JP3132774B2 - Data compression / decompression device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data compression / decompression device, and more particularly to a data compression / decompression device for compressing and decompressing data by universal encoding and decoding.

In recent years, due to remarkable technological development, the processing speed, storage capacity, and the like of computers have been dramatically improved. However, since the computer handles data such as vector information and image information, the amount of data handled is increasing more than ever. In order to cope with such a large increase in the data amount, various data compression methods have been proposed for reducing the data amount without losing the data contents.

[0003] These data compression schemes are schemes for compressing data by encoding without redundant portions included in the data. With the data compression method, the data amount can be reduced, and as a result, the storage capacity can be reduced. In communication, by transmitting compressed data, the same information can be transmitted at high speed.

[0004] The definition of "Character" and "Character String" are defined in JIS-C6230.
In addition to following the names used in information theory,
Data composed of one word unit is called a "character", and data composed of an arbitrary word unit is called a "character string".

[0005]

2. Description of the Related Art Conventionally, a universal coding method has been proposed as an example of compressing the above data.
As a typical example of the universal encoding method, LZ (Le
mpel-Ziv) coding method and arithmetic coding method. Also, L
Z encoding methods include universal type and incremental decomposition type (Increm
ental persing) algorithm has been proposed. Furthermore, there are LZSS encoding methods belonging to the universal type and LZW (Lempel-Ziv-Welch) encoding methods belonging to the incremental decomposition type as encoding methods in which these algorithms are improved.

The LZ encoding method is described in, for example, “Lempel-Ziv Data Compression Method” by Seiji Munakata, Information Processing, pp. 2-6, V
ol.26, No.1, 1985. Also, L
ZSS encoding is described in TC Bell, "Better OPM / L Text Co.
mpression ", IEEE Trans.onCommu., Vol.COM-34, No.1
2, Dec.1986. Furthermore, LZW
The encoding method is described in TA Welch, "A Technique for High-Perf
Ormance Data Compression ", Computer, Jun. 1984. An incremental decomposition type encoding method and an LZW encoding method are described in JP-A-59-231683, U.S. Pat.
o. 4,558,302.

[0007] Among these coding methods, the LZW coding method has been generally used because of its advantages of high-speed processing and simple algorithm. The LZW encoding method has a rewritable dictionary and performs encoding by the following processing. First, a new input character string is divided into different partial character strings, and if this partial character string is not registered in the dictionary, it is assigned an identification number in the order of appearance and registered in the dictionary. At the same time, of the currently input partial character strings, a partial character string that matches the longest partial character string is selected from the dictionary, and is encoded with the identification number assigned to the selected partial character string.

The details of a data compression circuit and a data decompression circuit using the LZW encoding method will be described below. FIG. 15 is a diagram showing a conventional data compression circuit. In the figure, a conventional data compression circuit includes a dictionary search unit 121,
It comprises a dictionary registration unit 122, a universal encoding unit 123, and a dictionary D12.

The dictionary search means 121 searches the registered character strings registered in the dictionary D12 for a registered character string that matches the input character string. Dictionary registration means 122
Adds an identification number to an input character string that could not be searched by the dictionary search means 121 and registers it in the dictionary 122. The universal encoding unit 123 encodes an identification number assigned to the registered character string searched by the dictionary search unit 121 or an identification number newly assigned to the input character string by the dictionary registration unit 122 into an output code.

[0010] Dictionary search means 121, dictionary registration means 122
The encoding process of the universal encoding unit 123 is performed sequentially. FIG. 16 shows an example of the encoding processing procedure.

FIG. 16 is a diagram showing a processing procedure by a conventional data compression circuit. In the figure, the processing character number column 125,
Dictionary search field 126, dictionary registration field 127, and encoding field 12
8 shows a table.

The number-of-characters-to-be-processed column 125 indicates the order of characters from the beginning to be processed when the input character string is encoded.
The dictionary search field 126 shows the contents processed by the dictionary search means 121 in FIG. The dictionary registration column 127 indicates the content processed by the dictionary registration unit 122. The encoding column 128 indicates the content processed by the universal encoding means 123. The number in the circle indicates the order of the processing procedure. Hereinafter, “the number X in a circle” is referred to as “cycle X”.

First, in cycles 1 to 3, a process in which the dictionary registration means 122 searches the dictionary D12 for the first character of the input character string is shown. More specifically, as a result of the dictionary search, the input character string and the registered character string do not match in cycles 1 and 2, but indicate that they match in cycle 3.

Next, in cycle 4 to cycle 6, the dictionary registration means 1 registers the second character of the input character string.
Reference numeral 22 denotes a process of searching the dictionary D12. In particular,
The input character string and the registered character string do not match in cycles 4 and 5, and cycle 6 indicates that there is no character registered in the dictionary.

At this time, the dictionary registration means 122 adds an identification number to the input character string and registers it in the dictionary D12. This is the “registration” shown in cycle 7. Further, in order to perform encoding, the universal encoding unit 123 encodes an identification number given at the time of dictionary registration. This is the “encoding” shown in cycle 8. Hereinafter, the encoding process is performed similarly.

FIG. 17 is a diagram showing a conventional data restoration circuit. In the figure, the conventional data restoration circuit
Universal decryption means 131, dictionary search means 132,
It comprises a stack storage unit 133, a dictionary registration unit 134, and a dictionary D13.

The universal decoding means 131 decodes the input code. The dictionary search unit 121 searches the registered character codes registered in the dictionary D13 for a character code that matches the decoded character code. The stack storage unit 133 stores the characters searched by the dictionary search unit 121 in a stack, and outputs all the stored characters when the search is completed. Dictionary registration means 134
If the restored character string output by the stack storage unit 133 is not registered in the dictionary D13, the restored character string is assigned a new code and registered in the dictionary D13.

These universal decoding means 131,
The decryption processing of the dictionary search unit 132, the stack storage unit 133, and the dictionary registration unit 134 is performed sequentially. FIG. 18 shows an example of this decoding processing procedure.

FIG. 18 is a diagram showing a processing procedure by a conventional data restoration circuit. In the figure, the processing code number column 136,
Decryption field 137, search / accumulation field 138, character string output field 1
39 shows a table composed of 39 and a dictionary registration field 140.

The processing code number column 136 indicates the number of codes to be processed when the input code is decoded. Decryption field 137
Indicates the contents to be processed by the universal decoding means 131 in FIG. The search / accumulation column 138 stores the dictionary search means 13
2 shows a process in which the stack storage means 133 stores the searched characters. The character string output column 139 shows a process of outputting all the characters accumulated by the stack accumulation means 133. The number in the circle indicates the order of the processing procedure. Hereinafter, similarly to FIG. 16, “the number X in a circle” is described as “cycle X”.

First, in a cycle 1 to a cycle 8, a process from decoding of an input first input code string to outputting as an output character string is shown. Specifically, each processing procedure is as follows. In cycle 1, the universal decoding means 131 decodes the input code. In cycles 2 to 4, the dictionary search unit 132 recursively searches for the character code decoded in cycle 1, and the stack storage unit 133 stores the searched character in the stack. In cycle 5 to cycle 7, the stack storage unit 1
33 outputs all the stored characters. In cycle 8,
The dictionary registration unit 134 assigns a new code to the output character string output by the stack
Register with. Hereinafter, the decoding process is performed similarly.

[0022]

However, in the conventional data compression circuit, after performing a dictionary search, dictionary registration and encoding for a certain character string, the next character string is encoded.
Such compression processing was performed by batch serial processing. Similarly, the data restoration circuit performs restoration processing by batch serial processing, such as decoding a certain code, searching a dictionary, accumulating stacks, and registering a dictionary, and then decoding the next code. I was

Therefore, in such batch serial processing, there is a problem that it is difficult to speed up the encoding and decoding processes. The present invention has been made in view of such a point, and an object of the present invention is to provide a data compression circuit that speeds up the compression processing of input data by parallel processing.

It is another object of the present invention to provide a data restoration circuit which speeds up restoration processing of an input code by parallel processing. Still another object of the present invention is to provide a data compression / decompression device that speeds up the compression processing of input data and the decompression processing of input codes by parallel processing.

[0025]

According to the present invention, in order to achieve the above object, as shown in FIG. 1, a data compression circuit comprises a first dictionary retrieval unit 1, a first dictionary registration unit 2, an encoding unit. 3. Initial dictionary access means 4, normal dictionary access means 5, initial dictionary D1, and normal dictionary D2. First
The dictionary search means 1 searches the registered character strings registered in the initial dictionary D1 via the initial dictionary access means 4 for the longest registered character string that matches the input character string. If the search cannot be performed using the initial dictionary D1, the search is terminated. If the search is successful, the search is performed from the registered character string registered in the normal dictionary D2 via the normal dictionary access means 5. The first dictionary registration unit 2 attaches an identification number to a character string obtained by adding one character to the longest registered character string searched by the first dictionary search unit 1, and sends the normal dictionary D 2 via the normal dictionary access unit 5. Register with. The encoding means 3
The identification number assigned to the longest registered character string is encoded as an output code.

The initial dictionary D1 and the normal dictionary D2 are
It is configured to perform data search and data registration using an external hash. Further, the initial dictionary D1 is configured to perform data search and data registration for character strings up to the second character, and the normal dictionary D2 is configured to perform data search and data registration for character strings after the third character.

Then, the initial dictionary D1 and the normal dictionary D2
Is configured to perform data search and data registration using a complete hash. Then, the first dictionary search means 1,
The first dictionary registration unit 2 and the encoding unit 3 are connected by a pipeline and are processed in parallel.

In addition, as shown in FIG. 3, the data restoration circuit includes a decoding unit 11, a second dictionary search unit 12, a first storage unit 13, a second storage unit 14, an output selection unit 15, 2 dictionary registration means 16 and dictionary access means 17
And the dictionary D3. The decoding unit 11 decodes the input code. Second dictionary search means 12
Searches a registered character string registered in the dictionary D3 via the dictionary access unit 17 for a character string that matches the character string restored by decoding. First storage means 13
The second storage unit 14 stores the character string restored by the search. The output selection unit 15 selects a restored character string to be output from the stored character strings. The second dictionary registration unit 16 registers character strings that are not registered in the dictionary D3 among the selected restored character strings with new codes.

The decoding means 11, the second dictionary search means 12, the output selection means 15, and the second dictionary registration means 16
Are connected in a pipeline and processed in parallel. And
As shown in FIG. 5, the data compression / decompression device includes a data compression circuit and a data decompression circuit. The data compression circuit includes a first dictionary search unit 1, a first dictionary registration unit 2, an encoding unit 3, an initial dictionary access unit 4, a normal dictionary access unit 5, an initial dictionary D1, and a normal dictionary D2. The data restoration circuit includes a decoding unit 11, a second dictionary search unit 12, a first storage unit 13, a second storage unit 14, an output selection unit 15, a second dictionary registration unit 16,
It comprises a dictionary access means 17 and a dictionary D3.

[0030]

In the data compression circuit, the first dictionary retrieval means 1 retrieves the input character string from the initial dictionary D1 via the initial dictionary access means 4. If the search is not performed from the initial dictionary D1, the search is similarly performed from the normal dictionary D2 via the normal dictionary access means 5. Then, the first dictionary registration unit 2 attaches an identification number to the character string obtained by adding one character to the longest registered character string searched, and registers the character string in the normal dictionary D2 via the normal dictionary access unit 5. The encoding means 3 encodes the identification number given to the longest registered character string searched into an output code.

The initial dictionary D1 and the normal dictionary D2 are
Perform data search and data registration by external hash.
Further, data search and data registration are performed for the character string up to the second character in the initial dictionary D1, and data search and data registration are performed for the character string after the third character in the normal dictionary D2.

Then, the initial dictionary D1 and the normal dictionary D2
Performs data search and data registration using a complete hash. Then, the first dictionary search means 1, the first dictionary registration means 2, and the encoding means 3 are connected by a pipeline to perform parallel processing.

Further, in the data restoration circuit, the decoding means 11 decodes the input code, and the second dictionary search means 12 transmits the dictionary D through the dictionary access means 17.
3, a character string that matches the character string restored by decoding is searched, and the restored character string is output to the first storage unit 13 and the second storage unit 14. Output selection means 1
5 selects a restored character string to be output, the second dictionary registration means 16 assigns a new code to the unregistered restored character string, and registers the new character string in the dictionary D3 via the dictionary access means 17.

The decoding means 11, the second dictionary search means 12, the output selection means 15, and the second dictionary registration means 16
Are connected in a pipeline to perform parallel processing. In the data compression / decompression device, the data compression circuit
Dictionary search means 1 via initial dictionary access means 4
The input character string is searched from the initial dictionary D1. If the search is not performed from the initial dictionary D1, the search is similarly performed from the normal dictionary D2 via the normal dictionary access means 5.
Then, the first dictionary registration unit 2 attaches an identification number to the character string obtained by adding one character to the longest registered character string searched, and registers the character string in the normal dictionary D2 via the normal dictionary access unit 5.
The encoding means 3 encodes the identification number given to the longest registered character string searched into an output code. In the data restoration circuit, the decoding means 11 decodes the input code, and the second dictionary search means 12 matches the character string restored by decoding in the dictionary D3 via the dictionary access means 17. Then, the restored character string is output to the first storage unit 13 and the second storage unit 14. The output selecting unit 15 selects the restored character string to be output, and the second dictionary registration unit 16 assigns a new code to the unregistered restored character string and registers the new character string in the dictionary D3 via the dictionary access unit 17.

[0035]

An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram showing an embodiment of the data compression circuit of the present invention. In the figure, the data compression circuit comprises a first dictionary search unit 1, a first dictionary registration unit 2, an encoding unit 3, an initial dictionary access unit 4, a normal dictionary access unit 5, an initial dictionary D1, and a normal dictionary D2. You.

The first dictionary search means 1 is one of the registered character strings registered in the initial dictionary D1 via the initial dictionary access means 4 or registered in the normal dictionary D2 via the normal dictionary access means 5. The longest registered character string that matches the input character string input from the information source is searched for among the registered character strings. The first dictionary registration unit 2 adds an identification number to a character string obtained by adding one character to the searched longest registered character string, and assigns an initial dictionary D1 via the initial dictionary access unit 4.
Or via the normal dictionary access means 5 in the normal dictionary D2. The initial dictionary access unit 4 is a unit for accessing the initial dictionary D1.
The logical address designated by the dictionary search means 1 and the first dictionary registration means 2 is converted into a physical address, and data is input / output to / from the initial dictionary D1. The normal dictionary access unit 5 is a unit for accessing the normal dictionary D2. For example, the normal dictionary access unit 5 converts a logical address specified by the first dictionary search unit 1 and the first dictionary registration unit 2 into a physical address. Performs data input / output with the normal dictionary D2. The encoding unit 3 uses, for example, an LZW encoding unit, and encodes, as an output code, an identification number given to the longest registered character string searched.

Here, the "character string obtained by adding one character to the longest registered character string" registered in the normal dictionary D2 is the following character string. For example, if the input character string is "abcd" and the longest registered character string already registered in the normal dictionary D2 is "ab", the input character string matches the longest registered character string. This is a character string “abc” to which the one character “c” next to “ab” is added.

The access from the initial dictionary access means 4 to the initial dictionary D1 and the access from the normal dictionary access means 5 to the normal dictionary D2 are performed by a complete hash method or an external hash method described later.

Each of the first dictionary search means 1, the first dictionary registration means 2, and the encoding means 3 are connected by a pipeline (not shown), and can be processed in parallel. it can. By this pipeline parallel processing, encoding processing by a plurality of means becomes possible, and encoding processing can be performed at high speed.

An exception in such parallel processing is that the first dictionary search means 1 and the first dictionary registration means 2
Is a request for access to the initial dictionary D1 or the normal dictionary D2 at the same time. For example, there is a case where the first dictionary search unit 1 and the first dictionary registration unit 2 request access to the initial dictionary D1 at the same time. A state in which an access request is made from two or more means at the same time is called "collision". When a collision occurs, for example, after the first dictionary registration unit 2 accesses the initial dictionary D1 via the initial dictionary access unit 4, the first dictionary search unit 1 similarly accesses the initial dictionary D1. Thus, the dictionary is accessed sequentially. However, such a collision is an extremely rare phenomenon in view of the entire encoding process. Therefore, since the processing time is negligible in view of the entire processing time, there is no problem even if the dictionary and the access processing are sequentially performed. Hereinafter, an example of this encoding processing procedure is shown in FIG.

FIG. 2 is a diagram showing a processing procedure by the data compression circuit of the present invention. The figure shows a table including a processing character number column 6, a dictionary search column 7, a dictionary registration column 8, and an encoding column 9. The number-of-processed characters column 6 indicates the character order from the head processed when the input character string is encoded. Dictionary search field 7
Shows the contents processed by the first dictionary search means 1 in FIG. The dictionary registration column 8 shows the contents processed by the first dictionary registration unit 2. The encoding column 9 shows the contents to be processed by the encoding means 3.

The numbers in the circle indicate the order of the processing procedure. Hereinafter, “the number X in a circle” is referred to as “cycle X”. If there is a number in the same circle, it indicates that the processing is performed in parallel at the same time. Then, the initial dictionary D1
The access to the normal dictionary D2 is performed through the normal dictionary access means 5, and the description of this process is omitted.

First, in cycle 1, the first dictionary search means 1 searches the initial dictionary D1 for the second character following the first character of the input character string input from the information source.
In cycle 2, the first dictionary search means 1 searches the normal dictionary D2 for the third character of the input character string. Cycle 3
The first dictionary search means 1 searches the normal dictionary D2 for the fourth character of the input character string. At this time, cycle 3
Shows a case where the fourth character of the input character string is not registered in the normal dictionary D2. These cycles 1 to 3
In this process, each process is performed sequentially.

Next, in cycle 4 of the dictionary search field 7, the first dictionary search means 1 searches the initial dictionary D1 for the fifth character following the fourth character of the input character string. At the same time, in cycle 4 of the dictionary registration section 8, cycle 3 of the dictionary search section 7
Then, an identification number is assigned to the character string of the first to fourth characters of the input character string that has not been registered in the normal dictionary D2, and the first dictionary registration unit 2 registers the character string in the normal dictionary D2.

Then, in cycle 5 of the dictionary search field 7,
For the sixth character following the fifth character of the input character string, the first dictionary search means 1 searches the normal dictionary D2. At the same time, in cycle 5 of the encoding field 9, cycle 4 of the dictionary registration field 8
Encodes the identification number given to the registered character string performed in step. Hereinafter, similar processing is continued.

As a result of searching the initial dictionary D1 or the normal dictionary D2 in the dictionary search field 7, if the input character string is not registered, the registration of the character string in the dictionary registration field 8 and the encoding The encoding in column 9 is performed in parallel with the dictionary search. Therefore, the encoding process can be performed at high speed.

FIG. 3 is a diagram showing an embodiment of the data restoration circuit of the present invention. In the figure, the data restoration circuit includes a decoding unit 11, a second dictionary search unit 12, a first storage unit 13, a second storage unit 14, an output selection unit 15, a second dictionary registration unit 16, and a dictionary access unit. Means 17 and dictionary D3
Consists of

As the decoding means 11, for example, LZW decoding means is used, and decodes an input code inputted from an information source. The second dictionary search means 12 is a dictionary access means 1
7, a character string that matches the character string restored by decoding is searched from the registered character strings registered in the dictionary D3. First storage means 13 and second storage means 1
No. 4 stores the character string restored by the search. The output selection unit 15 selects a restored character string to be output from the stored character strings. The second dictionary registration unit 16 registers character strings that are not registered in the dictionary D3 among the selected restored character strings with new codes. The dictionary access unit 17 is a unit for accessing the dictionary D3. For example, the dictionary access unit 17 converts a logical address instructed from the second dictionary search unit 12 and the second dictionary registration unit 16 into a physical address, and converts the logical address into a physical address. Data is input / output to / from D3.

The dictionary access means 17 sends the dictionary D3
Can be accessed by the full hash method or the external hash method described later. The "collision" described in the data compression circuit of FIG. 1 also occurs in this data restoration circuit. For example, this is a case where the second dictionary search means 12 makes a search request and the second dictionary registration means 16 makes a registration request at the same time. In this case, for example, the second dictionary registration unit 16
After the registration processing, the search processing of the second dictionary search unit 12 is sequentially performed. As in the case of the data compression circuit, collision is a very rare phenomenon from the viewpoint of the entire decoding process, and is only negligible in the overall processing time. There is no problem in performing the processing.

Then, the decoding means 11, the second dictionary search means 12, the first storage means 13, the second storage means 14,
Each of the output selection means 15 and the second dictionary registration means 16 are connected by a pipeline (not shown), and can be processed in parallel. By this pipeline parallel processing, encoding processing by a plurality of means becomes possible, and encoding processing can be performed at high speed. Hereinafter, an example of this decoding processing procedure is shown in FIG.

FIG. 4 is a diagram showing a processing procedure by the data restoration circuit of the present invention. The figure shows a table including a processing code number field 21, a decoding field 22, a search / accumulation field 23, a character string output field 24, and a dictionary registration field 25.

The number of processed codes column 21 indicates the number of codes to be processed when the input code is decoded. The decryption column 22 indicates the content processed by the decryption means 11 of FIG. The search / accumulation column 23 indicates a process in which the second dictionary search means 12 searches and the stack storage means 13 and 14 accumulate the searched characters. The character string output column 24 shows a process in which the output selection unit 15 outputs all the characters stored in the stack storage units 13 and 14. The dictionary registration section 25 stores the second dictionary registration unit 16
Indicates the content to be processed.

The numbers in the circle indicate the order of the processing procedure. Hereinafter, similarly to FIG. 2, “the number X in a circle” is described as “cycle X”. If there is a number in the same circle, it indicates that the processing is performed in parallel at the same time. The access to the initial dictionary D3 is performed by the dictionary access unit 17.
, The description of this process is omitted.

First, in cycle 1, the decoding means 11 decodes the first input code input from the information source. In cycle 2 of the decoding field 22, the decoding means 11 decodes the second input code. At the same time, cycle 2 of the search / accumulation column 23
Searches the dictionary D3 for code characters decoded by. Also,
The retrieved restored characters are stored in the stack by the stack storage unit 13. These cycles 2 are performed in parallel.

Next, in cycle 3 and cycle 4, the second dictionary search means 12 searches the dictionary D3 for code characters decoded by recursive decoding, which will be described later, and the stack storage means 13 restores the searched character. Accumulate characters on the stack.

Then, in cycle 5 of the decryption column 22,
The decoding means 11 decodes the third input code. At the same time, in cycle 5 of the search / accumulation column 23, the code character decoded by the second dictionary search means 12 is searched from the dictionary D3, and the stack storage means 14 stores the searched restored character in the stack. At the same time, the character string output field 24
In cycle 5, the first character of the character string accumulated in cycle 3 to cycle 4 is output. These cycles 5
Processing is performed in parallel.

Then, cycle 6 of the search / accumulation column 23
Retrieves the code characters decoded by the second dictionary retrieval unit 12 from the dictionary D3, and the stack storage unit 14 stores the retrieved restored characters in the stack. At the same time, in cycle 6 of the character string output column 24, the second character of the character string accumulated in cycle 3 to cycle 4 is output. These cycles 6
Are also performed in parallel. Hereinafter, similar processing is continued.

As described above, in parallel with the decoding performed in the decoding section 22, the character retrieved by one of the stack storage units is stored, and the character string stored by the other stack storage unit is output as a restored character string. Processing can be performed. Therefore, the decoding process can be performed at high speed.

FIG. 5 is a diagram showing an embodiment of the data compression / decompression device of the present invention. In the figure, the data compression / decompression device includes a data compression circuit and a data decompression circuit. The data compression circuit includes a first dictionary search unit 1, a first dictionary registration unit 2, an encoding unit 3, an initial dictionary access unit 4, a normal dictionary access unit 5, an initial dictionary D1, and a normal dictionary D2. The data restoration circuit includes a decoding unit 11 and a second dictionary search unit 1.
2, a first storage unit 13, a second storage unit 14, an output selection unit 15, a second dictionary registration unit 16, a dictionary access unit 17, and a dictionary D3.

The same elements as those in FIGS. 1 and 3 are denoted by the same reference numerals, and description thereof will be omitted. The operations of the data compression circuit and the data decompression circuit are the same as those shown in FIGS.

With this apparatus, the processing speed of the data compression processing and the decompression processing is improved. In addition, as in the processing procedure shown in FIGS. 2 and 4, a plurality of units can perform processing in parallel, so that the processing speed of the entire apparatus can be further improved.

Next, the dictionary search procedure and the dictionary registration procedure in each of the above embodiments will be described first with regard to the hash method. The hash method is one of the methods of storing data and retrieving data using a table called a hash table. In order to register data, a storage address is stored using an internal code ω of a search key. How to decide. For this reason, a function for obtaining an address from the internal code ω of the search key is required, and this function is called a “hash function”. The address H (ω) obtained by the hash function H is called “hash address”. In addition,
In the search, similarly to the case of registration, a hash address is obtained by a hash function and data of a target address is searched.

In the dictionary search and dictionary registration using the hash method, no matter how the hash function H is selected, the hash address is H (ω) for the internal codes ω ₁ and ω _{2 of} different search keys. ₁ ) = H (ω ₂ ). Such a state is called "collision", and an external hash method (also called "open hash method" or "chain method") is used to avoid the collision.
Is used. In addition, a method of preparing all possible values for the internal code ω of the search key in a table in advance so as not to cause a collision is a complete hash method. Hereinafter, the case using the external hash method and the case using the complete hash method will be described separately.

FIG. 6 is a diagram showing a data structure of the external hash method. In the external hash method, first, a hash function H
A table corresponding to the address obtained by the above, that is, an array called a bucket header (Bucket Headder) BH is prepared. In the figure, the bucket header BH is provided with b list headers having hash addresses from 0 to (b-1). Then, one of the list headers and one or more data elements having the same hash address, that is, the list L, are connected by a list structure. Here, one list includes a data storage area for storing data and a pointer storage area for storing a pointer to the next list.

For example, as shown in FIG. 6, the list header indicated by the hash address “0” is combined with the list L01 by setting a pointer to the list L01. This relationship is indicated by arrow A1. The list L01 is linked to the list L02 by setting a pointer to the list L02. This relationship is indicated by arrow A2. If the list L02 and subsequent lists are not linked by a pointer, the terminal symbol “0” is set. Hereinafter, the list headers having hash addresses 1 to (b-1) are similarly connected to the list L by the list structure.

The data search is performed in the following procedure. Here, the case where the hash address obtained by the hash function H is “3” will be described as an example. From the list header of the bucket header BH corresponding to the obtained hash address “3”, the first list L3
Obtain a pointer to 1. And the first list L3
Check with the data in 1. If the data does not match, a pointer to the next list L32 is obtained and collated with the data of the next list L32. If the data does not match, it indicates that the desired data was not found.
Note that if a pointer to the next list is set in the list L32, the target data is searched through the same search processing.

The data registration is performed in the following procedure. Similar to the data search, a case where the hash address obtained by the hash function H is “3” will be described as an example. The pointer to the first list L31 is obtained from the list header of the bucket header BH corresponding to the obtained hash address “3”. The list is followed until the pointer to the next list becomes "0". In the example of FIG. 6, since the list L32 is the last list, a new data list is generated here, and a pointer to the generated list is set in the pointer storage area in the list L32.

In a dictionary having such a data structure, data search and registration processing is performed as follows.
FIG. 7 is a diagram showing an example of a dictionary search procedure and a dictionary registration procedure using the external hash method. In the figure, the dictionary is first memory FM, next to speed up the search and registration process.
It is composed of three physical memories, a memory NM and an extention memory EM. These three memories are divided into an initial dictionary and a normal dictionary according to the contents of data to be registered.

First, data retrieval using this dictionary is performed in the following procedure. Here, a case where the character string to be searched is “K ₂₂ K ₃₂ K ₄₂ ” and the hash address of the character string obtained by the hash function H is “ω ₁ ” will be described as an example.

First, the algorithm obtained by the hash function H
Dress ω₁From the address of the first memory FM of the initial dictionary
Su ω₁Pointer ω to the next list stored in_{twenty one}To
get. And a pointer ω to the next list_{twenty one}From
 address ω of extention memory EM_{twenty one}Stored in
Data K_{twenty one}To match. In this case, the character string K_{twenty two}K₃₂K ₄₂
The first letter K_{twenty two}And the next memory NM
Address ω_{twenty one}Pointer to the next list ω_{twenty two}Get
I do.

Next, from the pointer ω ₂₂ to the next list,
To get a pointer ω ₃₁ to the next list that is stored in the first memory FM of the address ω ₂₂ of the normal dictionary. Less than,
A search similar to that of the initial dictionary is performed. Finally has been collated with the data K ₂₁ stored in the address omega ₄₁ of extention memory EM, matching string is terminated can not find. This process is indicated by an arrow in FIG.

The registration of data using this dictionary is performed in the following procedure. Here, a case will be described as an example where a character string “K ₂₂ K ₃₂ K ₄₂ ” that could not be searched above is registered. By the same processing as the above data search, the list is traced until the pointer to the next list becomes "0". In this case, the address of the next memory NM is ω _41. Here, to generate the character K ₄₂ to other addresses extention memory EM. In FIG. 7, this address is ω _42, the next memory NM of the same address is set to "0". Also, to concatenate the lists, ne
The address ω ₄₁ of xt memory NM, to register the ω ₄₂ is a pointer to the list, which was newly created. Thus, a new character string is registered.

In the initial dictionary for performing the data search and data registration, a character string up to the second character is set.
Normally, a dictionary is used for a character string after the third character. This allows the input strings to be optimally distributed,
The processing time when performing data search and data registration can be suppressed.

The initial dictionary is not limited to the character string up to the second character, and the normal dictionary is not limited to the character string up to the third character.
For example, the character string may be changed according to the properties of the input character string, such as a character string of up to the third character in the initial dictionary and a character string of the fourth and subsequent characters in the normal dictionary. By doing so, it is possible to construct an optimal dictionary corresponding to the properties of various input character strings.

FIG. 8 is a diagram showing an example of a dictionary search procedure and a dictionary registration procedure using the complete hash method. The complete hash method is a key search method using a hash function that does not cause any collision. The dictionary constructed by the perfect hash method has, for example, a tree structure shown in FIG.
It has a structure such that each "node" has 256 "branches".

In the figure, the dictionary is a complete hash memory P
It is composed of only M physical memories, and is divided into an initial dictionary and a normal dictionary according to the contents of data to be registered.
The address of the complete hash memory PM is composed of a hash address ω and a data address K in order to correspond one-to-one with data of each character string.

For example, if the hash address ω is an address represented by 20 bits and the data address K is an address represented by 8 bits, the sum of these two addresses is obtained.
An 8-bit address is used as an address of the complete hash memory PM. That is, each hash address ω
Indicates the head address of each of the 256 data addresses K. In FIG. 8, the complete hash memory P
The address of M is represented by “ω _nn · K _xx ”.

The data retrieval using this dictionary is performed in the following procedure. Here, the case where the character string to be searched is “K ₂ K ₃ K ₄ K ₅ ” and the hash address of the character string obtained by the hash function H is “ω ₁ ” will be described as an example.

First, an algorithm obtained by the hash function H
Dress ω₁And the first character K of the character string _TwoFrom the data of
Period dictionary address ω₁・ K_TwoF stored in the
Get L. The flag FL indicates whether this address is valid or not.
The effectiveness is indicated by “1” or “0”. In this case, the flag FL
Is set to "1", the character K_TwoIs valid
You. Therefore, the first character of the string to be searched matches the dictionary.
Indicates that you have done it.

Next, the second character K of the character string_ThreeIs valid or not
To check the address ω₁・ K _TwoStored in
Next pointer ω_{twenty one}To get. And the address ω_{twenty one}
And the second character K of the character string_ThreeFrom the data in
Dress ω_{twenty one}・ K_ThreeThe flag FL stored in the
You. Hereinafter, a search is performed by the same processing. Soshi
And finally the address ω₄₁・ K_FiveFlag stored in
Compared with FL, it ends without finding a matching character string
ing. This process is indicated by an arrow in FIG.

The registration of data using this dictionary is performed in the following procedure. Here, an example will be described in which a character string “K ₂ K ₃ K ₄ K ₅ ” that cannot be searched is registered. By the same processing as the above data search, the list is traced until the pointer to the next list becomes "0". In this case, the address of a complete hash memory PM is ω ₄₁ · K _5. Here, the flag FL is set to “1”. Thus, a new character string is registered.

Next, the LZW encoding means and LZW decoding means shown in the above embodiment will be described. FIG.
FIG. 9 is a diagram illustrating a specific example of a case where an input character string is encoded by an LZW encoding algorithm. This input character string is a character string composed of a combination of only three characters a, b, and c. First, initialization is performed in which only a, b, and c of one character are registered in the dictionary in association with codes 1, 2, and 3, respectively.

First, the input character string 71 is read one character at a time from left to right. The first character a is read, and this a is set as a prefix string. Next, the second character b is read, and ab obtained by adding the b to the initial letter a is compared with the registered character string in the dictionary. At this time, since there is no character string that matches ab in the dictionary, the corresponding code 1 of the initial character a is output as an encoded output. The output code is shown in the output code column 72. At the same time, the character string ab is registered in the dictionary in correspondence with the code 4. The contents registered in this dictionary are shown in the registered contents column 73. Here, let the second input character b be the initial character again.

Next, the third character a of the input character string 71 is read, and ba obtained by adding the letter a to the initial character b is compared with the registered character string in the dictionary. At this time, since there is no character string matching ba, the corresponding code 2 of the initial character b is output as an encoded output, and the character string ba is registered in the dictionary in correspondence with code 5. Further, the third input character a is set as the initial character again.

Further, the fourth character b is read, and ab obtained by adding the b to the initial letter a is collated with the registered character string in the dictionary. At this time, since a character string matching ab is registered in the dictionary, ab is set as the initial character string at this time.
Further, the fifth input character c is read and the initial character string a
The abc obtained by adding c to b is compared with the registered character string in the dictionary. At this time, since there is no character string matching abc in the dictionary, the corresponding code 4 of the initial character string ab is output as an encoded output, and the character string abc is registered in the dictionary in association with the code 6. Then, the fifth input character c is set as the initial character again.

Hereinafter, encoding and dictionary registration are continued by the same algorithm. Encoding is performed on the input character strings a, b, a, b, c,.
Codes 1, 2, 4, and 4 as shown in the output code field 72 in FIG.
Are output as encoded outputs. Then, a registered character string 91 and a corresponding code 92 as shown in FIG.
Is registered in the dictionary.

FIG. 10 is a flowchart showing the procedure of the above-described encoding process. In the figure, numbers following S indicate step numbers. [S101] By initialization, a code is registered in the dictionary in such a manner that a code is associated with all the characters that may be input. The address n to be registered next in the dictionary
Is set to, for example, 256. Here, n is equivalent to the total number of registered character strings when the codes are assigned 0, 1, 2,... Corresponding to the character strings registered in the dictionary. Further, the input character string is read, and the first character input is set as a prefix string ω.

[S102] The next input character K is read. [S103] In step S102, it is determined whether or not input character data exists. If the input character data exists, the process proceeds to step S105, and if not, the process proceeds to step S104.

[S104] The initial character string ω is collated with the dictionary, the corresponding code code (ω) is read, and output as an encoded output. At this time, the number of bits of the code (ω) is converted to a binary code of [log ₂ n] and output. here,
The symbol [x] represents the minimum integer among integers equal to or larger than x. Hereinafter, the symbol [x] will be used in this sense. In this step, since there is no character string to be processed, the present procedure is terminated after executing this step.

[S105] Step S is added to the initial character string ω.
The character string ωK to which the character K read in 102 is added is collated with the dictionary, and it is determined whether or not the character string ωK is registered in the dictionary. If registered, the process proceeds to step S106, and if not registered, the process proceeds to step S107.

[S106] The character string ωK is newly set as the initial character string ω. Then, the process returns to step S102 again. In this way, by repeating steps S102 to S106, a character string that matches the input character string
The longest character string among the character strings registered in the dictionary is searched.

[S107] The initial character string ω is collated with the dictionary, the corresponding code code (ω) is read, and output as an encoded output. The number of bits of the code code (ω) at this time is [log ₂ n]. Further, the value of n is associated with the character string ωK and registered in the dictionary. That is, the dictionary address n
Is stored as a character string ωK. Further, the character K read in step S102 is set as the initial character string ω, and the dictionary address n is incremented to prepare for execution of the next new input character string from step S102.

FIG. 12 is a diagram showing a specific example in the case where the encoded output shown in FIG. 9 is restored. In the dictionary on the restoration side, only codes 1, 2, and 3 are registered in association with characters a, b, and c, respectively, by initialization.

First, the input code 81 is read one by one from left to right. The first code 1 is read, and the character string a is restored with reference to the dictionary. The character string restored at this time is shown in a restored character string column 821. The first code is always registered in the dictionary by initialization. Then, the second code 2 is read, and the character string b is restored with reference to the dictionary. At this time, the previous input code 1 and the first character b of the currently decoded character string
The code 4 is associated with “1b” obtained by combining the above and registered in the dictionary. The registered content at this time is stored in the registered content column 8
3 is shown.

Next, the third code 4 of the input character string 81 is read, and the corresponding "1b" is read with reference to the dictionary. Further, the code | symbol 1 of "1b" reads the corresponding character a with reference to a dictionary. Such a series of read repetition operations is called “recursive decoding”. This is referred to as a recursive decryption field 8
It is shown in FIG. Thus, the character string ab is restored and output as a restored character string. The output character string is shown in the restored character string column 831. At the same time, “2” is a combination of the previous input code 2 and the first character a of the character string restored this time.
"a" is registered in the dictionary in correspondence with the reference numeral 5.

Thereafter, restoration of a character string and registration of a dictionary are continued by the same algorithm. Thus, the input code 1,
Are restored for 2, 4, 3, 5,...
Character strings a, b, a as shown in the restored character string column 821 of FIG.
are output as restored character strings.
Then, the correspondence between the registration code 93 and the corresponding character string 94 as shown in FIG. 11B is registered in the dictionary.

FIG. 13 is a flowchart showing the procedure of the decoding process exemplified above. In the figure, numbers following S indicate step numbers. [S111] Characters are registered in the dictionary in advance by associating characters with codes which may be input by initialization. Also, the address n to be registered next in the dictionary is
For example, it is set to 256. Here, n is equivalent to the total number of registered character strings when the codes are assigned 0, 1, 2,... Corresponding to the character strings registered in the dictionary. Next, the input code is read, and the first input code CODE (binary code) is set to 1
It is converted into the input code ω of the 0-base number. In this case, ω is an input character string in the encoding of FIG.
Is an input code. And this ω
OLD ω. At the same time, since the code to be input first is already registered in the dictionary, the character D corresponding to the input code ω
(Ω) is retrieved from the dictionary and output as restored characters. The output character is set in FINchar for exception processing in step S116 described later.

[S112] The next input code CODE is read. [S113] In step S112, it is determined whether or not the input code data exists. If it exists, the process proceeds to step S115, and if it does not exist, this processing procedure ends.

[S114] The input code CODE is converted into the input code ω, and the input code ω is converted to INω.
Set to. [S115] The input code ω is compared with n. That is, it is determined whether or not the input code is registered in the dictionary (ω ≧ n). If ω is smaller than n, step S11
7 and when ω is equal to or more than n, the process proceeds to step S116. Note that ω is equal to or larger than n when, for example, the input code field 81 in FIG. 12 is “8”.

[S116] OLD ω and FINchar set in step S111 or last time in step S119
(OLD ω, FINchar) is replaced with ωK. That is, the value set in OLD ω is set to ω, and the value set in FINchar is set to K. Then, K is pushed (PUSH) onto the stack. Note that ω is set in step S117.
Is decrypted.

[S117] Normally, since the input code ω has been registered in the dictionary in the previous processing, the character string D (ω) corresponding to the input code ω is read from the dictionary. The read character string D (ω) is decomposed into ω _i K. ω _i is a sign and K is a decoded character. Then, it is determined whether or not the character string D (ω) is one character that cannot be decomposed into ω _i K. D (ω)
If can be decomposed into ω _i K, the process proceeds to step S118,
If D (ω) is one character, the process proceeds to step S119.

[0102] [S118] to push the character K to temporarily stack, also the sign ω _i as a new ω, returns to step S117 again. The execution of steps S117 and S118 is repeated until D (ω) reaches one character.

[S119] Each character pushed on the stack in step S118 is transferred to the LIFO (Last In Fast Out).
Pop (POP) in the format and output the restored character string. For example, if the input code field 81 in FIG. 12 is “5”,
The data is pushed onto the stack in the order of a and b, and a restored character string ba is output. At the same time, the first character of the restored string is set to FINchar, and the previously set OLDω and
A character string consisting of a pair with FINchar (OLDω, FINchar) is registered in the dictionary in correspondence with the value of n. That is, this character string is stored at the address n of the dictionary. Further, n is incremented and the IN set in step S114 is incremented.
ω is set to OLD ω to prepare for execution of the next step S112 and subsequent steps.

As described above, in the decoding process, the data before encoding is restored by repeatedly performing steps S117 to S119 in FIG. That is, since the input code ω has been registered in the dictionary in the previous processing,
The character string D (ω) corresponding to the input code ω is read from the dictionary. Further, the read character string D (ω) is decomposed into ω _i K, and the character K is temporarily saved on the stack. Then, the character string D (ω) corresponding to the input code ω is read from the dictionary again, using the code ω _i as a new input code ω. These procedures are recursively repeated until the new input code ω becomes one character. Then, the character saved on the stack is L
This is a method of popping and outputting in the IFO format.

FIG. 14 is a diagram showing an example of the tree structure of the dictionary used in the above processing procedure and the like. The tree structure of this dictionary is
FIG. 3 illustrates an internal structure of a dictionary used for encoding and decoding by an algorithm realized in the LZW encoding unit and the LZW decoding unit. In FIG. 14, the numbers in the circles indicate identification numbers, and the places where the numbers in the circles are attached are called “nodes (nodes)”.

The dictionary 50 has a root (root) 51 as a starting point. No characters are assigned to this route 51. Then, the first character is registered at one level below the root 51, that is, at the first level 52. The registration of the first character is performed by registering different characters.
Is performed at the time of initialization of Although three characters “a”, “b” and “c” are registered in the figure, 256 characters that can be represented by 8-bit data are actually registered.

[0107] The layers below the second layer 53 are characters registered by learning a character string input from an information source. Note that a node having a lower hierarchy is called a "branch", and a node having a lower hierarchy is called a "leaf". Therefore, in the figure, the numbers 25, 26, 13, 14, 27, 28, 16,
The nodes 6,..., 22, 23, and 24 are “leaves”, and the other nodes are “branches”.

Even if a certain node is currently a “leaf”, it may become a “branch” by learning. For example,
When the character string “acd” is registered in the dictionary 50, the character string “ac” is “a” (the number 1 in the circle) in the first hierarchy 52 and “c” (the number 6 in the circle) in the second hierarchy 53. ), “D” is newly registered in the third hierarchy 54 below “c” of the second hierarchy 53. At this time, the node of number 6 in the circle changes from “leaf” to “branch”.

In the above description of the embodiment, in the data compression circuit, the code is output as the output code having the minimum integer bit number of log ₂ n or more.
Bit fraction compensation, Phasin, as disclosed in No. 23
g in Binary Codes or an output code composed of multi-level arithmetic codes may be used.

Although the initial dictionary D1, the normal dictionary D2, and the dictionary D3 are constructed based on an external hash or a complete hash, and the dictionary 50 is constructed based on a tree structure, the dictionary may be constructed based on another construction method. . For example, a dictionary may be constructed by a binary tree method, and data such as a character string registered in the dictionary may be searched by a binary search.

Each of the above embodiments is applied to compression and decompression of character codes, vector information, image data, and the like in a workstation or the like, and the required storage capacity can be greatly reduced.

Further, the present invention can be applied to data transmission / reception using a communication line, and the communication time can be reduced. For example, it can be applied to communication equipment such as a modem and a facsimile.

[0113]

As described above, according to the present invention, in the data compression circuit, the first dictionary search means searches the initial dictionary or the normal dictionary for the longest registered character string that matches the input character string, and The first dictionary search means attaches an identification number to the character string obtained by adding one character to the longest registered character string and registers it in the normal dictionary, and the encoding means uses the identification number assigned to the longest registered character string as an output code. Since the encoding is performed, the encoding process can be performed quickly.

In the data compression circuit, the initial dictionary and the normal dictionary are configured to perform data search and data registration using an external hash, so that data search and data registration become faster, and the encoding process can be performed faster. it can.

Further, in the data compression circuit, data search and data registration are performed for the character string up to the second character in the initial dictionary, and data search and data registration are performed for the character string after the third character in the normal dictionary. The processing time required for data search and data registration can be reduced.

In the data compression circuit, the initial dictionary and the normal dictionary are configured to perform data search and data registration by using a complete hash, so that data search and data registration become faster, and encoding processing can be performed faster. it can.

Since the first dictionary search means, the first dictionary registration means, and the encoding means perform parallel processing by the pipeline, the encoding processing can be performed at high speed.
In addition, in the data restoration circuit, the decoding means decodes the input code, and at the same time, the second dictionary search means searches the dictionary for a character string restored by decoding, and the first storage means and the second storage means Accumulates the restored string,
Since the output character selecting unit selects the restored character string to be output, and the second dictionary registration unit adds a new code to the unregistered restored character string and registers it in the dictionary, the decoding process can be performed quickly.

Also, the decoding means, the second dictionary search means,
Since the output selection means and the second dictionary registration means perform the parallel processing by the pipeline, the decoding processing can be performed at high speed.

In the data compression / decompression device,
In the data compression circuit, the first dictionary search means searches the initial dictionary or the normal dictionary for the longest registered character string that matches the input character string, and at the same time, the first dictionary search means adds one character to the longest registered character string. The assigned character string is assigned an identification number and registered in a normal dictionary, and the encoding means encodes the identification number assigned to the longest registered character string into an output code. In the data restoration circuit, the decoding means decodes the input code, and at the same time, the second dictionary search means searches the dictionary for a character string restored by decoding, and the first storage means and the second storage means The restored character string is stored, the output selection means selects the restored character string to be output, and the second dictionary registration means assigns a new code to the unregistered restored character string and registers it in the dictionary. The decryption process and the decryption process can be performed quickly.

Therefore, the processing speed of each circuit and the entire device can be improved.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating the principle of a data compression circuit according to the present invention.

FIG. 2 is a diagram showing a processing procedure by a data compression circuit of the present invention.

FIG. 3 is a diagram illustrating the principle of a data restoration circuit according to the present invention.

FIG. 4 is a diagram showing a processing procedure by the data restoration circuit of the present invention.

FIG. 5 is a diagram illustrating the principle of a data compression / decompression device according to the present invention.

FIG. 6 is a diagram showing a data structure of an external hash method.

FIG. 7 is a diagram showing an example of a dictionary search procedure and a dictionary registration procedure using the external hash method.

FIG. 8 is a diagram showing an example of a dictionary search procedure and a dictionary registration procedure using the complete hash method.

FIG. 9 is a diagram illustrating a specific example of encoding.

FIG. 10 is a diagram showing an encoding processing procedure.

FIG. 11 is a diagram illustrating a correspondence relationship between a character string and a code.

FIG. 12 is a diagram illustrating a specific example of decoding.

FIG. 13 is a diagram showing a decoding processing procedure.

FIG. 14 is a diagram illustrating an example of a tree structure of a dictionary.

FIG. 15 is a diagram showing a conventional data compression circuit.

FIG. 16 is a diagram showing a processing procedure by a conventional data compression circuit.

FIG. 17 is a diagram showing a conventional data restoration circuit.

FIG. 18 is a diagram showing a processing procedure by a conventional data restoration circuit.

[Description of Signs] 1 Dictionary search means 2 Dictionary registration means 3 Encoding means 4 Initial dictionary 5 Normal dictionary 11 Decoding means 12 Dictionary search means 13 First storage means 14 Second storage means 15 Output selection means 16 Dictionary registration Means 17 Dictionary

──────────────────────────────────────────────────続 き Continuation of the front page (72) Inventor Hirotaka Chiba 1015 Kamikodanaka, Nakahara-ku, Kawasaki City, Kanagawa Prefecture Inside Fujitsu Limited (56) References JP-A-3-247167 (JP, A) JP-A-62-263529 (JP, A) JP-A-4-267630 (JP, A) JP-A-4-129429 (JP, A) JP-A-4-123619 (JP, A) JP-A-4-76727 (JP, A) Kaihei 3-262331 (JP, A) JP-A-61-13340 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G06F 5/00 H03M ^7/ 30-7/46

Claims (8)

(57) [Claims] A data compression circuit for compressing an input character string input from an information source by encoding the character string, an initial dictionary (D1) for registering the character string, and a character string registered in the initial dictionary. Included in the initial dictionary
Normal dictionary that registers character strings longer than
And 2), the initial dictionary (D1) or the normal of the registered character strings registered in the dictionary (D2), the first dictionary to find the longest registered character string that matches the inputted input character string search Means (1); initial dictionary access means (4) for accessing the initial dictionary (D1); normal dictionary access means (5) for accessing the normal dictionary (D2); and the longest registration A first dictionary registration unit (2) for registering a character string obtained by adding one character to a character string to the ordinary dictionary (D2) by adding an identification number; and an identification number assigned to the longest registered character string. has a coding means (3) for encoding an output code, the initial dictionary (D1) of the first dictionary search means (1)
Search and the first dictionary registration means (2)
A data compression circuit characterized in that registration in the dictionary (D2) is performed simultaneously . 2. The data compression circuit according to claim 1, wherein the initial dictionary (D1) and the normal dictionary (D2) are configured to perform data search and data registration using an external hash. 3. The data compression circuit according to claim 1, wherein the initial dictionary (D1) and the normal dictionary (D2) are configured to perform data search and data registration using a complete hash. 4. A data search and data registration for a character string up to a second character in the initial dictionary (D1), and a data search and data registration for a character string after a third character in the normal dictionary (D2). 2. The data compression circuit according to claim 1, wherein the data compression circuit is configured to perform the following. 5. The first dictionary search unit (1), the first dictionary registration unit (2), and the encoding unit (3)
5. The data compression circuit according to claim 1, wherein the data compression circuit is processed in parallel by a pipeline. 6. A decoding means (11) for decoding an input code inputted in a data restoration circuit for restoring an input code inputted from an information source by decoding.
A second dictionary search means (12) for searching a registered character string registered in the dictionary (D3) for a registered character string that matches the character string restored by the decoding; To the first storage means (13), to store the registered character string, and to the first storage means (13) or the second storage means (14). Output selecting means (15) for selecting any of the stored registered character strings from the stored registered character strings and outputting the selected registered character string as a restored character string; home,
A second dictionary registration unit (16) for searching for a registered character string that matches the restored character string, assigning a new code to the unregistered restored character string, and registering the restored character string in the dictionary (D3); And dictionary access means (17) for accessing the dictionary (D3). 7. The decoding means (11), the second dictionary search means (12), the output selection means (15) and the second dictionary registration means (16) are processed in parallel by a pipeline. 7. The data restoration circuit according to claim 6, wherein: 8. An initial dictionary (D1) for registering a character string in a data compression / decompression device that compresses and decompresses an input character string or an input code input from an information source by encoding or decoding. , containing the string and the initial dictionary registered in the initial dictionary
Normal dictionary that registers character strings longer than
And 2), the initial dictionary (D1) or the normal of the registered character strings registered in the dictionary (D2), the first dictionary to find the longest registered character string that matches the inputted input character string search Means (1), initial dictionary access means (4) for accessing said initial dictionary (D1), and said normal dictionary (D2)
A normal dictionary access means (5) for accessing the first dictionary, and a first dictionary registration means for registering a character string obtained by adding one character to the longest registered character string to the normal dictionary (D2) with an identification number (2) and encoding means (3) for encoding the identification number given to the longest registered character string into an output code, wherein the initial dictionary of the first dictionary search means (1)
(D1) search and the first dictionary registration means (2)
The registration to the normal dictionary (D2) is performed simultaneously.
Decoding means for decoding and the data compression circuit, the inputted input code (11)
And of the registered character strings registered in the dictionary (D3)
A second dictionary search unit (12) for searching for a registered character string that matches the character string restored by the decoding, a first storage unit (13) for storing the registered character string, Storage means (14) for accumulating the data, and the first storage means (13) or the second storage means (14)
Output selection means (15) for selecting one of the stored registered character strings from the registered character strings stored in the dictionary and outputting as a restored character string; and a registered character string registered in the dictionary (D3). A second dictionary registration unit (16) for searching for a registered character string that matches the restored character string, assigning a new code to the unregistered restored character string, and registering the restored character string in the dictionary (D3). And a data decompression circuit comprising: a dictionary access unit (17) for accessing the dictionary (D3).