)
Initially register 1-0 (indicating that characters are registered only in the ext ₁ column) in [1, M]).

The head address n of the dictionary is set to M + 1 (M + 1 → n). In addition, the contents of the (1) first ₁ column of all addresses in the dictionary first ₁ [1, NMAX], (2) the contents of the first ₂ column
_{first 2 [1, NMAX],} (3) the content of the next ₂ column _{next 2 [1, NMAX],} (4)
ext ₂ column content ext ₂ [1, NMAX], (5) ω ₂ column content ω ₂ [1, NMA
X] are all initialized to 0, and addresses M + 1 to NMA
Initialize the contents ext ₁ [M + 1, NMAX] of the (6) ext ₁ column and (7) the contents ω ₁ [N + 1, NMAX] of the ω ₁ column of X to 0, and (8) flag Column flag
[N + 1, NMAX] are all initialized to 0-0 (indicating that no characters are registered in the ext ₁ and ext ₂ columns).

Further, the search switching parameter T and the registration switching parameter U are set to 0, respectively, and the first input character K is input, the reference number of the character is set to i, and this is referred to as a prefix string. I do. Note that T = 0 is fir
means that from st ₁ column address indicated by the address data and outputs the reference number of the candidate character when it and the code word output reading the following plurality of candidate characters, T = 1, the
reading the next plurality of candidate characters from the address indicated by the address data in the first ₂ column;
This means to output the reference number of the candidate character. Also,
U = 0 is to register characters in ext ₁ Box during dictionary registration, U = 1 means that register with ext ₂ column.・ ・
Step 401

In this state, the next input character K is input (step 402), and i is substituted for ω (i → ω, and the reference number up to the character immediately before the input character K is ω). , J = 0 (step 403). Then T
= 0 (step 404), and if T = 0, the first ₁ address (first ₁ (i)) at the address i storing the immediately preceding character is set as a new i (step 40).
5) If T = 1, the first ₂ address (first ₂ (i)) at the address i for storing the immediately preceding character is newly set to i (step 406).

Thereafter, it is checked whether i = 0 (step 407). If i ≠ 0, the i-th address
It is checked whether the first candidate character ext ₁ (i) in the ext ₁ column matches the input character K (step 408), and if it matches, T = 0 (step 409) and the process jumps to step 410. If they do not match, the process jumps to step 421 to be described later, where the input character is compared with the second candidate character.

After setting T = 0 in step 409, i is changed to ω
(Step 410). That is, the address i at which the first candidate character that matches the input character is stored is ω. Next, it is checked that the data has been completed (step 411). If the data is not completed, the process returns to step 402, the next character K is input, and the subsequent processing is repeated to search for the longest matching character string. On the other hand, if the data is completed, it is checked whether T = 0 (step 412), and T = 0
If so, the reference number ω of the first candidate character held in step 410 is output as a code word code (ω) (step 4
13), end the encoding process.

On the other hand, at step 408, the i-th address
If the first candidate character ext ₁ (i) in the ext ₁ column does not match the input character K, it is checked whether flag (i) of the i-th address is 1-0, that is, whether a second candidate character exists ( Step 42
1). If the second candidate character exists, ext _{2 of} the i-th address
It is checked whether the second candidate character ext ₂ (i) in the column matches the input character K (step 422), and if it matches, T = 1 (step 423), and the routine jumps to step 410. If they do not match, the process jumps to step 425 to be described later, where the input character is compared with another candidate character.

After setting T = 1 in step 423, i is changed to ω
(Step 410). That is, the address i for storing the second candidate character that matches the input character is ω. Next, it is checked that the data has been completed (step 411). If the data is not completed, the process returns to step 402, the next character K is input, and the subsequent processing is repeated to search for the longest matching character string. On the other hand, if the data is completed, it is checked whether T = 0 (step 412), and T = 1
, The reference number ω ₂ (ω) of the second candidate character is replaced with the code word cod
Output as e (ω ₂ (ω)) (step 414), and end the encoding process.

At step 421, the fl of the i-th address
If ag (i) is 1-0, in other words, if the second candidate character does not exist, there is no candidate character that matches the input character, and U = 1 (step 424). Then, the output of the code word and the dictionary registration processing after step 426 are performed. On the other hand, if the second candidate character ext ₂ (i) in the ext ₂ column of the i-th address does not match the input character K in step 422, in other words, the first and second candidate characters must match the input character. For example, i is substituted for j, U = 0, and the storage address next of candidate characters other than the first and second candidate characters
₂ (i) as a new _{i (next 2 (i) →} i). If there is no other candidate character (candidate character linked in the next ₂ direction), next ₂ (i) = 0 and i = 0. ..Step 425

Thereafter, the flow returns to step 407 to check whether i = 0, and if i ≠ 0, there is another candidate character, so that the processing from step 408 onward is repeated. However, if another candidate character does not exist, i = 0, and thereafter, the output of the code word and the dictionary registration processing after step 426 are performed. If there are no more candidate characters matching the input character, it is checked in step 426 whether T = 0. T = 0
If so, the reference number ω of the first candidate character held in step 403 is output as a codeword code (ω) (step 427), and if T = 1, the reference number ω ₂ of the second candidate character
(Ω) is output as a code word code (ω ₂ (ω)) (step 428).

After outputting the code word, i is substituted for p, and n
Is substituted for i, and n is further incremented by 1 (step 429), and it is checked whether U = 0 (step 43).
0). Note that U = 1 only when the flag is 1-0 and only the first candidate character is stored at the target address. If U = 1 in step 430, the input character K is assigned to the p-th address.
(K → ext ₂ (p)) is written in the ext ₂ column of (the address where the last input character was stored), and 1-1 is written in the flag column.
(1-1 → flag (p)). This registers that the input character K is linked to the immediately preceding character. ... Step 4
31

Then, i is written in the ω ₂ column of the p-th address (the address where the last character was stored) (i → ω
₂ (p)). Thereby, the reference number of the string up to the current input character K is registered in the omega ₂ column (Step 43
2). Hereinafter, the reference number of the character K is set to i, and
Is substituted for ω, and T and U are set to 0 (step 43).
3) Then, it is checked that the data is completed (step 411). If the data is not completed, step 40
Returning to step 2, the next input character is read, and the subsequent processing is repeated. On the other hand, if the data has been completed, it is checked whether T = 0 (step 412).
The final character ω held in 33 is output as the code word code (ω) (step 413), and the encoding process ends.

On the other hand, if U = 0 in step 430,
This input character K is written in the ext ₁ column of the i-th address (a new address in which nothing is stored) (K → ext
₁ (i)), 1-0 is written in the flag column (1-0 → flag (i)).
As a result, a character string in which the current input character K is linked to the last character of the previous character string (the immediately preceding input character) is registered. Step 441 Then, it is checked whether j = 0 (step 442).
If i = 0 after the process of step 425, that is, the first and second candidates exist, they do not match with each other, and ne
j ≠ 0 when the xt ₂ column is 0, and j = 0 otherwise.

If j ≠ 0, i (the storage address of the current character) is written into the next ₂ column of the address j (i → ne
xt ₂ (j), step 443), and thereafter, the processing of step 433 and thereafter is repeated. If j = 0, that is, if the first candidate character does not exist, or if it matches the first or second candidate character and there is no character linked to these matching candidate characters, check if T = 0. (Step 444), T
If = 0, i (the storage address of the current character) is written in the first ₁ column of the storage address ω of the immediately preceding input character (i
→ fitst ₁ (ω), step 445), if T = 1, i
(The storage address of the current character) is written in the first ₂ column of the storage address ω of the immediately preceding input character (i → fitst ₂ (ω), step 446), and thereafter the processing of step 433 and thereafter is repeated.

In summary, the first dictionary search allows
1, second candidate character ext₁, Ext_TwoTogether with multiple references
Address next where the candidate character to be stored is next_Two, first₁, f
irst _TwoAnd (1) the input character is the first candidate character
First if matches₁Should be referenced next to address
Read multiple candidate characters and addresses immediately and enter next
Performs comparison and comparison with characters, and (2) the input character is a second candidate
First if matches character_TwoRefer to next from address
Read multiple candidate characters and addresses immediately
The input character is compared and collated, and (3) the input character is
If they do not match both the first and second candidate characters, next_Two
Multiple candidate characters and addresses to be referenced next to the address
Immediately read out and compare and match with this input character,
If no matching character is found in the first or next direction, the dictionary
Terminates the search and outputs the code word of the longest matching character string.
To register the dictionary.
Start searching. According to the above flow chart, the top of FIG.
As the character string shown in the column is encoded and output, the dictionary shown in FIG.
Is created.

FIG. 6 shows a first example of the dictionary search circuit according to the present invention.
This is an embodiment of the present invention. An MPU (microprocessor unit) 12 inputs an input character K via a DMA circuit (not shown).
Is read and stored in the first register 14a of the comparison and collation unit 14, and the dictionary memory 11 is accessed using the reference number of the immediately preceding input character as an address, and the following data (1) The first character (ext ₁ ), (2) the number of the character string up to the first character (ω ₁ ), and (3) the second character (ext ₂ ) which is a character connected to the predetermined character and which is different from the first character
And (4) the number of the character string up to the second character (ω ₂ ), and (5)
A storage address (next ₂ ) of a third character different from the first and second characters connected to the predetermined character, (6) a storage address (first ₁ ) of a fourth character connected to the first character, and ( 7) The storage address (first ₂ ) of the fifth character to be connected to the second character, and (8) the first,
A flag (fla) indicating how many of the second characters are stored
g), and issues a dictionary search command to the controller 16. As a result, the controller 16 collectively stores predetermined data of the data in the register unit 13. That is, the flag data is stored in the register 13a, and the first ₁
The address is stored in the register 13b, the first ₂ address is stored in the register 13c, the next ₂ address is stored in the register 13d, and the first
The candidate character ext ₁ (= K ₁ ) is latched in the register 13e and the second candidate character ext ₂ (= K ₂ ) is latched in the register 13f at a time.

Then, under the control of the controller 16, the first and second comparison circuits 14b and 14c of the comparison and collation unit 14 input the first and second candidate characters K ₁ and K ₂ and the input latched in the register 14a. The character K is compared and matched at the same time. Note that the comparison circuits 14b and 14c receive the flag data, and
It is recognized whether (1) both the first and second candidate characters exist, (2) only the first candidate character exists, and (3) whether both the first and second candidate characters do not exist. ing.

As a result of the comparison and collation, if the input character matches the first candidate character, the controller 16 causes the address selecting section (multiplexer) 15 to select and output the first ₁ address stored in the register 13b.

[0069] connection detecting unit 17 determines whether the first ₁ address is 0, since the candidate character is not present in the longer first ₁ direction if 0, notifies the no candidate character MPU 12, first ₁ addresses If not 0, the address is M
Notify PU12. When notified of the first ₁ address, the MPU 12 reads the next input character, and
And the dictionary memory 11 is accessed using the first ₁ address, and the read data is stored in the register unit 13 under the control of the controller 16, and thereafter the above-described comparison and collation operation is repeated.

On the other hand, the MPU 12 detects the
When receiving no candidate character, it notifies the controller 16 that the search for the longest matching character string has been completed, creates a code word and outputs it from an I / O port (not shown), and registers the dictionary in the dictionary memory 11. I do. Thereafter, if the input data is not completed, a dictionary search is instructed to the controller 16 and the same operation is repeated for the next input character string.

The above is the case where the input character matches the first candidate character in the comparison operation by the comparison / matching unit 14, but if the input character matches the second candidate character, the controller 16 sets the address selection unit. Do 15 and register 13
Select and output the first ₂ address stored in c. The connection detecting unit 17 determines whether or not the first ₂ address is 0. If the address is 0, there is no candidate character in the first ₂ direction anymore, so the MPU 12 notifies the MPU 12 that there is no candidate character, and fi
If the rst ₂ address is not 0, the address is
Notify. MPU12 is if it is notified first ₂ address, dictionary memory 11 with stores read the next input character to the register 14a, the first ₂ address
And the read data is stored in the register section 13 under the control of the controller 16, and thereafter the above-described comparison and collation operation is repeated.

On the other hand, the MPU 12 outputs
When receiving no candidate character, it notifies the controller 16 that the search for the longest matching character string has been completed, creates a code word and outputs it from an I / O port (not shown), and registers the dictionary in the dictionary memory 11. I do. Thereafter, if the input data is not completed, a dictionary search is instructed to the controller 16 and the same operation is repeated for the next input character string.

If the input character does not match both the first and second candidate characters as a result of the comparison and collation, the controller 16 operates the address selecting section 15 to change the next ₂ address stored in the register 13d. Select and output. The connection detecting unit 17 determines whether or not the next ₂ address is 0. If it is 0, no candidate character exists in the next ₂ direction, so the MPU 12 notifies the MPU 12 that there is no candidate character, and the next ₂ address must be 0. If so, the MPU 12 is notified of the next ₂ address. When notified of the next ₂ address, the MPU 12 accesses the dictionary memory 11 using the address, and
The read data is stored in the register section 13 under the control of the controller 16, and thereafter, the above-described collating operation is repeated.

On the other hand, the MPU 12 outputs
When receiving no candidate character, the controller 16 notifies the controller 16 that the search for the longest matching character string has been completed, creates and outputs a code word, and registers the dictionary in the dictionary memory 11. Thereafter, if the input data is not completed, a dictionary search is instructed to the controller 16 and the above operation is repeated.

FIG. 7 shows another embodiment of the dictionary search circuit according to the present invention, and the same parts as those in FIG. 6 are denoted by the same reference numerals.
The difference from the configuration of FIG. 6 is that (1) the reference numbers ω ₁ and ω ₂ of the first candidate character and the second candidate character are not stored in the dictionary memory 11, and (2) from the first address to the reference number (address). (3) that the MPU 12 can easily perform the code word output process. The address data (first ₁ (i)) written in the first ₁ column and the reference number (ω ₁ ) of the first candidate character specified by the address data match, and the address data written in the first ₂ column The address data (first ₂ (i)) corresponding to the address matches the reference number (ω ₂ ) of the second candidate character specified by the address data. Therefore, even if the reference numbers ω ₁ and ω ₂ of the first candidate character and the second candidate character are not stored in the dictionary memory 11, they can be obtained from the address data first ₁ (i) and first ₂ (i).

Therefore, the input character matches the first candidate character.
The first in the conversion memory₁(i) is the first reference number W₁
And the input character matches the second candidate character
Has first_Two(i) is the first reference number W₁Stored as Soshi
And the first reference number W₁First or second candidate sentence specified by
If the character matches the next input character, the first reference number W
₁To the second reference number W_TwoNishi (W₁→ W_Two), And then,
New first reference number W in the manner of₁Is searched for and memorized.
However, the first reference number W₁1st or 2nd weather indicated by
If the complement does not match the next input character,
Second reference number W _TwoCodeword of the longest matching character string searched for
Is input to the MPU 12. In this way, MP
No need to remember the reference number ω that U12 matched each time
Only, shortens the time between entering characters for the next search after a match
can do.

The present invention has been described with reference to the embodiments.
The present invention can be variously modified in accordance with the gist of the present invention described in the claims, and the present invention does not exclude these.

[0078]

As described above, according to the present invention, a plurality of candidate characters are read out by one dictionary search, and a plurality of candidate characters are
Since the dictionary search is performed by matching one input character at a time, the dictionary search can be performed at high speed. Further, according to the present invention, a plurality of addresses are read together with a plurality of candidate characters by a single dictionary search, and a plurality of candidate characters and one
Comparison result with two input characters (match with first candidate, second match
A plurality of candidate characters to be referred next are immediately read out from the predetermined address and compared and collated based on the candidate, matching none, etc. Search can be performed at high speed.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating the principle of the present invention.

FIG. 2 is a diagram illustrating the structure of a dictionary memory according to the present invention.

FIG. 3 is an explanatory diagram of dictionary contents according to the present invention.

FIG. 4 is a first flowchart of an encoding process according to the present invention.

FIG. 5 is a second flowchart of the encoding process of the present invention.

FIG. 6 is a first configuration diagram of a dictionary search circuit according to the present invention.

FIG. 7 is another configuration diagram of the dictionary search circuit according to the present invention.

FIG. 8 is an explanatory diagram of LZW encoding.

FIG. 9 is an explanatory diagram of a dictionary configuration.

FIG. 10 is a flowchart of LZW encoding.

FIG. 11 is a flowchart of LZW decoding.

FIG. 12 is an explanatory diagram at the time of exception of LZW decoding.

FIG. 13 is an explanatory diagram of LZW decoding.

FIG. 14 is an explanatory diagram of an external hash method

FIG. 15 is an explanatory diagram of a data structure by an external hash method.

FIG. 16 is an explanatory diagram of a dictionary structure by an external hash method.

FIG. 17 is a flowchart of dictionary search and dictionary registration for LZW decoding by the external hash method.

FIG. 18 is a first explanatory chart showing a state of dictionary registration.

FIG. 19 is a second explanatory diagram showing a state of dictionary registration.

FIG. 20 is a third explanatory diagram showing a state of dictionary registration.

FIG. 21 is a configuration diagram of a conventional dictionary search circuit using an external hash method.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 11 Dictionary memory 12 MPU 13 Register part 14 Comparison collation part 15 Address selection part

────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Hirotaka Chiba 1015 Kamiodanaka, Nakahara-ku, Kawasaki City, Kanagawa Prefecture Inside Fujitsu Limited (56) References JP-A-60-116228 (JP, A) JP-A-3-68219 (JP, A) JP-A-3-204233 (JP, A) JP-A-59-231683 (JP, A) JP-A-58-155589 (JP, A) JP-A-61-13340 (JP, A) ( 58) Field surveyed (Int.Cl. ⁷ , DB name) H03M 7/40

Claims (2)

(57) [Claims] 1. A character string that has already been encoded is divided into different partial character strings, the partial character strings are registered in a dictionary, and a partial character string that is the longest match with the input character string is searched from the dictionary. In the dictionary search method in data compression in which the number of the longest matching character string is designated and encoded, a plurality of data elements are added to the searched character string so that a plurality of candidate characters connected to the searched character string can be searched. Is stored in the storage area designated by the user, and the partial character string is registered in the dictionary. When searching for the longest matching character string, the data elements corresponding to the searched character string are collectively read from the dictionary and included in the data element. A plurality of candidate characters are compared with one input character for matching and, if there is a candidate character that matches the input character, a data element is read from the storage area specified by the candidate character and the next input character is read. Letters and Dictionary search method characterized by continuing the longest match search processing by comparing the plurality of candidate characters in the data element. 2. The data element includes a first character (ext ₁ ) connected to a searched character string, and a second character (ext) connected to the searched character string and different from the first character. ₂ ), a storage address (next ₂ ) of a third character different from the first and second characters connected to the searched character string, and a fourth address connected to the first character.
It has a character storage address (first ₁ ), a fifth character storage address (first ₂ ) linked to the second character, and a flag (flag) indicating how many of the first and second characters are stored. When the input character and the first and second characters are different from each other in matching and matching one input character and the two first and second characters which are a plurality of candidate characters, the following is performed based on the address (next ₂ ). The data element is read and matched for matching. If the input character matches the first character, the data element for the next input character is read based on the address (first ₁ ), and the longest match search process is continued. 2. The dictionary search method according to claim 1, wherein when the character and the second character match, the data element for the next input character is read based on the address (first ₂ ) and the longest match search process is continued. .