JP2799228B2 - Dictionary initialization method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Table of Contents] Outline Industrial field of application Conventional technology Problems to be solved by the invention Means for solving the problem Operation Embodiment Effects of the invention [Overview] Compress data In order to prevent the reduction of the data compression rate due to the initialization of the dictionary without lengthening the required time, different character strings were registered in the dictionary before the relevant character strings. While sequentially registering in the dictionary represented by the reference number of the character string and the increment, the dictionary is initialized based on the search notification introduced in the dictionary initialization method when encoding the input character string and compressing the data. Counting means for counting the number of times each element registered in the dictionary is searched, and, in response to the initialization instruction, detecting an infrequently used element based on the counting result by the counting means, and storing at least one storage area of the dictionary. Frequent use Thereby divided into a plurality of blocks separated by low elements of,
A dividing unit that outputs division information regarding division of these blocks, a moving unit that moves a storage location in the dictionary for each of a plurality of blocks based on the division information, and a moving unit that moves based on the division information. Means for changing the reference number included in each element of the dictionary according to the storage location of the corresponding element after the movement of the corresponding element, deleting the infrequently used elements and changing the storage location of the other elements in the dictionary. In other words, it is configured to secure a storage location for registering a new element.

[Industrial applications]

The present invention relates to a dictionary initialization method during incremental decomposition type Ziv-Lempel encoding and decoding.

In recent years, various types of data such as character codes, vector information, and image information have been handled by computers, and the amount of data handled has rapidly increased.

When storing or transmitting such a huge amount of data, it is desirable to reduce the amount of data by omitting redundant portions included in the data. Therefore, there is a demand for a method of efficiently compressing data regardless of the type of data.

The universal encoding method has a feature that it can be applied to the above-described various data compression because it is not necessary to define a code table in advance.

Here, in this specification, one word unit of data is referred to as “character”, and data of a plurality of continuous words is referred to as “character string”.
Called.

The Ziv-Lempel code is a typical method of the universal code described above (see "Ziv-L" written by Munakata).
empel data compression method ", Information Processing, Vol. 26, No. 1, 1985), a universal type algorithm and an incremental decomposition type algorithm have been proposed. Furthermore, as an improvement of the universal algorithm, LZSS codes (TCBell, “Bet
ter OPM / L Text Compression ", IEEE Trans.on Commun.,
Vol.COM-34, No. 12, Dec. 1986). As an improvement of the incremental decomposition type algorithm, LZW code (TAWelch, “AT
ethnique for High-Performance Date Compression ", C
omputer, June 1984).

Among these encoding methods, the LZW code is used for file compression of a storage device or the like because high-speed processing is possible and the algorithm is simple.

[Conventional technology]

The incremental decomposition type algorithm decomposes an input character string into a series of components formed by adding one character as an increment to a substring already registered in the dictionary, and decomposes this series of components into a registered substring. The input character string is encoded by expressing the reference number and the increment corresponding to. Further, the above-described components are registered in the dictionary as new subsequences, and are used in subsequent encoding processing.

Further, in the LZW code, the above-described increment is incorporated in the next subsequence.

Hereinafter, for the sake of simplicity, “a”, “b”,
A character string consisting of three characters "c""ababcbababaaaaa ...
When "" (see FIG. 6 (a)) is input,
The LZW encoding method will be described.

In this case, the three characters "a", "b", and "c" described above are registered in the dictionary addresses "1", "2", and "3", respectively, and the encoding process is started.

First, the first character (for example, the character "a") of the above-mentioned input character string is read, this character is searched from the dictionary, and the address of the dictionary (for example, "1") where this character is stored is referred to as a reference number ω. I do.

Thereafter, the second and subsequent characters of the input character string are sequentially read, and this character is set as an extended character K corresponding to the increment described above, and is represented by the combination (ωK) of the reference number ω and the extended character K described above. A subsequence (ωK) (hereinafter, the combination (ωK) is referred to as a subsequence expression) is searched from the dictionary. If the corresponding subsequence (ωK) is found, the code corresponding to the subsequence (ωK) is replaced by a new reference number ω
Then, the next character of the input character string is read, and the above-described processing is repeated.

In this way, the character string to be encoded is sequentially extended one character at a time, and this character string is sequentially searched from the dictionary. The subsequence that matches the part of interest the longest is searched, and the reference number ω corresponding to this subsequence is searched.
Is output as the corresponding code. At this time,
A subsequence obtained by adding the extended character K to the subsequence (ω) corresponding to the reference number ω is represented by a combination (ωK) of the reference number ω and the extended character K, is given a reference number, and is a dictionary as a new subsequence. Registered in.

In this way, the character string shown in FIG. 6 (a) is decomposed into the underlined subsequences in the figure, and as shown in FIG. Sign "1",
"2", "4", ... are output. FIG. 6 (c)
Table 1 shows the correspondence between the input character strings and the substrings registered in the dictionary, and Table 1 shows an example of the dictionary created.

Further, the dictionary created as described above has a tree-like configuration as shown in FIG. 7, and each of the elements of the dictionary corresponds to each node of the dictionary tree. In FIG. 7, the numbers in parentheses at each node indicate the reference numbers of the corresponding dictionary elements. In the encoding process of the character string shown in FIG. The path followed by each node is indicated by a thick solid line.

Given a code, the LZW is obtained by tracing the dictionary tree in reverse from the node corresponding to the branch or leaf of the dictionary tree whose reference number is the reference number toward the root.
A code decoding process is performed.

First, the surface (ω′K) of the subsequence registered in the dictionary is determined using the input code ω as a reference number. At this time, the extended character K is stored in the stack, and then the substring expression registered in the dictionary corresponding to the reference number ω ′ is searched. The search is sequentially performed in this manner, and the above-described processing is repeated until the expression of the obtained partial string is only the extended character K. Thereafter, by popping up and outputting the extended characters K held in the stack in order from the one held last, the partial sequence corresponding to the input code ω is restored.

Here, the extended character K obtained at the end is the first character of the character string corresponding to the code ω, and a combination (ω _OLD K) of the extended character K and the code _OLD immediately before is represented by a new subsequence. It is registered in the dictionary as an expression.

As described above, in the LZW encoding method, by performing the encoding process and registering a new code in the dictionary, while learning the statistical properties of the input data,
The amount of data is compressed efficiently.

Therefore, if the statistical properties of the input data change, the dictionary stored so far cannot efficiently compress the input data having the new properties, so that it is necessary to perform learning again. .

Here, if a sufficiently large capacity is allocated to the dictionary, all the learning histories can be stored and re-learning on the properties of new data can be performed. Resources are finite. For this reason, usually, when registering up to the capacity allocated to the dictionary, the compression rate obtained as the ratio of the data amount before compression to the data amount after compression is checked, and the compression rate is reduced. Determined that re-learning was necessary, and initialized the dictionary in the same manner as when the above-described encoding process was started.

[Problems to be solved by the invention]

By the way, in the above-described conventional method, when the dictionary is initialized, all the learning histories up to that point are discarded. However, when the number of subsequences registered in the dictionary is small, that is, in the initial stage of the above-described learning, the compression ratio is generally low, and when the number of times of the initialization processing is large, a dictionary having a sufficient capacity is required. There is a problem that the compression ratio is reduced as compared with the ideal case provided.

As a technique for suppressing a decrease in the compression ratio due to the initialization processing, the present applicant has already proposed Japanese Patent Application No. 2-45164 “Data Compression Method”.

The dictionary initialization method proposed by the present applicant provides a means for counting the frequency of use for each of the elements registered in the dictionary, leaving the frequently used elements in the dictionary without deleting them, and deleting the low frequency elements. Thus, an area for registering a new code is secured, and a part of the learning history up to that time is left.

However, in the dictionary initialization method proposed by the present applicant, every time an element to be deleted is detected, a process of changing the storage area and a process of changing the expression of the subsequence are performed for all the elements of the dictionary. Therefore, when the number of elements dictionary and n, n ² times the access process with respect to the dictionary is required, it has the disadvantage that the time required for initialization is prolonged.

The present invention has been made in view of such a point, and a dictionary initialization method for preventing a decrease in compression rate due to dictionary initialization without increasing a time required for dictionary initialization processing. The purpose is to provide.

[Means for solving the problem]

 FIG. 1 is a block diagram showing the principle of the present invention.

(I) In the figure, different character strings are represented by a reference number corresponding to a storage location of a character string registered in the dictionary 110 before the character string and an increment of one character. Counting in the dictionary initialization method when compressing data by encoding the input character string by the reference number of the character string registered in the dictionary 110 while sequentially registering the expression of the character string as a new element in the dictionary 110 The means 111 counts the number of times each element registered in the dictionary 110 has been searched, based on the introduced search notification.

The dividing unit 121 detects an infrequently used element based on the counting result by the counting unit 111 in response to the initialization instruction,
While dividing the storage area of the dictionary 110 into a plurality of blocks separated by at least one infrequently used element,
The division information regarding the division of these blocks is output.

The moving means 131 moves the storage location in the dictionary 110 for each of a plurality of blocks by an infrequently used element that separates the blocks, based on the division information output by the dividing means 121.

The changing unit 141 changes the reference number included in each element of the dictionary 110 moved by the moving unit 131 based on the division information, in accordance with the storage location of the corresponding element after the movement.

As a whole, the configuration is such that the infrequently used elements are deleted, and the storage locations of the other elements in the dictionary 110 are reduced to secure storage locations for registering new elements.

(Ii) The invention of claim 2 According to the invention of claim 2, in the dictionary initialization method according to the invention of claim 1, the changing means 141 sets the block to which the element corresponding to the reference number included in each element belongs by two. It is configured to detect by a minute search and change the reference number based on the movement amount of the storage location of the corresponding block.

(Operation)

(I) The invention of claim 1 In the invention of claim 1, the number of times each element registered in the dictionary 110 has been searched is counted by the counting means 111 based on the introduced search notification, and an initialization instruction is issued. In response to the,
Based on the counting result by the counting means 111, the dividing means 1
According to 21, an element that is used less frequently is detected. The dividing unit 121 divides the storage area of the dictionary 110 into a plurality of blocks separated by at least one infrequently used element, and outputs division information on the division of these blocks.

Based on the division information output by the division unit 121, the storage unit in the dictionary 110 is moved by the movement unit 131 by the infrequently used elements separating the blocks for each of the plurality of blocks described above. .

Further, based on the division information described above, the reference number included in each element of the dictionary 110 moved by the moving unit 131 by the changing unit 141 is changed to the reference number corresponding to the storage location of the corresponding element after the movement. Be changed.

In this way, the less frequently used elements are deleted, the storage locations of the other elements in the dictionary 110 are reduced, and the storage locations for registering new elements are secured.

According to the first aspect of the present invention, before performing the moving process of each element, the storage area of the dictionary 110 is divided into a plurality of blocks divided by the infrequently used elements to be deleted,
The storage location of the elements of the dictionary 110 is moved for each of these blocks. Therefore, every time an element to be deleted is detected, the processing amount is greatly reduced as compared with the case where the storage locations of all the elements registered in the subsequent storage locations are moved. As a result, it is possible to save a part of the learning history without increasing the time required for the initialization processing, and to prevent a reduction in the compression ratio due to the initialization.

(Ii) Invention of Claim 2 According to the invention of claim 2, the block to which the element corresponding to the reference number included in each element belongs is detected by the changing means 141 by a binary search, and The reference number is changed based on the amount of movement of the storage location.

According to the second aspect of the present invention, the time required for the reference number changing process can be reduced by detecting the corresponding block using the binary search method, and the time required for the initialization process is further reduced. can do.

〔Example〕

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 2 shows a configuration of a data compression apparatus to which a dictionary initialization method according to one embodiment of the present invention is applied.

FIG. 5 shows a configuration of a data restoration apparatus to which a dictionary initialization method according to one embodiment of the present invention is applied.

Here, the correspondence between FIG. 1 and the embodiment will be described.

 The dictionary 110 corresponds to the dictionary 221.

 The counting means 111 corresponds to the counter 222.

The division unit 121 corresponds to the division processing unit 231 of the initialization processing unit 230.

The movement unit 131 corresponds to the movement processing unit 232 of the initialization processing unit 230.

The change unit 141 corresponds to the change processing unit 233 of the initialization processing unit 230.

The configuration and operation of the embodiment will be described below assuming that there is the above correspondence.

In FIG. 2, 210 is an encoding unit, 220 is a memory,
Reference numeral 230 denotes an initialization processing unit, and in the memory 220, reference numeral 221 denotes a dictionary.

The dictionary 221 is divided into N _max areas, and each area is assigned an address of “1” to “N _max ”. Each of these areas is Different substrings are registered.

Further, corresponding to each of the N _max number of regions of the dictionary 221 as described above, N _max number of counters 222 _1, ..., 222 _Nmax is provided, these counters 222 _1, ..., the 222 _Nmax, “0” is stored as the initial value of the count value. Hereinafter, these counters 222 ₁ ,..., 222 _Nmax will be simply referred to as the counter 222.

FIG. 3 is a flowchart showing a data compression operation according to the embodiment.

For each of the characters constituting the input character string, a character string consisting of one character is registered in the dictionary 221 to initialize the dictionary 221 (step 301), and thereafter, the encoding process is started.

For example, the characters "a", "b",
When encoding a character string consisting of “c”, the character string consisting of one character of each of these characters “a”, “b”, and “c” is dictionaryd.
As elements of the 221, the addresses “1”, “2”, and “3” are registered, and “4” is set to the registration start address n indicating the area of the dictionary 221 to be registered next.

First, the first character of the input character string is read, and the address of the dictionary 221 in which the character is registered is set as a reference number ω (step 302). It is set to K (step 303).

If it is determined in step 304 that there is still a character to be read (in the case of a positive determination), the dictionary 221 searches the dictionary 221 for a substring represented by the combination (ωK) of the reference character ω and the extended character K described above ( Step 305).

If the corresponding substring is registered in the dictionary 221,
An affirmative determination is made in step 306, and the address at which the substring represented by the combination (ωK) is registered is set as a new reference number ω (step 307), and the count of the counter 222 corresponding to this reference number ω is counted. Increment (step 308), and return to step 303.

As described above, in the data compression device according to the embodiment, the process of incrementing the count of the corresponding counter 222 in step 308 is added to the conventional LZW encoding algorithm (step 301 to step 307).

For example, when encoding the first character “a” of the character string shown in FIG. 6A, the combination (1b) is added to the address “4” of the dictionary 221 and the second character “b” "When encoding
The combination (2a) is registered at the address “5” of the dictionary 221. Therefore, when encoding the third character "a", the subsequently read character "b" is combined with the extended character K and the reference number "1" corresponding to the character "a" (1b). Has already been registered, the affirmative determination is made in step 306 described above, and the counter 222 corresponding to the address “4” of the dictionary 221 is registered.
The count value of ₄ is incremented.

Also, when encoding by focusing on the character string starting with the eighth character “b” shown in FIG. 6A, first, the next character “a” is extended to the character “b”. The expression (2a) representing the substring “ba” added as the character K is searched from the dictionary 221. Further, a substring expression (5b) representing "bab" to which the next character "b" is added is searched. Therefore, in this case, the counter 222 _5, 222 ₈ corresponding to each of the address "5" and the address "8" is incremented.

In this way, every time each substring is searched in step 305, the count value of the corresponding counter 222 is incremented by one. Accordingly, the count value of each of the counters 222 is the number of times the branch of the dictionary tree has been extended via the corresponding node of the dictionary tree created by the above-described encoding process, that is, the corresponding reference number is Indicates the frequency used in the conversion process.

On the other hand, in the case of a negative determination in step 306, the reference number ω is output as a code (step 309), and the combination (ωK) described above is registered as an element of the dictionary 221 at the registration start address n (step 310). Thus, the reference number “n” corresponding to the subsequence represented by the combination (ωK)
Is defined.

The address where the extended character K is registered at this time is set as a new reference number ω (step 311), and the registration start address n is incremented (step 312).

Then, in step 313, the registration start address n
Is compared with the maximum address _{Nmax of the} dictionary 221. If the registration start address n is smaller (positive determination), the process returns to step 303 described above, and in step 304, there is no character to be read (negative determination). If so, the reference number ω at that time is output as a code (step 314), and the process ends.

On the other hand, in the case of a negative determination in step 313 described above, it is determined that a new code cannot be registered in the dictionary 221 and the encoding unit 210 performs an initialization process described below to the initialization processing unit 230. Request (step 315).

In response to this initialization request, the initialization processing unit 230
A process is performed to delete a part of the elements registered in 21 and pack the elements stored in the empty area without being deleted.

In the initialization processing unit 230, reference numeral 231 denotes a division processing unit;
2 indicates a movement processing unit, and 233 indicates a change processing unit.

The division processing unit 231 determines an element to be deleted and an element to be stored based on the counting result of the N _max counters 222 described above, divides the dictionary 221 into a plurality of blocks, and includes an element included in each block. The amount of movement of the storage location is calculated.

Based on the movement amount calculated by the division processing unit 231, the movement processing unit 232 moves the storage location of the element to be stored.

Further, the change processing unit 233 changes the expression of the substring stored in the dictionary 221 based on the new address.

FIG. 4 (a) is a flowchart showing the operation of the division processing unit 231, FIG. 4 (b) is a flowchart showing the operation of the movement processing unit 232, and FIG. 4 (c) shows the operation of the change processing unit 233. 4 shows a flowchart.

The division processing unit 231 previously sets the minimum address N _min after compressing the dictionary 221 as the initial value of the variable i indicating the address of interest of the dictionary 221. For example, as described above, when the input character string is composed of three characters, "4" may be set as the minimum address _Nmin .
Further, an initial value “0” is set for the variable s and the variable k.

Also, an array Slid consisting of the amount of change in storage location corresponding to each of a plurality of blocks generated by the division process
e and an array Pos consisting of the last address of each block are defined, and the “0” th component Slide of the array Slide described above is defined.
A numerical value “0” is set as [0].

First, the division processing unit 231 compares the count value Count [i] of the counter 222 _i corresponding to the area of the address i of the dictionary 221 with a predetermined threshold Th, and determines whether the element of the dictionary 221 is referred to with high frequency. It is determined whether or not this element should be deleted based on whether or not (step 401).

When the count value Count [i] is larger than the threshold Th (negative determination in step 401), the division process 231 determines that the corresponding element is an element that is frequently used and should be stored, and the variable i Is incremented (step 40
2) Return to step 401.

On the other hand, in the case of an affirmative determination in step 401, the value of variable i is set to another variable j (step 403), and it is determined whether or not the j-th element should be deleted in the same manner as in step 401 described above. (Step 404).

For example, the count value Count [6] of the counter 222 corresponding to the element registered at the address “6” of the dictionary 221 is equal to the threshold value.
If it is equal to or less than Th, an affirmative determination is made in step 401 described above, and an affirmative determination is made in step 404, and the variable i and the variable s are each incremented (step 405), and the process returns to step 404.

Further, the counter 222 ₇ corresponding to the element of the address “7”
If the count value Count [7] is equal to or smaller than the threshold value Th, Step 4
At 05, the variables j and s are further incremented.

In this way, by incrementing the variable s each time an element to be deleted is detected, the number of elements to be deleted that have been detected so far is counted. Therefore, this variable s indicates the number by which the address of the next element to be stored should be moved, that is, the amount of change in the address.

Thereafter, when a frequently used element to be stored is detected, a negative determination is made in step 404. In this case, the division processing unit 231 substitutes the value obtained by subtracting the value “1” from the variable i into the k-th component Pos [k] of the array Pos,
The variable k is incremented (step 406).

Here, the element corresponding to the address indicated by the variable i is the element to be deleted. Therefore, from the variable i, the numerical value "1"
Is subtracted as the final address of the k-th block, whereby the preceding portion where the elements to be stored are continuous is separated as the k-th block. Also,
As described above, the number of blocks generated by division is counted by incrementing the variable k every time the block is divided into the element blocks of the dictionary 221.

For example, if the element registered at the address “6” is the first detected element to be deleted, the array
The address “5” is set to the 0th component Pos [0] of Pos, and the variable k is incremented to “1”.

The value of the variable s indicating the number of elements to be deleted detected up to the k-th block is the k-th component Slide [k of the array Slide as the amount of change in the address of the element included in the k-th block. ] (Step 407).

For example, if the elements registered at addresses “6” and “7” in the dictionary 221 are determined to be elements to be deleted, the value of the variable s is “2”, and this value is
Is assigned to the first component Slide [1] of the array Slide.

Next, the division processing unit 231 sets the value of the variable j to the variable i (step 408), compares the value of the variable i with the maximum address _Nmax (step 409), and sets the variable i to the maximum address _Nmax.
If smaller than (Yes in step 409)
Returns to step 401 and repeats the above-described processing.

In this way, the dictionary 221 is divided into a plurality of blocks separated by the element to be deleted, and each block is sequentially given the block number k, and the variable i is set to the maximum address.
When N _max is reached, a negative determination is made in step 409.

In this case, division processing unit 231 substitutes the maximum address N _max to k-th component Pos [k] of SEQ Pos, holds the variable k at this time as a division number k _m (step 410), the processing To end.

At this time, the division processing unit 231 also
de, movement processor arrays Pos and division number k _m as division information
Introduced in 232, requesting a transfer process.

Here, the dictionary 221 and the count value of the counter 222 are shown in the second table in correspondence with each other.
Is shown as a result of performing the dividing process with “3”. In Table 2, numerals in the block column indicate the numbers of blocks to which each element belongs.

In response to the above-described movement processing request, the movement processing unit 232
First, the initial value “0” is assigned to the variable k, and 0 of the array Pos is assigned to the variable p.
The value of the nth component Pos [0] is set to the number of elements n of the dictionary 221 after the initialization processing from the above-described maximum address _Nmax to Slide [km].
Are subtracted from each other, and the minimum address N _min (for example, “4”) is set as an initial value in a variable i (step
421). Also, an array W composed of reference numbers ω and an array K composed of extended characters K registered in the dictionary 221 as a subsequence expression (ωK) are defined. Hereinafter, the elements corresponding to the address i of these arrays W and K are denoted by W, respectively.
[I] and K [i].

Next, the movement processing unit 232 compares the variable i with the variable p and determines whether or not the element of the dictionary 221 corresponding to the variable i is included in the k-th block (Step 422).

If the value of the variable i does not exceed the value of the variable p (a positive determination in step 422), it is determined that the corresponding element belongs to the k-th block. In this case, the movement processing unit 232 determines whether the component W [i + Slide
[K]] to W [i] and component K [i + Slide of array K
[K]] is substituted for K [i] (step 423). This means that in the dictionary 221, the address “i + Slide
This corresponds to moving the expression (ωK) of the substring registered in [k] ”to the address“ i ”.

At this time, the address “i + Slide
[K]], the count value Count [i + Sl of the counter 222.
ide [k]] is set to the counter 222 _i corresponding to the address “i”.
(Step 424).

For example, based on the result of the division processing shown in Table 2,
When performing the movement processing, in step 421, the variable i
Is set to the initial value "4", and the variable p is set to the address "5". In this case, an affirmative determination is made in step 422, and the above-described movement processing is performed. However, in this case, the change amount Slide [0] corresponding to the 0th block
Is "0", the storage location of the corresponding element does not change.

On the other hand, if the value of the variable i exceeds the value of the variable p, it is determined that the corresponding element is included in the next block, and a negative determination is made in step 422. In this case, the variable k indicating the block is incremented, and Pos [k] is incremented by S
The value obtained by subtracting lide [k] is set as a variable p (step 425).

Therefore, the value of the variable p set in step 425 indicates the address of the last element belonging to the k-th block after the movement of the element of the dictionary 221 is completed.

For example, when the value of the variable i becomes “6”, step 422
In step 425, the variables k and p indicating the block number are updated, and the variable p
Is set to the address of the last element of the second block. Accordingly, an affirmative determination is made in step 422, and the above-described movement processing in steps 423 and 424 is performed.
In this case, the change amount Slide corresponding to the first block
Since the value of [1] is “2”, the storage locations of the corresponding elements are each reduced by “2”.

In this way, the moving process of the storage location is performed for each block divided by the division processing unit 231.

After performing the above-described movement processing, the movement processing unit 232
It requests the change processing unit 233 to change the value of the component W [i] of the i component of the array W (step 426),
There is an end notification from the expression changing unit 233 (step 427).

By the way, in the LZW encoding method, a subsequence registered in the dictionary 221 is represented by a reference number ω and an extended character K corresponding to a node closer to the root than a node of the dictionary tree corresponding to itself. . Therefore, when the reference number corresponding to the node changes due to the above-described movement processing, it is necessary to change the expression of the subsequence with a new reference number corresponding to the node.

In response to the change processing request described above, the change processing unit 233
First, an initial value "0" to the variable L indicating the lower limit of the range of binary search, it sets a division number k _m described above as an initial value to a variable H indicating the upper limit (step 441).

Next, an average value m of the variable L and the variable H is obtained (step
442), the value of W [i] described above and the m-th component of array Pos
The value of Pos [m] is compared with the value of Pos [m] (step 443), and it is determined whether the reference number ω registered as the i-th element belongs to section [L, m] or section [m, H]. .

When it is determined that the value of W [i] is equal to or less than the value of Pos [m] (a positive determination in step 443), the change processing unit 233
Determines that reference number ω belongs to section [L, m],
A value obtained by subtracting the numerical value “1” from the average value m is set to the variable H (step 444).

On the other hand, when the value of W [i] is determined to be larger than the value of Pos [m] (negative determination in step 443), the change processing unit 233 determines that the reference number ω belongs to the section [m, H]. Then, a value obtained by adding “1” to the average value m is set to the variable L (step 445).

After that, the change processing unit 233 compares the variable L with the variable H, and determines that the variable L is equal to or less than the variable H (step S1).
The affirmative determination in 446) determines that the corresponding block has not been detected, returns to step 442, and repeats the above-described processing.

On the other hand, when the variable L exceeds the variable H (negative determination in step 446), the change processing unit 233 determines that the corresponding block has been detected, and the i-th component W [i] of the array W
Slide of the array Slide corresponding to the L-th block from
The value obtained by subtracting the value of [L] is set as the element W [i] (step 447), and the above-described movement processing unit 232 is notified of the end of the change processing.

As described above, by using the binary search technique, the corresponding block can be efficiently searched.

In response to the end notification described above, the movement processing unit 232 restarts the process (Yes in step 427), increments the variable i (step 428), and compares the variable i with the number n of elements (step 429). ).

If it is determined that the variable i is smaller than the number of elements n (a positive determination in step 429), the process returns to step 422, and the above-described processing is performed until the variable i becomes equal to the number of elements n and a negative determination is made in step 422. repeat.

After the above-described processing is completed, the arrays W and K
Are sequentially stored in the respective addresses of the dictionary 221 and the initialization process is completed with the number of elements n as the registration start address. In response to the completion notification, the encoding unit 210 restarts the encoding process. Good.

As described above, the amount of movement of each block is calculated in advance, and by moving the storage location for each block,
The time required for the process of moving the storage location can be set to a time proportional to the number of elements registered in the dictionary 221. As described above, it is possible to greatly reduce the processing amount as compared with the case where the entire element of the dictionary 221 is searched in a double loop.

Further, by using the binary search in the change processing unit 233, it is possible to efficiently search for a block including an element corresponding to the reference number included in each element, and based on the corresponding reference number, The processing for changing the surface of the partial row can be speeded up.

As described above, without increasing the time required for initialization, it is possible to save the learning history and prevent a reduction in the compression ratio due to the initialization processing, and to perform data compression efficiently.

In the above-described embodiment, the case where the present invention is applied to a data compression device has been described. However, the present invention may be applied to a data decompression device.

In this case, as shown in FIG. 5, a decoding unit 510 may be provided instead of the encoding unit 210.

The decryption unit 510 uses the conventional LZW
Each time the dictionary 221 is searched, the count value of the counter 222 corresponding to the corresponding element is incremented each time the code is decoded. Also,
The decryption unit 510 requests the initialization processing unit 230 to perform an initialization process when attempting to register in the dictionary 221 exceeding the maximum address _Nmax .

〔The invention's effect〕

As described above, according to the first aspect of the present invention, the dictionary storage area is divided into a plurality of blocks before each element is moved, and the storage location of the dictionary elements is moved for each of these blocks. Is performed, every time an element to be deleted is detected, the processing amount can be significantly reduced as compared with the case where the storage locations of all the elements registered in the subsequent storage locations are moved, Without lengthening the time required for the initialization process, a part of the learning history can be saved, and a reduction in the compression ratio due to the initialization can be prevented, so that the data can be efficiently compressed and restored.

According to the second aspect of the present invention, the time required for the reference number changing process can be reduced by detecting the corresponding block using the binary search method, and the time required for the initialization process can be further reduced. Can be shortened.

[Brief description of the drawings]

1 is a block diagram of the principle of the present invention, FIG. 2 is a configuration diagram of a data compression apparatus to which a dictionary initialization method according to an embodiment of the present invention is applied, FIG. 3 is a flowchart showing an encoding operation according to the embodiment, FIG. 4 is a flowchart showing an initialization operation according to the embodiment, FIG. 5 is a configuration diagram of a data restoration apparatus to which a dictionary initialization system according to the embodiment is applied, FIG. 6 is an explanatory diagram of an LZW encoding system, FIG. FIG. 3 is a diagram showing a configuration of a dictionary. In the figure, 110 is a dictionary, 111 is counting means, 121 is decomposing means, 131 is moving means, 141 is changing means, 210 is an encoding unit, 220 is a memory, 221 is a dictionary, 222 is a counter, and 230 is initialization processing. 231 is a division processing unit, 232 is a movement processing unit, 233 is a change processing unit, and 510 is a decoding unit.

────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Hirotaka Chiba 1015 Kamiodanaka, Nakahara-ku, Kawasaki City, Kanagawa Prefecture Inside Fujitsu Limited (58) Field surveyed (Int.Cl. ⁶ , DB name) G06F 5/00 H03M 7 / 30

Claims (2)

(57) [Claims] A different character string is represented by a reference number corresponding to a storage location of a character string registered in the dictionary (110) before the character string and an increment of one character, and the character string is represented. In the dictionary initialization method when the input character string is encoded by the reference number of the character string registered in the dictionary (110) and the data is compressed while sequentially registering as a new element in the dictionary (110). A counting means (111) for counting the number of times each element registered in the dictionary (110) is searched based on the search notification to be introduced; and a counting means (111) in response to an initialization instruction. Based on the counting result, the less frequently used elements are detected and the dictionary (11
A dividing unit (121) that divides the storage area of (0) into a plurality of blocks separated by at least one infrequently used element, and outputs division information regarding the division of these blocks; The dictionary (110) for each of the plurality of blocks, based on the infrequently used elements separating the blocks, based on the division information output by (110).
A moving means (131) for moving a storage location in the dictionary, and a reference number included in each element of the dictionary (110) moved by the moving means (131) based on the division information after the corresponding element is moved. And a changing means (141) for changing in accordance with the storage location of the storage device. In order to delete the infrequently used elements and to store the storage locations of the other elements in the dictionary (110), A dictionary initialization method characterized in that it is configured to secure a storage location of a dictionary. 2. The change means (141) detects a block to which an element corresponding to a reference number included in each element belongs by a binary search, and based on a movement amount of a storage location of the corresponding block, 2. The dictionary initialization method according to claim 1, wherein reference numbers are changed.