JP2002084198A - Method and device for decoding variable-length code, and decoding table - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enable fast decoding process, using only a table without bit calculation, such as bit shifting with respect to a variable-length code. SOLUTION: With a table in accordance with a condition S of current where a data D is taken out of a variable-length code stream which is to be decoded for each single byte a record of a number indicated by the data D is referenced, so that a decoding data stream V and a next state S' are acquired. By repeating this, a variable-length code stream is decoded sequentially. If, for example, when a data 243 '11110011' is taken out in a normal state 22, remaining bit stream being '1100', a record number 243 on the table 22 is made reference to. The record decodes '11110011' after '1100' and a decoding data stream '11, 3, 0' indicating values provided by decoding codes 11001, 111, and 001 as well as next state 10 where a remaining bit streams is '1' are stored. Thus, the decoded data stream '11, 3, 0' is acquired, before shifting to the next state 10.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a decoding method and apparatus for a variable-length code for performing high-speed decoding using a table, and a decoding table, and more particularly to a moving image or a still image encoded using a variable-length code. This is suitable for decoding data such as data by software.

[0002]

2. Description of the Related Art With the recent development of network environments such as computers and the Internet, distribution and transmission of digital data obtained by digitizing moving images, still images, and the like have been actively performed. Such digital data is generally compressed to reduce the amount of data. For example, MPEG (Moving Picture) which is one of the moving image compression methods
Coding Experts Group), after orthogonally transforming the frame data of a moving image by discrete cosine transform (DCT),
The numerical value of the T coefficient is encoded with a variable length code. The variable-length code can reduce the entire code amount by assigning a code word having a shorter code length to a more frequent numerical value. For this reason, variable length codes are widely used in data compression.

FIG. 34 shows an example of a variable length code. In FIG. 34, values and codeword lengths are shown in decimal numbers, and codewords are shown in binary numbers. In this variable length code, for example, a value “0” is a code word “00” having a code word length of 3.
The value "10" is encoded into a code word "10001" having a code word length of 5.

FIG. 35 shows data which is variable-length coded by the variable-length code shown in FIG. 34 and data which is obtained by decoding the variable-length code. The variable-length coded data is a 20-bit bit string of “01100011010111100010”, and the data obtained by decoding the variable-length code is five decimal numbers “1,4,2,1,15”. The first three bits “011” correspond to the value “1” in FIG. 34, and the next four bits “0001” correspond to the value “4”. Less than,
The bit strings of 3, 3, and 6 bits sequentially represent the values "2, 1, 1
5 ".

Next, a method of performing decoding using a table will be described. FIG. 36 shows an example of a table for decoding a variable length code. This table has records from 0 to 63. In each record, when the record number is represented by a 6-bit binary number, the value of the variable length codeword included first and the codeword length are stored as the number of bits used. Record numbers are shown in binary notation in parentheses. For example, for record numbers 8 to 15, the first three bits are "00".
In the variable length code shown in FIG. 34, the value corresponds to a code having a value of 0 and a code word length of 3 bits, and stores a value of 0 and the number of used bits of 3. In order to perform decoding using this table, first, a 6-bit bit string is extracted, and a decoded value and the number of bits used are obtained by referring to a record having a number obtained by converting the bit string into a numerical value. Second, the extraction start position is advanced by the number of bits used. This procedure is repeated to decode the variable length code.

When decoding the variable-length coded data shown in FIG. 35, first, the 6-bit "011000" is taken out, and the 24th record, which is a numerical value, is referred to as a decoding target. The 24th record has the value 1 and the number of used bits 3
Is stored. Here, 1 is obtained as the decoded value.
Next, when the extraction start position is advanced by 3 bits and 6 bits are extracted, the decoding target is “000110”. Next, referring to the sixth record in which the decoding target is set to a numerical value, a value 4 and a used bit number 4 are stored. Here, 4 is obtained as the decoded value. When the extraction start position is advanced by 4 bits and 6 bits are extracted, the decoding target is “101011”. If the decoding result is described in the order (<value>, used bit length) referring to the table, the above-mentioned (<1>, 3),
(<4>, 4) followed by (<2>, 3), (<1>,
3), (<15>, 6), and refer to the table a total of five times to obtain "1, 4, 2, 1, 15" as a decoded value. In FIG. 36, the record in which “-” is described in the value and the number of bits used is unnecessary when decoding the variable length code in FIG.

In order to execute the above procedure on a computer, a variable length code to be decoded, a table for decoding, and an area for sequentially storing decoded values are allocated on a memory. In addition, a general C that uses 8 bits or more as a processing unit
In the case of a PU, the unit of access to the memory is 1 byte = 8 bits.

FIG. 37 shows an example of a processing flow of the above-described decoding method using a computer. A conventional decoding method will be described below with reference to FIG. First, in step 3701, a takeout start position is initialized. Here, the address on the memory of the extraction start position is stored in ReadP, and the bit number of the address of ReadP is stored in ReadB. For example, if the retrieval start position is the 86th bit from the beginning of the memory, ReadP = 10, ReadB =
It becomes 6. Next, in step 3702, ReadP and ReadB
6 bits are extracted from the extraction start position indicated by, to obtain a decoding target. Then, in step 3703, by referring to the record of the number indicated by the decoding target, the decoded value and the number of used bits UB are obtained. Thereafter, in step 3704, ReadP and ReadB are updated, and the extraction start position is advanced by UB bits. For example, if UB = 4 is obtained when ReadP = 10 and ReadB = 6, ReadP = 11 and ReadB = 2 after updating.
Then, in step 3705, it is determined whether or not decoding has been completed.
If the decoding is not completed, the process returns to step 3702. If the decoding is completed, the process is terminated.

FIG. 38 shows step 3702 in FIG.
Shows an example of a processing flow for obtaining the decryption target of
The decoding target acquisition procedure will be described below with reference to FIG. First, in step 3801, 2 bytes from the ReadP byte of the memory are loaded into the 16-bit register A.
Here, the decoding target for 6 bits from the extraction start position indicated by ReadP and ReadB is stored in the register A. For example, when ReadB = 6, the content of 16 bits of the register A is “xxxxxXXXXXXxxxxx” when the decoding target is indicated by X and the rest is indicated by x. Next, step 3802
Then, the register A is shifted to the right by (11-ReadB) bits. Here, the decoding target is stored at the position of 6 bits from the least significant bit of the register A. In the above example, ReadB =
In the case of 6, the contents of 16 bits of the register A are “0000
0xxxxxXXXXXX ". Then, in step 3803, the lower 6 bits of the register A are masked. That is, register A is masked with “0000000000111111” and the lower 6
Enable bits. In the above example, register A is "00
00000000XXXXXX ".

FIG. 39 shows a step 3704 in FIG.
FIG. 39 shows an example of a processing flow for updating the takeout start position of the program. The procedure for updating the takeout start position will be described below with reference to FIG. First, in step 3901, Read
The number of used bytes UB obtained in step 3703 is added to B. Next, in Step 3902, it is determined whether or not ReadB is 9 or more. If ReadB ≧ 9, the process proceeds to Step 3903, and if ReadB is 8 or less, the process ends. In step 3903, 1 is added to ReadP, and 8 is subtracted from ReadB.

The variable-length code can be decoded by the processing flow of FIGS. 37 to 39 described above.

As a method for decoding variable-length codes at high speed, there is known a "table storage method and a decoding method and apparatus" described in JP-A-9-261076. In this method, decoding is performed using a table composed of decoded records having a plurality of values, so that the number of table references can be reduced, and variable-length codes can be decoded at high speed.

When this prior art is used for decoding the variable length code shown in FIG. 34, a table as shown in FIG. 40 may be used. This table has records from 0 to 63. Each record is obtained by sequentially decoding not only the value obtained first but also its bit string when the record number is represented by a 6-bit binary number. The value sequence and the number of bits decoded there are stored as the number of bits used.
In FIG. 40, there are a plurality of values obtained by decoding in tables No. 9, 11, 13 and so on, which is different from the table of FIG. For example, the 11th record is 1
Since the binary notation “001011” of 1 can be decoded into the variable length code “011” of value 1 following the variable length code “001” of value 0, the value is “0, 1” and the number of bits used is “6”. Is stored.

The flow of processing for decoding using the table of FIG. 40 is the same as the processing flow described with reference to FIGS. is there. For example, when decoding the example of the variable-length-encoded data shown in FIG.
(<1>, 3), (<4>, 4), (<2, 1>,
6), (<15>, 6), and the decoded value “1, 4, 2, 1, 15” is obtained by referring to the table a total of four times. This is because the number of times the table is referred to is one less than when the table of FIG. 36 is used. By obtaining a plurality of values by one table reference in this way, the number of table references can be reduced and variable-length codes can be decoded at high speed.

[0015]

FIG. 36 described above.
Alternatively, in the conventional variable-length code decoding method using the table of FIG. 40, every time one variable-length code is decoded or each time the table is referred to, a bit string at a predetermined bit position is extracted from the variable-length code string as described above. Had to be taken out.
However, in a computer including a general CPU and a memory, data is accessed in byte units, and a process of arbitrarily extracting a bit string at a predetermined position cannot be directly performed.
It is necessary to perform a bit operation such as a bit shift or a bit mask as exemplified in the processing flow of No. 8. FIG.
As exemplified in the processing flow of No. 9, it is necessary to manage both the byte unit and the bit unit for the extraction start position of the variable-length code string to be decoded.

As described above, in the conventional variable-length code decoding method, a bit operation is required every time to extract a bit string at a predetermined position, or a processing time is required to update a bit string extraction start position. There is a problem that processing cannot be performed at high speed.

The present invention has been made in view of the above circumstances, and there is no need to perform a bit operation such as a bit shift on a variable-length code to be decoded. It is an object of the present invention to provide a decoding method and apparatus for a variable-length code and a decoding table, which can be decoded only by itself and can perform a decoding process at high speed.

[0018]

According to the present invention, first, N bits (N is an integer of 1 or more) from a variable-length code string to be decoded are provided.
And the table corresponding to the current state S is referred to a record corresponding to the extracted bit string B, and the decoded data string V
And a procedure for obtaining a next state S ′ to transition to, and a procedure for repeating these procedures to sequentially obtain decoded data of a variable-length code sequence.

Second, there are a plurality of tables respectively corresponding to the state S at the time of decoding, and in each table, a plurality of records respectively corresponding to the bit string B extracted from the variable length code string in byte units are provided. Includes a decoded data string V indicating data obtained when the bit string B is decoded, and a next state S 'indicating a next transition state after referring to the record. Using a table to extract a bit string B of N bytes (N is an integer equal to or greater than 1) from a variable-length code string to be decoded; Referring to the corresponding record, and obtaining a decoded data string V and the next state S 'from this record. A method of decoding variable length code to obtain the decoded data of the variable length code sequence.

Third, the state S is composed of a plurality of initial states SFi in which the i-th bit (i is an integer of 1 or more) of the extracted bit string B is decoded and the previous bits are not to be decoded. , And a plurality of normal states SNx in which a bit string that cannot be decoded occurs as a result of the previous decoding.

Fourth, there is provided a decoding method for extracting and decoding a bit string for each N bits (N is an integer of 1 or more) from a variable-length code string to be decoded. i = 1 to N) and a plurality of initial states SFi in a state where decoding is performed, and a plurality of normal states in which the remaining bit strings R cannot be decoded in the previous decoding result and the remaining bit strings R can take. SNx and a plurality of tables corresponding to the respective SNx. These tables have a number of records that can be represented by an N-bit binary number. In a corresponding table, a decoded data string V obtained by decoding the bit string b from the i-th bit and a next state S ′ corresponding to the next remaining bit string r are stored. Serial decoded data sequence V in the table corresponding to the normal state SNx obtained by decoding the bit string obtained by coupling the bit string R of the odd bit sequence b
Extracting a N-bit bit string B from a variable-length code string to be decoded using a decoding table configured to store the next state S ′ corresponding to the next remaining bit string r;
Referring to a record corresponding to the bit string B in the table corresponding to the current state S, and obtaining a decoded data string V and a next state S ′ from the record;
A variable length code decoding method is characterized in that decoded data of a variable length code sequence is sequentially obtained by repeating these procedures.

Fifth, there is a decoding method in which a bit string is extracted from the variable-length code string to be decoded every N bytes (N is an integer of 1 or more) and decoded. i = 1 to N × 8), and a plurality of initial states SFi in a state where decoding is performed from a previous decoding result and a plurality of possible bit strings R in a state where there is a surplus bit string R that cannot be decoded as a result of the previous decoding. It has a plurality of tables corresponding to each of the normal states SNx. These tables have a number of records that can be represented by an N × 8-bit binary number. In the table corresponding to the initial state SFi, the decoded data string V obtained by decoding the bit string b from the i-th bit and the next state S ′ corresponding to the next remaining bit string r In a table corresponding to the normal state SNx, a decoded data string V obtained by decoding a bit string obtained by combining the remaining bit string R and the bit string b and a next state corresponding to the next remaining bit string r are stored. A procedure for extracting a bit string B of N bytes in byte units from a variable-length code string to be decoded using a decoding table configured to store S ′ and S ′, and a step of extracting the bit string B in the table corresponding to the current state S Referring to a corresponding record and obtaining a decoded data string V and the next state S 'from this record, and repeating these steps to sequentially obtain decoded data of a variable-length code string. A method for decoding a variable length code is provided.

Sixth, if the variable-length code sequence to be decoded contains both variable-length codes and fixed-length codes, and a fixed-length identification code CS is a code indicating the start of the fixed-length code, In a record corresponding to the bit string b, a record in which the bit string b includes at least a part of the fixed-length identification code CS includes the decoded data string V
The data obtained by decoding the code string up to the fixed-length identification code CS is stored, and the state corresponding to the remaining bits is stored as the next state S ', indicating the detection of the fixed-length identification code. A variable length code decoding method for decoding a fixed length code portion using the record is provided.

Seventh, if the variable-length code sequence to be decoded contains variable-length codes and fixed-length codes, and a fixed-length identification code CS is a code indicating the start of the fixed-length codes, the state S A variable length that includes a plurality of fixed-length decoding states SCy in which decoding of fixed-length codes is performed, and decodes fixed-length codes subsequent to the fixed-length identification code CS with reference to a table corresponding to the fixed-length decoding states SCy. A code decoding method is provided.

Eighth, if the variable-length code sequence to be decoded contains variable-length codes and fixed-length codes, and a fixed-length identification code CS is a code indicating the start of the fixed-length code, the normal state SNx In the corresponding table, in the record indicated by each bit string b, and the record in which the bit string b includes at least a part of the fixed length identification code CS, the code string up to the fixed length identification code CS is described. A decoded data string V obtained by decoding, a numerical value v1 determined by a fixed-length code part after the fixed-length identification code CS in the bit string b, and a next decoding starting from a certain bit of the fixed-length code And a fixed-length decoding state SCy indicating whether there is a bit string in a table corresponding to the fixed-length decoding state SCy indicating decoding from the y-th bit of the fixed-length code. In the record indicated by, a numerical value v2 obtained by decoding the bit string b as the y-th bit of the fixed-length code, a decoded data string V obtained by decoding a code string after the fixed-length code, A decoding table is formed by storing the next state S 'indicating the state of (1), and when decoding the fixed-length code, after obtaining the numerical values v1 and v2, using v1 + v2. A variable length code decoding method for decoding a fixed length code is provided.

Ninth, the variable-length code and the fixed-length code in the variable-length code sequence to be decoded are the run and level of the DCT coefficient in the MPEG image compression method.
Provided is a variable length code decoding method for decoding the run and level of the variable length coded DCT coefficients of the EG.

Tenth, there are a plurality of tables respectively corresponding to the state S at the time of decoding, and in each table, a plurality of records respectively corresponding to the bit string B extracted from the variable length code string in byte units are provided. The decoding of a variable-length code storing at least a decoded data string V indicating data obtained when the bit string B is decoded and a next state S 'indicating the next transition state after referring to the record Provide a table.

Eleventh, a plurality of initial states in which decoding is started from the ith bit (i = 1 to N) of a bit string B of N bits (N is an integer of 1 or more) extracted from the variable-length code string to be decoded. It has a plurality of tables respectively corresponding to a state SFi and a plurality of normal states SNx in which there is a surplus bit string R that cannot be decoded as a result of the previous decoding, and the surplus bit string R can take. The table has a number of records that can be represented by an N-bit binary number. In the record indicated by the N-bit bit string b, in the table corresponding to the initial state SFi, the bit string b is decoded from the i-th bit. And the next state S 'corresponding to the next remaining bit string r are stored in the table corresponding to the normal state SNx. Decoded data string V obtained by decoding the bit stream bound to Rino bit string R and the bit string b
And a next state S ′ corresponding to the next remaining bit string r.

Twelfth, a plurality of N-bit (N is an integer of 1 or more) bits extracted from a variable-length code string to be decoded is decoded from the ith bit (i = 1 to N × 8) of a bit string B. And a plurality of tables respectively corresponding to a plurality of normal states SNx in a state in which a surplus bit string R that cannot be decoded has occurred in the result of the previous decoding and the surplus bit string R can take, These tables have a number of records that can be represented by an N × 8-bit binary number, and a record indicated by an N × 8-bit bit string b includes the bit string b in the table corresponding to the initial state SFi. The decoded data string V obtained by decoding from the i-th bit and the next state S 'corresponding to the next remaining bit string r are stored in a table corresponding to the normal state SNx. A decoded data string V obtained by decoding a bit string obtained by combining the remaining bit string R and the bit string b, and a next state S 'corresponding to the next remaining bit string r. Provide a variable length decoding table.

Thirteenth, a variable-length code storage means for storing a variable-length code string to be decoded, comprising: a decoding table according to any one of the above, and a decoded data storage for storing a data string obtained by decoding. Means, extracting a bit string in a predetermined number of bits or a predetermined byte unit from the variable length code string stored in the variable length code storage means, sequentially performing decoding corresponding to the bit string with reference to the decoding table, And a variable-length code decoding means for storing the decoded data in the decoded data storage means.

In the decoding method and apparatus described above, a predetermined number (N bits) of bit strings, such as byte units, are extracted from the variable-length code string to be decoded and correspond to the extracted bit strings of the table corresponding to the current state S. Decoding can be performed only by referring to the record. Therefore, the decoding process of the variable length code can be performed without performing a bit operation such as a bit shift. In particular, when N is a multiple of 8, decoding can be performed without performing a bit operation while extracting a variable-length code string in byte units. Further, even in a variable length code string in which a variable length code and a fixed length code are mixed, decoding can be performed without referring to a table and without performing a bit operation. As a result, the speed of decoding the variable length code can be increased.

[0032]

Embodiments of the present invention will be described below with reference to the drawings. In the present embodiment, a decoding method and a decoding method of a variable-length code string in which variable-length codes and fixed-length codes are mixed, such as a variable-length code used in MPEG, which is one of the video data compression methods, are described. The device will be described.

[First Embodiment] FIG. 1 is a block diagram showing a configuration of a variable length code decoding apparatus according to an embodiment of the present invention. The variable-length code decoding apparatus according to the present embodiment includes a variable-length code storage unit 101 that stores a code string encoded by a variable-length code, a variable-length code decoding unit 102 that decodes a variable-length code, Decryption table 1 for decryption
03, and a decoded data storage unit 104 for storing a data string obtained by decoding.

The variable length code decoding means 102 is provided for the variable length code storage means 101 realized by a semiconductor memory or the like.
Access can be performed in units of 1 byte = 8 bits. In the present embodiment, the most significant bit of one byte is described as a first bit, and the least significant bit is described as an eighth bit. The variable-length code decoding means 102 includes a general-purpose or special-purpose processor, and operates either by a software program or partially or entirely by hardware having a decoding processing function. included.

In the present embodiment, as a basic operation, the variable-length code decoding means 102 extracts data D for each byte from a variable-length code data sequence (variable-length code sequence) and stores the data D as a decoding table 103. Data D of the table corresponding to the current state S from the plurality of tables
With reference to the record having the number, the decoding result and the next state S 'are obtained. Then, the variable length code is decoded by repeating this process until the end of the data sequence. Here, the state S indicates a state of “remainder bit string” that cannot be decoded in the immediately preceding data D.

FIG. 2 shows an example of a variable length code used in the present embodiment. This variable length code can represent a numerical value from 0 to 255, and EO indicating the end of the variable length code.
It has a B (End Of Block) code "100001". FIG. 2A shows correspondence between a variable length code (code word and code word length) of 6 bits or less and a value. Here, the E
An OB code and an escape code “110001” are included. An escape code (fixed length identification code) indicating an escape sequence indicates the start of a code having a value from 14 to 255 represented by a variable length code of 6 bits or more. This 14 to 25
Codes with values up to 5 have a code length of 14 bits in total consisting of 6 bits of escape code followed by a fixed length portion of 8 bits. FIG. 2B shows the correspondence between an 8-bit code (codeword and codeword length) following an escape code and a value.

In the decoding table of this embodiment, the decoding of the variable length code (value of 0 to 13) having a code word length of 3 to 6 bits in FIG. It can be performed only by accessing in byte units without performing it. As for decoding of a value of 14 or more, an escape code is detected by a decoding process using a decoding table, the next 8 bits (1 byte) are extracted, decoded without using the decoding table, and decoded again using the decoding table. Return to processing.

Next, the state S taken by the variable length code decoding means 102 will be described. The state S has five states roughly classified into an initial state, a normal state, an EOB detection state, an escape detection state (fixed length decoding state), and an error detection state.

The initial state is a state in which decoding is performed from any one of the extracted 1-byte data.
In the initial state, in “decoding from the ith bit”, i = 1
There are eight types of ~ 8. FIG. 3 is an explanatory diagram showing the state of decoding in the initial state. FIG. 3A shows a state of “decoding from the first bit”, in which decoding is performed from the first bit of the extracted one byte. (B) shows a state of “decoding from the second bit”, in which 1 byte 2
This is a case where decoding is performed from the bit number. Similarly, (C) to (H) show states from "decoding from the third bit" to "decoding from the eighth bit". This initial state is a state that can be taken when decoding is started. The bit position where the variable length code to be decoded starts can be determined in the preprocessing. Here, when the head of the variable length code is not always on a byte boundary, the above eight states are required.
This initial state is also obtained after decoding the 8-bit code following the escape code.

The normal state is a state in which a “remainder bit string” is being generated during decoding, and has a state of the number of possible combinations of the “remainder bit string”. The “remainder bit sequence” will be described with reference to FIG.

In FIG. 4A, when the N-th byte is decoded, the code word and the code word can be decoded.
The codeword “1101” indicating the next value “7” cannot be decoded until the (N + 1) th byte is processed. At the time of processing the N-th byte, “110” from the 6th bit to the 8th bit of the N-th byte becomes the “remaining bit string”. When processing the (N + 1) th byte, in the normal state, “the remaining bit string: 11
The processing corresponding to the state “0” is performed. In (B), when the N-th byte is processed, “11” becomes the “remainder bit string”, and when the (N + 1) -th byte is processed,
In the normal state, a process corresponding to the state of “remaining bit string: 11” is performed. In (C), when the N-th byte is processed, “1” becomes “remainder bit string”, and N + 1
When processing the byte, a process corresponding to the state of “remaining bit string: 1” is performed in the normal state. In (D), when the N-th byte is processed, the code word and the code word can be decoded exactly, and there is no “remainder bit sequence” in the N + 1-th byte, and the decoding is started from the first bit. "Initial state: decoding from the first bit". In (E), when the N-th byte is processed, 8
The codeword, codeword, codeword `` 11
01 "can be decoded, and since there is no" remaining bit string "in the (N + 1) th byte and decoding is started from the first bit," initial state: decoding from the first bit ".

As described above, in (A) to (C) of FIG. 4, the normal state is obtained after processing the N-th byte.
"Remainder bit string: 110""Remainder bit string: 1"
1 "and" Remainder bit sequence: 1. "

As described above, for a code word having a code word length of M bits, M-1 kinds of "remainder bit strings" are generated. FIG. 5 shows a combination of “remainder bit strings” corresponding to the variable-length code in FIG.

FIG. 5A enumerates the "remainder bit string" for each code word. Since the code word “001” having a value of 0 has a code word length of 3, the “remainder bit sequence” is “0” having a 1-bit length and “00” having a 2-bit length. Since the code word “0101” having the value 5 has a code word length of 4,
The “remaining bit string” is 1-bit length “0”, 2-bit length “01”, and 3-bit length “010”. In (A), there is an overlapped “remaining bit string”. For example, the 1-bit length “remainder bit string” for the value 0 and the value 1 is “0”, which is an overlap. FIG. 5 (B)
Indicates the "remaining bit string" excluding the overlapping ones. For the 16 code words, the 1-bit length “remainder bit sequence” is two of “0” and “1”, and the 2-bit length “remainder bit sequence” is “00”, “01”, “1”. Ten",
There are four "11". Each of the 3-bit length, 4-bit length, and 5-bit length has four “remaining bit strings”, and the total number of combinations of “remainder bit strings” is 18.

The EOB detection state is a state indicating the detection of an EOB code.
There are eight states (“EOB detection: remaining bit number i” i = 0 to 7) depending on how many bits of data remain after B.

FIG. 6 is an explanatory diagram showing eight types of patterns in the EOB detection state. In this case, after processing the (N + 1) th byte, the state becomes the EOB detection state. FIG. 6A shows a state in which a “remainder bit string” is generated by a codeword after processing the Nth byte. At the time when the (N + 1) th byte is processed, the codeword can be decoded and EOB is reduced to 3 bits.
This is a case where detection is performed at the 8th to 8th bits. Therefore, the state "EOB" in which the EOB is detected and the number of remaining bits becomes 0 bits
Detection: Remaining bit number 0 ”. (B) shows a state in which a “remainder bit string” is generated by a codeword after processing the Nth byte. When the (N + 1) th byte is processed, the codeword can be decoded. Is detected. Therefore, the number of remaining bits after EOB is 1
The state becomes the bit "EOB detection: remaining bit number 1". (C) shows that after processing the Nth byte, there is no “remaining bit string” and the state becomes “initial state: decoding from the first bit”, and when the (N + 1) th byte is processed, the EOB is 1 to 6 bits. This is the case when the eye is detected. Therefore, EOB
Is detected and the remaining number of bits becomes 2 bits “EO
B detection: remaining bit number 2 ”. (D) shows that at the time when the first one bit of the EOB becomes “remaining bit string” and becomes “normal state: surplus bit string: 1” after processing the N-th byte, the N + 1-th byte is processed. , EOB are detected. Therefore, EOB
Is detected and the remaining number of bits becomes 3 bits "EO
B detection: remaining bit number 3 ”. (E)-(H)
After processing the N-th byte in the same manner as in (D), the head portion of the EOB becomes a “remaining bit string”, and the remaining portion of the EOB is detected when the (N + 1) -th byte is processed. Therefore, EOB is detected and the number of remaining bits is 4 to 7
The state of the bit becomes “EOB detection: remaining bit number 4 to 7”.

The escape detection state indicates the detection of an escape code.
Eight states (“escape detection: number of remaining bits i” i =
0-7).

FIG. 7 is an explanatory diagram showing eight types of patterns in the escape detection state. In this case, after processing the (N + 1) th byte, an escape detection state is set. The escape code is followed by an 8-bit fixed length part. FIG. 7A shows a state in which a “remainder bit string” is generated by a codeword after processing the Nth byte. At the time when the (N + 1) th byte is processed, the codeword can be decoded.
This is a case where detection is performed at the 8th to 8th bits. Therefore, the state where "escape" is detected and the number of remaining bits is 0 "escape
Detection: Remaining bit number 0 ”. (B) shows a state in which a “remainder bit string” is generated by the code word after processing the Nth byte. At the time when the (N + 1) th byte is processed, the codeword can be decoded, and escape is set to the 2nd to 7th bits. Is detected. Therefore, the number of remaining bits after escape is 1
The state becomes bit "escape detected: remaining bit number 1". (C) shows that after processing the N-th byte, there is no “remaining bit string” and the state is “initial state: decoding from the first bit”. When the (N + 1) -th byte is processed, escape is set to 1 to 6 bits. This is the case when the eye is detected. So escape
Is detected and the number of remaining bits becomes 2 bits "esca
pe detection: remaining bit number 2 ". (D) shows that, after processing the N-th byte, the first bit of escape becomes the “remaining bit string” and becomes “normal state: surplus bit string: 1”. , Escape are detected. So escape
Is detected and the remaining number of bits becomes 3 bits "esca
pe detection: remaining bit number 3 ". (E)-(H)
After processing the N-th byte in the same manner as in (D), the first part of the escape becomes the “remaining bit string”, and the remaining part of the escape is detected when the (N + 1) -th byte is processed. Therefore, escape is detected and the number of remaining bits is 4 to 7
The bit state becomes "escape detection: remaining bit number 4 to 7".

The error detection state is a state in which a code which cannot be assumed as a variable length code assumed in the fetched data is detected. For example, after a codeword, "111xxxxxx
When a bit string such as “x” continues, there is no variable length code corresponding to such a bit string, and an error detection state is set.

FIG. 8 shows the state numbers assigned to the states that can be taken by the variable length code decoding means 102, the meanings indicated by the states, and the numbers of the tables to be referred to corresponding to each state. 8A shows an initial state and a normal state, and FIG. 8B shows an EOB detection state, an escape detection state, and an error detection state. For example, the state of the state number 16 indicates that the “remainder bit string” is “010”,
The decoding table corresponding to that state is the 16th table. In the states of state numbers 1 to 26, decoding tables 1 to 26 are used, respectively.
The states of state numbers 27 to 43 indicate that the decoding table is not used. That is, the decoding table 103 used in the present embodiment includes tables 1 to 26 corresponding to eight types of initial states and eighteen types of normal states.

FIG. 9 shows an example of a processing flow for listing "remaining bit strings" that can be taken in a normal state. The procedure for listing remaining bit strings will be described below with reference to FIG.

First, in step 901, it is determined whether all codewords have been extracted. If all codewords have been extracted, the flow advances to step 908. If all codewords have not been extracted yet, the processing after step 902 is performed. Initially, the process proceeds to step 902.
In step 902, the next codeword C is extracted. Here, 16 code words indicating values 0 to 13 and EOB and escape in FIG. 2A are targeted, and the processing is performed on all 16 code words by this flow.

Next, in step 903, the codeword length of the extracted codeword C is substituted for the variable LEN. Subsequently, at step 904, 1 is substituted for a variable i. And step 90
In step 5, it is determined whether or not the variable i is equal to the variable LEN. Variable i and variable LEN
If are not equal, go to step 906. In step 906, the bit string s of the first i bits from the codeword C
And adds it to the remaining bit string set RS. Here, if there is already a bit string s in the remaining bit string set RS, it is not added. In step 907, the variable i is set to 1
And returns to step 905, where the variable i is set to the variable LEN
Steps 906 and 907 are repeated until is equal to

When the process returns to step 901 and all codewords have been extracted, in step 908,
The state number is made to correspond to the bit string of each element of the remaining bit string set RS, and the processing is ended.

A specific example in which the variable length code of FIG. 2 is applied to the processing flow of FIG. 9 will be described. First, the code word “001” in FIG. 2A is extracted, and LEN is set to 1 for this code word.
The bit strings “0” and “00” are extracted from to and 2, and are added to the remaining bit string set RS. Here, the content of the remainder bit string set RS is {0, 00}. Next, code word "010"
And the bit strings “0” and “01” are extracted for this codeword while LEN is between 1 and 2, and the remainder is added to the bit string set RS. At this time, since “0” is already included, only “01” is to add. Here, the content of the remainder bit string set RS is {0, 00, 01}. FIG. 2
When processing is performed on all 16 code words of (A),
The contents of the remainder bit string set RS are {0, 1, 00, 01, 10,
11,000,010,100,110,0000,0100,1000,1100,00
There are 18 types of 000, 01000, 10000, 11000｝. In addition,
The arrangement of the remaining bit string elements is shown in a sorted order, not in the order of addition. In step 908 of FIG. 9, the state number is made to correspond to the bit string of each element.
No. to No. 26.

In this way, the "remaining bit strings" that can be taken in the normal state can be enumerated and correspond to states 9 to 26. As a result, as shown in FIG. 8, the correspondence between the state number and the meaning indicated by the state can be obtained.

FIG. 10 shows an example of a table corresponding to the normal state as an example of the decoding table 103. "Remaining bit string" is "1100" in normal state
No. 22 is a table corresponding to the state 22 which is the state described above. This table has records 0 to 255 as shown in FIG. In parentheses of the record numbers, the record numbers are represented in binary numbers. Each record stores a “decoded data string” and a “next state”. For example, the 243rd record is obtained when the “remainder bit string” is “1100” in the Nth byte, and the bit string “11110011” indicating the record number 243 by an 8-bit binary number appears in the N + 1th byte. Yes, it is.

Regarding the details of the record of No. 243,
This will be described with reference to the right side of the table shown in FIG. In this case, when the N-th byte is extracted, the code word can be decoded, and the state becomes “normal state: surplus bit string: 1100”. Then, when extracting and decoding the (N + 1) th byte, the state 22 of “normal state: surplus bit string: 1100”
Is referred to the table 22 corresponding to. In this case, since the data of the (N + 1) th byte is 243 (“11110011” in binary), the record with the record number 243 is referred to. Here, when decoding is performed on “110011110011” obtained by concatenating the remaining bit string “1100” and the N + 1th byte data “11110011”, the variable-length codewords “11001”, “111”,
Since “100” is obtained, the data to be decoded is “11,
3,0 ". Finally, the bit string “1” remains.
Therefore, the record of No. 243 stores “11, 3, 0” in the “decoded data string”, and stores the state 10 indicating “normal state: surplus bit string: 1” in the “next state”.

In this table 22, for the record of No. 0, "-" indicating that there is no data to be decoded is stored in the decoded data string, and the next state is the state 43 of the error detection state. Is stored. In other words, "11
A bit string "11000000000" that connects "00" and "00000000"
With reference to FIG. 2, "0" indicates that there is no corresponding code word that can be decoded, and the error cannot be detected because the code word cannot be decoded.

In the table corresponding to the initial state, the bit strings to be decoded are extracted according to each of the eight states, and the "decoded data string" and the "next state" Is determined. A decoding table including tables corresponding to each of the initial state and the normal state is created in advance so as to correspond to the content of the variable-length code.

Here, a detailed procedure for creating a decoding table will be described with reference to FIG. FIG. 11 shows an example of a processing flow for creating a decoding table. In this processing flow, when a state S and a record number RN are given, a process of determining a “decoded data string” and a “next state” is performed. In this case, the state S is a state 1 to a state 26 representing each state in the initial state and the normal state.

First, in step 1101, the record number R
A bit string in which N is represented by an 8-bit binary number is substituted into BIT. Then, in step 1102, it is determined whether or not the state S is the initial state.
Otherwise, that is, in the case of the normal state, the process proceeds to step 1104. In step 1103, since the state S corresponds to decoding from the N-th bit in the initial state, C
The bit sequence from the Nth bit to the end of the BIT is substituted for mode, and the process proceeds to step 1106. On the other hand, step 11
In 04, since the state S is the normal state, a surplus bit string corresponding to the state S is obtained and substituted into R. Then, in step 1105, a bit string obtained by connecting BIT to the remaining bit string R is substituted for Code, and the flow advances to step 1106.

Next, in step 1106, it is determined whether there is a code word C that matches the code or a code word C that matches the head of the code. Here, all 16 code words in FIG. 2A are examined. For example, if the code word length is 3, the first 3
Checks for a match. This step 11
In step 06, if a codeword C that matches the head of the code is detected, the process proceeds to step 1107; otherwise, the process proceeds to step 1113.

In step 1107, step 1106
It is determined whether or not the code word C that matches with is escape or EOB. If the codeword C is escape or EOB, the process proceeds to step 1112; otherwise, the process proceeds to step 1108 and subsequent steps. In step 1108, the codeword C
Therefore, the data corresponding to the codeword C is added to the decoded data string of the record. Then, in step 1109, the part of the decoded codeword C is removed from the beginning of the Code, and Code
Substitute for e.

Next, in step 1110, it is determined whether or not the length of the code is 0. If the length of the code is not 0, the process returns to step 1106 and the same processing is repeated. C
When the length of the mode becomes 0, the process proceeds to step 1111. When proceeding to Step 1111, the record RN
Since there is no "remaining bit string" after decoding, the state of "initial state: decoding from the first bit" (that is, state 1) is stored in the "next state" of the record, and the process ends.

When the process proceeds to step 1112,
This is a state in which an escape code or an EOB code has been detected, the escape detection state or the EOB detection state is stored in the next state, and the process ends. Here, when an escape code is detected, states 35 to 42 shown in FIG. 8 are stored in the “next state” of the record according to the number of bits obtained by removing the escape code from Code. When an EOB code is detected, the state 27 to the state 34 shown in FIG. 8 are stored in the “next state” of the record according to the number of bits obtained by removing the EOB code from the Code.

In step 1113, it is determined whether Code matches the head of the code word C. Here, all 16 code words in FIG. 2A are examined. If it is detected in this step 1113 that the Code matches the head of the codeword C, the flow proceeds to step 1114; otherwise, the flow proceeds to step 1115. If the process proceeds to step 1114, the Code is in a state of remaining bits, and the state in which the remaining bits are Code in the normal state is stored in the “next state” of the record, and the process ends. Here, the correspondence between the state number shown in FIG. 8A and the remaining bit string obtained by the processing flow of FIG. 9 is used.

On the other hand, when the process proceeds to step 1115,
Since Code cannot be decoded even if any subsequent bit string continues, an error is detected. Therefore, the error detection state (state 43) is stored in the “next state” of the record, and the process ends.

FIG. 12 shows an example of the decoding table. A specific example of the procedure for creating the decoding table will be described below with reference to FIGS. 12 and 11. The format of the table is the same as in FIG. Table 1 shown in FIG.
Is a table used in state 1, that is, "initial state: decoding from bit 1".

In this table 1, record number 0
The record (record 0) is used when the code "00000000" appears. Since such a code does not have a code corresponding to the variable length code in FIG. 2 and is not a correct code, the "next state""43" is stored (error detection state). The procedure for obtaining this record will be described with reference to FIG.

In this case, first, "00000000" is assigned to BIT (step 1101). Since state 1 is the initial state, the flow branches from step 1102 to step 1103 of Yes. Then, since the state S is decoding from the first bit, when the first to last bits of the BIT are substituted for Code, the Code becomes "00000000" (step 11).
03). Next, since the code word C cannot be detected at the head of the code, steps 1106 to 11
Branch to 13. Further, since the head of the code word C cannot be detected in the code, the process branches from step 1113 to step 1115 of No. Then, the error detection state is stored in the “next state” (step 1115), and the process ends.

The record No. 44 in Table 1 (record 44) is used when the code “00101100” appears. In this case, "001" indicating the value 0
And the code “011” indicating the value 1 is decoded, and the bit string “00” remains, so the value of the “decoded data string” is “0,
In “1” and “next state”, 11 (“remaining bit string in normal state: 00”) is stored. The procedure for obtaining this record will be described with reference to FIG.

In this case, first, “00101100” is assigned to BIT (step 1101). Since state 1 is an initial state, the process branches from step 1102 to step 1103 of Yes. Then, since the state S is decoding from the first bit, when the first to last bits of the BIT are substituted for Code, the Code becomes “00101100” (step 11).
03). Next, since the code word “001” can be detected at the head of the code, the process branches from step 1106 to step 1107 of Yes. Subsequently, the detected codeword is es
Since it is not cape or EOB, No from step 1107
Branch to step 1108. And the codeword "001"
Is added to the decoded data sequence (step 1108), and the codeword “00” is read from the beginning of the code.
Excluding “1”, Code becomes “01100” (step 1).
109).

Next, since the length of the code is not 0, the process returns from step 1110 to step 1106 of No. Then, since the code word “011” can be detected at the beginning of the code, the steps 1106 to 1107 of Yes
Branch to Next, the detected code word is escape or EO
Since it is not B, the process branches from step 1107 to step 1108 of No. Next, data “1” corresponding to the code word “011” is added to the decoded data sequence (step 1).
108), excluding the code word “011” from the beginning of the code, the code becomes “00” (step 1109). Then, since the length of the code is 2, step 1110
The process returns to step 1106 of No.

At this time, since a codeword cannot be detected at the head of the code, the process branches from step 1106 to step 1113 of No. In this case, since the Code is “00” and matches the first two bits of the code words “001”, “0001”, “00001”, etc.
It branches to step 1114 of es. Subsequently, the state where the remaining bit string becomes Code “00”, that is, state 11
Is stored in the “next state” (step 1114), and the process ends.

In the record 45 in the table 1, the "decoded data string" has the same value "0, 1" as the record 44. In this case, since the remaining bit string is "01", the "next state" 12 is stored. Also, record 92
Is a case where the code “0101” indicating the value 5 is decoded and the bit string “1100” remains, so “5” is stored in the value of the “decoded data string” and 22 is stored in the “next state”. .

The record 93 in the table 1
Is a case where the code “0101” indicating the value 5 and the code “1101” indicating the value 7 are decoded, and the remaining bit string is lost.
The state becomes the same as the initial state of decoding from the bit number. Therefore, “5, 7” is stored in the value of the “decoded data string”, and 1 (“initial state: decoding from the first bit”) is stored in the “next state”. The procedure for obtaining this record will be described with reference to FIG.

In this case, first, "01011101" is assigned to BIT (step 1101). Since state 1 is the initial state, the flow branches from step 1102 to step 1103 of Yes. Next, since the state S is decoding from the first bit, when the first bit to the last bit of the BIT are substituted into Code, the Code becomes “01011101” (step 11).
03). Then, the code word “0101” is added at the beginning of the code.
Can be detected, and the process branches from step 1106 to step 1107 of Yes. Subsequently, the detected codeword is
Since it is not escape or EOB, N from step 1107
Branch to step 1108 of o. Next, the codeword "010
Data “5” having a value corresponding to “1” is added to the decoded data string (step 1108), and the code becomes “1101” by removing the codeword “0101” from the beginning of the code (step 1109).

Next, since the length of the code is not 0, the process returns from step 1110 to step 1106 of No. Then, since the code word “1101” can be detected at the head of the code, the steps 1106 to 110
Branch to 7. Subsequently, the detected codeword is escape or E
Since it is not OB, the process branches from step 1107 to step 1108 of No. Then, when the value data “7” corresponding to the code word “1101” is added to the decoded data sequence, the decoded data sequence becomes “5, 7” (step 1108).
At this time, if the codeword “1101” is removed from the beginning of the code, the code becomes “” (no content) (step 11).
09). Next, since the length of the code is 0, the process branches from step 1110 to step 1111 of Yes.
Then, the state 1 indicating "initial state: decoding from the first bit" is stored in the "next state" (step 1111), and the process ends.

The record 255 in the table 1
Is a case where two codes “111” indicating the value 3 are decoded and the remaining bit string becomes “11”. The values of the “decoded data string” are “3, 3”, Is stored in.

Table 5 shown in FIG. 12B is a table used in state 5, that is, "initial state: decoding from bit 5". In this table 5, the record of record number 0 (record 0) is "00
This code is used when the code “00000” appears. However, the code to be decoded is “0000” after the fifth bit. Since the corresponding code is not determined here, the value of the “decoded data string” is Is also not stored. Since the remaining bit string is “0000”, the “next state” is 19. In the record 44 and the record 92, the code after the fifth bit is also “1100” and one or more values cannot be decoded.
Nothing is stored in the value of the "decoded data string". The “next state” is 22 since the remaining bit string is “1100”. In the record 45 and the record 93, the code after the fifth bit is also “1101”, and the value 7 is decoded and there is no excess bit string. Therefore, "decrypted data string"
Is stored in the value of “7”, and 1 in the “next state”. In the record 255, the code after the fifth bit is “1111”, and the value 3 is decoded and the bit string “1” is left. Therefore, the value of the “decoded data string” is “3”, and the value of the “next state” is 10 is stored.

Table 10 shown in FIG. 12 (C) shows that in state 10, that is, in the normal state, the "remaining bit string" is "1".
It is a table used in the state. This table 1
At 0, record 8 is a record used when the code "00001000" appears. In this case, "100001000" with the surplus bit string "1" prefixed is to be decoded. Here, “100001” indicating the EOB can be decoded, and the remaining bit string becomes 3 bits of “000”. Therefore, the value of the “decoded data string” is not stored, and the “next state” is “EOB detection: remaining. State 30 indicating "bit number 3" is reached. The procedure for obtaining this record will be described with reference to FIG.

In this case, first, "00001000" is assigned to BIT (step 1101). Since the state S is not the initial state, the process branches from step 1102 to step 1104 of No. Next, since the “remainder bit string” corresponding to the state 10 is “1”, “1” is substituted for R (step 1104). Then, R and BIT are linked to “100001
"000" is substituted for Code (step 1105). Next, since the EOB code “100001” can be detected at the beginning of the code, the steps 1106 to 1
Branch to 107. Then, the detected code word is EOB
Therefore, Step 1107 to Step 1 of Yes
Branch to 112. At this time, since the remaining bit string excluding EOB from the Code is 3 bits, a state 30 indicating that the number of remaining bits is 3 bits in the EOB detection state is stored in the “next state” (step 1112), The process ends. The record 9 is the same as the record 8, and the value of the “decoded data string” is “−”, and the “next state”
Stores a state 30 indicating "EOB detection: remaining bit number 3".

The record 44 in the table 10
Is a record used when the code “00101100” appears. In this case, “100101100” prefixed with the surplus bit string “1” is to be decoded. Here, “1001” indicating the value 6 and “011” indicating the value 1 are decoded, and the bit string “0” is decoded.
Since “0” remains, the value of “decoded data string” is “6,
"1" and "next state" include "bit string remaining in normal state:
00 is stored. Also, record 14
Reference numeral 2 denotes a record used when the code “10001110” appears. In this case, “110001110” with the surplus bit string “1” prefixed is to be decoded. Where first esc
"110001" indicating ape can be decoded, and the remaining bit string is "1
Since it is 3 bits of "10", the value of the "decoded data string" is not stored, and the "next state" stores the state 38 indicating "escape detection: remaining bit number 3".

FIG. 12D shows a table 6
Record 44 of “(initial state: decoding from 6th bit)”
(E) shows the record 195 of the table 17 (“the remaining bit string in the normal state: 100”). In these tables and the table 22 in FIG. 10, each record is obtained based on the processing flow in FIG. 11 in the same manner as described above. By the explanation of the record acquisition procedure of the table mentioned above,
A specific example of all the operations in the processing flow of FIG. 11 has been described.

Next, the operation of the variable length code decoding means 102 in the present embodiment will be described. The variable length code decoding means 102, when given the position of the bit to start decoding by the address (the number of bytes from the head of the memory) and the number of the bit in the address, (1) storing the variable length code By designating a read address to the means 101, the code of one byte of the address is extracted,
(2) decoding the variable length code with reference to the decoding table 103; and (3) storing the decoded data in the decoded data storage unit 1.
The decoding of the variable-length code is performed by repeating the process of outputting the variable-length code.

FIG. 13 shows an example of the processing flow of the operation of the variable length code decoding means 102. The decoding procedure of this embodiment will be described below with reference to FIG.

First, in step 1301, the address of the bit position to start decoding (the decoding start address on the memory) is stored in the variable ReadP indicating the read address. Substitute the state number corresponding to the position. As described above, 1 to 8 are assigned to the decoding start bit positions 1 to 8, respectively, as the initial state numbers.

Next, in step 1302, an address ReadP is instructed to the variable-length code storage means 101, and 1-byte data of the corresponding address is obtained from the codes stored in the variable-length code storage means 101. Assign to variable D. Then, in step 1303, by referring to the D-th record in the table corresponding to the state S in the decoding table 103, the column of the decoded data stored in the “decoded data column” and the “next state” Snext are obtained. . If there is a decoded data string, the data is stored in the decoded data storage unit 104.
Output to Next, in steps 1304 to 1306, Snext
Is determined.

In step 1304, it is determined whether or not Snext is EOB detection. If Snext is 27 to 34, EOB detection is performed, and the process ends. Step 13
In 05, it is determined whether or not Snext is an escape detection,
If the value of Next is 35 to 42, escape detection is performed, and the process proceeds to step 1308. In step 1306, the Snext
Is determined to be an error detection. If Snext is 43, an error has been detected and the process is terminated.

Next, at step 1307, 1 is added to ReadP, and the read position of the data to be decoded is advanced by one byte. Also, Snext is assigned to the state S, and the state is shifted.
Thereafter, the flow returns to step 1302 to repeat the same processing.

Steps 1308 to 1309 are processing in the escape detection state. In step 1308, esca
The fixed length part following the pe code is decoded, and the decoded data is output to the decoded data storage unit 104. In this case, the fixed length is 8
Since it is a bit, the address of ReadP + 1 is instructed to the variable length code storage means 101, and 1-byte data of the corresponding address is obtained. And the esca of the variable D that we have already obtained
Eight bits are extracted from the part after the pe code and the newly obtained 1-byte data, and are decoded according to the table in FIG.

Next, in step 1309, the state S and the read address ReadP are updated so that the state becomes the corresponding state after decoding the fixed length part. Here, step 130
If there are remaining bits after the fixed-length part of the newly obtained 1-byte data in step 8, ReadP + 1 is substituted for ReadP. : Decoding from the 9th-nth bit "
Set 8. On the other hand, if there are no remaining bits after the fixed length part, ReadP + 2 is substituted for ReadP, and
State 1 of "initial state: decoding from first bit" is set.

Next, a specific example of the operation of the variable length code decoding means 102 will be described in detail with reference to FIG. 13 taking the case of decoding the variable length code shown in FIG. 14 as an example. FIG. 14 shows an example of the variable-length code stored in the variable-length code storage means 101. The code content of each byte is shown in a rectangular frame, and the address is above the frame, and the address is decoded below the frame. Indicates the value. Regarding the contents of each byte, the upper part in the square frame shows a numerical value in decimal number, and the lower part shows a bit string in binary number. Here, the bit position to start decoding is from the fifth bit of the address 1001. Further, all tables and records referred to in the following description are shown in FIG.
And FIG.

First, the start address 1001 is substituted for ReadP,
In the state S, 5 is substituted because the decoding is from the fifth bit in the initial state (step 1301). Then, the address of ReadP, that is, 1-byte data “92 (01011100)” of the address 1001 is obtained from the variable-length code storage unit 101 and substituted into D (step 1302). Next, S is 5,
Since D is 92, the record 92 of the table 5 in the decoding table 103 is referred to (step 130).
3). In this case, there is no value to be decoded, and the state is set as the “next state” in state 22 (“bit string remaining in normal state: 1100”).
Is obtained, 22 is substituted for Snext. Where S
Since next = 22, the determination results in steps 1304 to 1306 are all No, and the process proceeds to step 1307. Then, 1 is added to ReadP to obtain 1002, and Snext is substituted for S to obtain 22 (step 130).
7) Return to step 1302.

In step 1302 for the second time, 1-byte data “243
(11110011) ". Then, the record 243 of the table 22 in the decoding table 103 is referred to (step 1303). In this case, a value “11,3,0” is obtained as a decoded data string, and a state 10 (“remaining bit string: 1 in normal state”) is obtained as a “next state”. The value “11,3,0” of the decoded data string is stored in the decoded data storage unit 1.
04 and substitute 10 for Snext. Where Sne
Since xt = 10, the determination results in steps 1304 to 1306 are all No, and the process proceeds to step 1307.
Then, 1 is added to ReadP to obtain 1003, and S
Is set to 10 (step 1307), and the process returns to step 1302.

In the third step 1302, as the variable D, 1-byte data “142” of the address of ReadP = 1003
(10001110) ". Then, reference is made to the record 142 of the table 10 in the decoding table 103 (step 1303). In this case, there is no value to be decoded, and a state 38 (“escape detection: remaining number of bits 3”) is obtained as the “next state”, so 38 is substituted for Snext. Here, since Snext = 38, the determination result of step 1305 is Yes, it is determined that the escape is being detected, and the process proceeds to step 1308. Then, the fixed length portion after the escape of the extracted data is decoded (step 13).
08). At this time, first, the address 1004 of the next ReadP + 1
Is obtained, then the remaining three-bit bit string “110” after the escape code in data D = 142 and the first five bits “00101” of the next data “44” are obtained. To obtain an 8-bit bit string "11000101". Then, referring to the table in FIG. 2B, since the 8-bit data corresponds to the value “197”, the value “197” is output to the decoded data storage unit 104 as the decoded data. Thereafter, the state S indicating the next state and the read address ReadP are updated (step 1).
309). In this case, the decoding starts next from the sixth bit of the address 1004. Therefore, 1004 is assigned to ReadP, and the state 6 indicating “initial state: decoding from the sixth bit” is assigned to S.

In the fourth step 1302, 1-byte data “44 (0
0101100) ”. Then, the record 44 of the table 6 in the decoding table 103 is referred to (step 1).
303). In this case, there is no value to be decoded, and the next state is set to state 17 (“the remaining bit string: 10 in the normal state”).
0 ”) is obtained, so 17 is substituted for Snext. Here, since Snext = 17, steps 1304 to 13
All the judgment results of No. 06 are No, and step 1307
Proceed to. Then, 1 is added to ReadP to obtain 1005, and Snext is substituted for S to obtain 17 (step 130).
7) Return to step 1302.

In the fifth step 1302, as the variable D, 1-byte data “195” at the address of ReadP = 1005
(11000011) ". Then, the record 195 of the table 17 in the decoding table 103 is referred to (step 1303). In this case, the value “6” is set as the decoded data string, and the state 28 (“EOB detection:
The number of remaining bits: 1) is obtained. The value “6” of the decoded data string is output to the decoded data storage unit 104,
Is assigned to 28. At this time, since Snext = 28, the result of the determination in step 1304 is Yes, and the process ends when the end of the variable length code is detected.

By the above operation, "11, 3, 0, 197, 6" is output to the decoded data storage means 104 as a value obtained by decoding the addresses 1001 to 1005, and the EOB code is detected at the address 1005. A state in which the number of surplus bits is 1 can be obtained.

As described above, when decoding a variable length code, data can be handled in units of bytes, and decoding can be performed only by referring to a table without performing a bit operation. It is also possible to detect a fixed-length code accompanied by an escape code. After decoding the fixed-length code, decoding of the variable-length code can be started again in the same procedure. As a result, the decoding process of the variable length code can be performed easily and at high speed, so that the processing time can be reduced and the load on the processor that performs the decoding process can be reduced.

In the first embodiment described above, the escape
An 8-bit fixed-length code is defined after the code.
In the case where all the codes as shown in FIG. 4 are variable length codes, the decoding processing of the fixed length part (steps 1305, 1308, 1309) in the processing flow of FIG. 13 can be omitted.

Further, in the present embodiment, a fixed length part in data is taken out in order to decode a fixed length code starting with an escape code, and a bit operation is required for this processing. To avoid this, es
A decoding table is created by treating a cape code and a code consisting of a total of 14 bits of a fixed-length portion of 8 bits thereafter in the same manner as a normal variable-length code, and sets all codewords including the values 14 to 255 as bits. The decoding may be performed without performing the shift. With this configuration, the processing time can be further reduced. In this case, the number of combinations of “remaining bit strings” generated when decoding is performed by the variable-length code decoding unit 102 increases, so that more states and corresponding tables are required.

In this embodiment, an example of decoding a variable length code as shown in FIG. 2 has been described. However, the present invention can be applied to any variable length code, The decoding table and the decoding means can be configured according to the contents of. For example, in a configuration for decoding MPEG, a set of data including MPEG runs and signed levels is enumerated as what is stored in the decoded data string of each table in the decoding table. The bit string of the EOB code is “10”, indicating the beginning of the fixed-length code.
The bit sequence of the escape code is “000001”, and the fixed-length code portion has a signed level of 8 bits or 16 bits following a 6-bit run, so that these may be decoded.

As described above, in the present embodiment, the decoding table has a plurality of tables corresponding to each state at the time of decoding processing, and the record RN of each table decodes the bit string indicated by RN in that state. A variable-length code is decoded by referring to the decoding table while taking out data for each byte, using the data storing the "decoded data string" and the "next state", respectively. With such a decoding method, a variable-length code can be decoded without performing a bit operation.
The decoding processing time can be reduced, and the practical effect is great.

[Second Embodiment] In the second embodiment, the MPE
FIG. 3 shows an example of a configuration and a method for decoding a variable length code of a run and a level of a DCT coefficient in G1. The configuration of the variable-length code decoding device is the same as that of the first embodiment shown in FIG. The variable-length code decoding unit 102 is provided with a variable-length code storage unit 101 implemented by a semiconductor memory or the like.
Access can be performed in units of bytes = 8 bits.
In the present embodiment, the most significant bit of one byte is described as a first bit, and the least significant bit is described as an eighth bit.

FIGS. 15A and 15B, FIGS.
(B) shows a variable length code of DCT coefficient run and level in MPEG1 as an example of a variable length code to be decoded.
In FIG. 15, the code “10” is an EOB (End Of Block) code indicating the end of the variable length code. The code `` 00000
"1" indicates run with fixed length code, esc indicating start of level
ape code (fixed-length identification code). Other signs are
Corresponding to one set of run and level, if the last bit s of the code is 0, the level is a positive value, and if the bit s is 1, the level is a negative value. For example, the code "01000" is run =
0, level = 2, and the code “01001” is run = 0, level = −2. The code “1s” is used only as the first code of the variable length code of the DCT coefficient. That is, when run = 0 and level = ± 1, the code “1s” is used for the first code, but the code “11s” is used for the second and subsequent codes. FIG. 16 shows the run following the escape code,
Indicates the fixed length code of the level. The code length of the run portion is 6 bits, and the code length of the level portion is 8 bits.
May be a bit.

In this embodiment, the variable length code decoding means 10
2 is the data D for each byte from the data string of the variable length code.
, And with reference to the record of the number corresponding to the data D of the table corresponding to the state S, the “decoded data sequence” is extracted and decoded, and the decoding process is executed while repeatedly obtaining the “next state”. I do.

The state S that can be taken by the variable-length code decoding means 102 is roughly divided into an initial state SFn, an EOB detection state SEOB, an error detection state SERR, a run decoding state SRRUNn, a level decoding state SLVLn, and a normal state SNn. There is a state. In the second embodiment, as the fixed-length decoding state, a run decoding state SRRUNn and a level decoding state SLV
Ln has two states.

FIG. 17 shows the states that the variable-length code decoding means 102 can take, the respective states, the meanings indicated by the states, and the decoding tables corresponding to the respective states. 17A shows an initial state, an EOB detection state, an error detection state, a run decoding state, and a level decoding state, and FIG. 17B shows a normal state.

The initial state SFn is a state at the start of decoding, and has eight types of states (SF1 to SF8) depending on the bit number of 1-byte data to be decoded. Here, the state of “initial state: decoding from the n-th bit” is referred to as state SFn. In the initial state, one of the decoding tables TF1 to TF8 is referred to corresponding to each state. The state of decoding in the initial state is the same as in the case of the first embodiment shown in FIG. However, when there is a bit string “1x” at the decoding start position, decoding is performed as a codeword “1s” used only for the first code in the DCT of MPEG1.
Different from the embodiment. Also, unlike the first embodiment, the second
The initial state in the embodiment is a state that can be taken only when decoding is started.

The EOB detection state SEOB is a state indicating detection of an EOB code. In this case, the decoding table is not referred. This EOB detection state is the same as in the case of the first embodiment shown in FIG. However, the EOB code is "10"
The difference from the first embodiment is that the code has a bit length.

The error detection state SERR is a state in which an error is detected when a code that cannot be assumed as a variable length code assumed in the fetched data is detected. Also in this case, the decoding table is not referred.

The run decoding state SRUNn is a state in which an escape code is detected and the fixed-length run shown in FIG. 16 is decoded. Since the code length of the fixed-length code of the run portion is 6 bits, six types (SRU) from the first bit to the sixth bit depend on which bit of the data D includes the bit after the run code.
N1 to SRUN6). In this run decoding state, the decoding table TRUN1 corresponds to each state.
~ TRUN6.

The level decoding state SLVLn is a state in which the escape code is detected, the fixed length run shown in FIG. 16 is decoded, and then the fixed length level decoding is performed. Data D
Include eight bits from the first bit to the eighth bit (SLVL1 to SLV
L8). In this level decoding state, the decoding tables TLVL1 to TLVL correspond to the respective states.
8 is referred to. There are two types of code lengths of the fixed length code of the level portion: 8 bits and 16 bits. In the case of 16 bits, the decoding state of the nth bit (n> 8) is n-8.
Assign to the decoding state of the bit.

FIG. 18 is an explanatory diagram showing all the states of the run decoding state and the level decoding state. (A) of FIG.
Is a case where the escape code “000001” is provided from the third bit to the last (eighth bit) of the (N + 1) th byte. In this case, when processing the (N + 2) th byte, the decoding is performed from the first bit of the run portion, so that the state is “run decoding: from the first bit”. Since the (N + 2) th byte includes the first two bits of the level portion, when processing the (N + 3) th byte, decoding is performed from the third bit of the level, so that the state is “level decoding: from the third bit”.

FIG. 18B shows a case where an escape code is present in the second to seventh bits in the (N + 1) th byte and includes the first bit of the run portion. In this case, N
When processing the +2 byte, the decoding is performed from the second bit of the run portion, so that the state becomes “run decoding: from the second bit”. Since the (N + 2) th byte includes the first three bits of the level portion, when processing the (N + 3) th byte, decoding is performed from the fourth bit of the level.
From the bit position. " (C) is a case where the escape code is present in the first to sixth bits in the (N + 1) th byte and includes the first two bits of the run portion. In this case, when processing the (N + 2) th byte, the 3
Since decoding is performed from the bit, the state becomes “run decoding: from the third bit”. Since the (N + 2) th byte includes the first 4 bits of the level portion, when processing the (N + 3) th byte, decoding is performed from the 5th bit of the level, so that the state is “level decoding: from the 5th bit”. In (D), the escape code starts from the 8th bit of the Nth byte.
In the + 1st byte, there is an escape code up to the 5th bit, which includes the first 3 bits of the run portion from the 6th bit. In this case, when processing the (N + 2) th byte,
To perform decoding from the fourth bit of the run part, "Run decoding:
From the fourth bit ". In the N + 2 byte,
Since the first 5 bits of the level portion are included, when processing the (N + 3) th byte, decoding is performed from the 6th bit of the level, so that the state is “level decoding: from the 6th bit”.

FIG. 18E shows a case where the escape code starts from the 7th bit of the Nth byte, the escape code is up to the 4th bit in the N + 1th byte, and the first 4 bits of the run portion are included. is there. In this case, when the N + 2 byte is processed, the state becomes “run decoding: from the 5th bit”, and when the N + 3 byte is processed, the state becomes “level decoding: from the 7th bit”. (F) is N
The escape code starts from the 6th bit of the byte,
In the (N + 1) th byte, there is an escape code up to the third bit, which includes the first 5 bits of the run portion. In this case, when processing the (N + 2) th byte, “run decoding:
From the 6th bit, the state becomes "level decoding: from the 8th bit" when processing the (N + 3) th byte. (G) shows a case where the escape code starts from the fifth bit of the Nth byte. In this case, the (N + 1) th byte is in the state of “normal state: surplus bit string: 0000”, and in this case, it is possible to decode up to the run part. When processing the (N + 2) th byte, the decoding is performed from the first bit of the level, so that the state is “level decoding: from the first bit”. (H) is a case where the escape code starts from the fourth bit of the Nth byte. In this case, the (N + 1) th byte is in the state of “normal state: surplus bit string: 00000”, and in this case, even the run portion can be decoded. When processing the (N + 2) th byte, the decoding is performed from the second bit of the level, so that the state is “level decoding: from the second bit”.

The normal state SNn is a state in which a “remainder bit string” that cannot be decoded occurs, and has a state of the number of possible combinations of the “remainder bit string”. In this case, a combination of “remainder bit strings” is determined by the same method as the processing flow shown in FIG. 9 of the first embodiment. In the second embodiment, when the "remaining bit string" is "1", there is a state of SN3 immediately after the start of decoding (initial), in addition to the normal SN4, and the normal state is different from the initial state. There is a state SN1 in which there is no “remaining bit string” in the state.

FIG. 19 is an explanatory diagram showing an example of the state SN1. In this case, the decoding is started from the data before the (N + 1) th byte, and the code up to the end (the 8th bit) of the Nth byte is decoded. It is SN1.
FIG. 19A shows a case where the codeword “10” (EOB code) is decoded at the beginning of the (N + 1) th byte. On the other hand, in the state SF1 of “initial state: decoding from the first bit”, the case where s = 0 of the codeword “1s” is decoded, and run = 0,
Obtain level = 1 data. (B) shows a case where, at the beginning of the (N + 1) th byte, the case where s = 0 of the codeword “11s” is decoded, and data of run = 0 and level = 1 are obtained. On the other hand, the state SF1 of the “initial state: decoding from the first bit”
Then, the case where s = 1 (for 2 bits) of the code word “1s” is decoded, and data of run = 0 and level = −1 are obtained. (C)
Is s = 1 of the code word “11s” at the beginning of the (N + 1) th byte.
Is decoded to obtain data of run = 0 and level = -1. On the other hand, in the state SF1 of “initial state: decoding from the first bit”, the case where the code word “1s” is s = 1 (for 2 bits) is decoded, and the data of run = 0 and level = −1 are decoded. obtain. In cases other than (A) to (C) in FIG. 19, decoding is performed in the same manner as in state SF1.

FIG. 20 is an explanatory diagram showing an example of states SN3 and SN4. The state SN3 shown in FIG.
This is a state that can be taken only immediately after the start of decoding (initial), and is a state where the decoding start position is the 8th bit of an address, and if the 8th bit is “1”, the state shifts to the next. In the example of the figure, decoding starts from the 8th bit of the Nth byte, and since the 8th bit is “1”, when processing the (N + 1) th byte, first, only the first coefficient of the DCT is used. Is decoded (run = 0, level = ± 1).

The state SN4 shown in FIG. 20B is a state that can be taken in a normal state, and is a case where “1” remains after decoding one or more codes. In the example shown in the figure, decoding is started before the 8th bit of the Nth byte, and one or more codewords are decoded. The Nth byte is decoded and “1” is set as “the remaining bit string”. Indicates the state in which the error has occurred. In this case, when processing the (N + 1) th byte, first, the codeword “10” (escape code) or the codeword “11s” (run =
0, level = ± 1).

In the normal state, a total of 225 types of states (SN1 to SN
225). In each of these states, the decoding tables TN1 to TN225 are respectively referred to. The possible combinations of the “remaining bit strings” are obtained by the same method as the processing flow shown in FIG. 9 of the first embodiment, and then
The special state described above, "Remainder bit string: None"
(SN1) and “Remainder bit string: 1 (initial)” (SN
3) can be obtained by renumbering.

As described above, the states that the variable-length decoding means 102 can take include the “initial state” 8 as shown in FIG.
Type, EOB detection status 1 type, Error detection status
1 type, “Run decoding status” 6 types, “Level decoding status”
There are eight types and 255 types of “normal state”.

FIGS. 21 to 23 show examples of the decoding table in the second embodiment. Each table has records from 0 to 255,
21 to 23 show only some of the records. In parentheses of the record numbers, the record numbers are represented in binary numbers. In each record, when an 8-bit binary digit string corresponding to the record number appears in the state of the table, the row of the decoded data (a set of a run and a level) corresponding to the bit string is described in the record “ The next state is stored in the "next state" of the record. Also, in the run and the level following the escape code, there are cases where the level is 8 bits and there are cases where the level is 16 bits. Stored. Therefore, each record is composed of “record number” and corresponding “data”, “next state”, “next state 2”, and “level 2”. These tables are created in advance in correspondence with the contents of the variable length codes and stored in the decoding table 103.

In the second embodiment, in addition to the decoding of the variable-length code described in the first embodiment, the decoding of the fixed-length code following the escape code is performed by merely extracting the bit string in byte units and referring to the table. You can do it.

Here, FIG. 24 shows an example of a bit string including an escape code, and the operation principle of the variable length code decoding means of the second embodiment will be briefly described. In the processing of the data at the address 2002, when decoding the data after the escape, as the run, the first to third bits are "010" and the fourth to sixth bits are "010000". The value 16 of the bit string is obtained, and the “next state” is set as a decoding state from the fourth bit of the run (“run decoding: from the fourth bit”). Then, in the processing of the data at the address 2003, the value 3 of the bit string "011" is obtained as the run, and 3 is added to the previously obtained 16 to obtain the complete run value 19. Next, a value 40 of a bit string of “00101000” in which the first to fifth bits are “00101” and the sixth to eighth bits are 0 is obtained, and the “next state” is set to the level. The decoding state starts from 6 bits (“level decoding: from the 6th bit”). Then, in the processing of the data at the address 2004, the value 2 of the bit string of “010” is obtained as the level, and 2 is added to the previously obtained 40 to obtain the complete level value 42. Thus, run = 19 and level = 42 can be obtained as decoded data.

Next, a detailed procedure for creating a decoding table will be described with reference to FIGS. The creation of the record with the record number RN of a certain table T can be obtained from the state S corresponding to the table T by the processing flows shown in FIGS. Here, different processing flows are used depending on the type of the state S.

FIGS. 25 and 26 show a processing flow for creating a record with the record number RN in the table TFn referred to in the initial state SFn. The procedure will be described below.

First, at step 2501, the record number R
A bit string in which N is represented by an 8-bit binary number is substituted into BIT. If the state S is “initial state: decoding from the Nth bit” in step 2502, BIT is added to Code.
From the Nth bit to the end. Then step 2
At 503, a codeword C other than "10" and "1s" of the variable length code shown in FIG. In step 2504, it is determined whether or not the above codeword C is found. If found, the process proceeds to step 2505, and if not found, the process proceeds to step 2515. Step 250
In 5, it is determined whether or not the codeword C is an escape code.
If it is a sign, the process proceeds to step 2601 in FIG. 26 (A). If not the escape code, the next step 2506
Proceed to.

In step 2506, the decoded data corresponding to the code word C is added to the "data" item of the record. Then, in step 2507, the codeword C is removed from the beginning of the code, and in step 2508, it is determined whether the length of the code is 0 or not. Here, the length of Code is 0
If it is, the process proceeds to step 2525; otherwise, the process proceeds to the next step 2509. In step 2509, the variable length code “1” shown in FIG.
Find codeword C other than "s". And step 251
At 0, it is determined whether or not the above codeword C is found. If found, the process proceeds to step 2511; otherwise, the process proceeds to step 2521. In step 2511, it is determined whether or not the codeword C is an EOB code. If the code word C is an EOB code, the process proceeds to step 2526; otherwise, the process proceeds to the next step 2512. In step 2512, it is determined whether or not the code word C is an escape code. If the code word C is an escape code, the process proceeds to step 2601 in FIG. 26 (A); otherwise, the process proceeds to the next step 2513. Step 251
In 3, the decoded data corresponding to the code word C is added to the "data" item of the record. And step 2514
Then, the code word C is removed from the beginning of the code, and the process returns to step 2508 to repeat the same processing.

In step 2504, "1"
No code word C other than "0" and "1s" is found, and
When proceeding to 15, the contents of the state S and Code are determined,
If the state S is SF8 and the code is “1”, the process proceeds to step 2519; otherwise, the process proceeds to the next step 2516. In step 2516, Cod
In step e, the head of the codeword C shown in FIG. 15 is found. Then, in step 2517, it is determined whether or not the head of the codeword C has been found.
If not found, go to step 2520. In step 2518, the remaining bit sequence is Code
Is stored in the “next state” of the record, and the process is terminated.

In step 2515, the state S is SF8 and C
When the mode is “1” and the process proceeds to step 2319, the normal state SN3 is stored in the “next state” of the record, and the process ends. If the head of the code word C is not found in step 2517 and the process proceeds to step 2520, the error detection state SER is displayed in the “next state” of the record.
R is stored, and the process ends.

In step 2510, “1” is added to the head of Code.
When the code word C other than "s" is not found and the process proceeds to step 2521, the head of the code word C other than "1s" shown in FIG. And step 2522
To determine whether the head of the codeword C has been found,
If found, go to the next step 2523; otherwise, go to step 2524. In step 2523, the normal state SNn in which the remaining bit string becomes Code is stored in the “next state” of the record, and the process ends. However, when the code is “1”, the normal state SN4 is stored as the “next state”.

If the head of the code word C is not found in step 2522 and the process proceeds to step 2524, the error detection state SERR is stored in the "next state" of the record, and the process ends. If the length of the code is 0 in step 2508 and the process proceeds to step 2525, the normal state SN1 is stored in the “next state” of the record, and the process ends. If the codeword C is an EOB code in step 2511 and the process proceeds to step 2526, the EOB detection state SEOB is stored in the “next state” of the record, and the process ends.

Step 2505 or step 2512
In FIG. 26, the code word C is an escape code.
When the process proceeds to 01, 0 is assigned to each of the run and the level. Then, in step 2602, Co
Step 26 removes the code word escape from the beginning of de.
In 03, the length of Code is substituted for Len. Next, in step 2604, the RU is substituted with the sixth bit from the beginning of the code. If the length of the code is less than 6 bits, the entire code is substituted. Then, in step 2605, “0” is added after the RU until the RU has 6 bits. For example, when the RU is “101”, three “0” s are added at the end to make “101000”.

Next, at step 2606, the RU is set to 2 in the run.
A decimal value (≧ 0) corresponding to the case of a decimal number is substituted. In step 2607, the value of Len is determined. If Len is smaller than 6, the next step 260
8; otherwise, to step 2610.
In step 2608, (run, level) at this time is added to the “data” of the record as decoded data. Then, in step 2609, the run decoding state SRUNn is stored in the “next state” of the record, and the processing is terminated. Here, n is Len + 1 (1 to 6).

When Len is 6 or more in step 2607 and the process proceeds to step 2610, 6 bits are removed from the head of the code, and in step 2611, the length of the code is substituted for Len. And step 2612
In step 2613, Code is substituted for LV.
“0” is added after LV until V becomes 8 bits. Next, in step 2614, a decimal value (≧ 0) corresponding to the case where LV is a binary number is substituted for level. Then, at step 2615, the (run, lev
el) is added to the “data” of the record as decrypted data. Subsequently, in step 2616, the level decoding state SLVLn is stored in the “next state” of the record, and the process ends. Here, n is Len + 1 (1 to 8).

FIG. 27 is a processing flow for creating a record with the record number RN in the table TNn referred to in the normal state SNn. The procedure will be described below.

In this case, first, in step 2701, a bit string in which the record number RN is represented by an 8-bit binary number is substituted into BIT. In step 2702, if the state S is “normal state: remaining bit string REM”, C
A value obtained by connecting REM and BIT in order is substituted for "mode". Next, in step 2703, it is determined whether or not the state S is SN3. If the state S is SN3, the process proceeds to step 2503 in FIG. 25 (B). If the state S is not SN3, the process proceeds to step 2509 in FIG. 25 (C).

FIG. 28 is a processing flow for creating a record of the record number RN in the table TRUNn referred to in the run decoding state SRUNn. The procedure will be described below.

In this case, first, in step 2801, a bit string in which the record number RN is represented by an 8-bit binary number is substituted for BIT. Next, at step 2802, 0 is assigned to each of the run and the level.
Then, in step 2803, when the state S is “run decoding: from the n-th bit”, (7-
Substitute the bits up to the (n) th bit, and set (8-
n) Substitute the bits after the first. Next, step 2804
Then, a decimal value (≧ 0) corresponding to the case where the RU is a binary number is substituted for run, and the process proceeds to step 2611 in FIG. 26 (D).

FIG. 29 is a processing flow for creating a record with the record number RN in the table TLVLn referred to in the level decoding state SLVLn. The procedure will be described below.

In this case, first, in step 2901, a bit string in which the record number RN is represented by an 8-bit binary number is substituted for BIT. Next, in step 2902, 0 is assigned to each of the run and the level.
Then, in step 2903, the state S becomes “level decoding: n
In the case of “from the bit”, the record is changed to “next state 2”.
LVLn is stored, and (9-n) from the head of BIT is stored in LV.
The bits up to the bit are substituted, and (10-
n) Substitute the bits after the first. Next, step 2904
Then, a decimal value (≧ 0) corresponding to the case where LV is a binary number is substituted into level. And step 2905
Then, (run, level) at this time is added to the “data” of the record as decoded data. Next, step 2906
Then, Code is substituted into LV2, and in step 2907,
“0” is added after LV2 until LV2 has 8 bits. Then, in step 2908, “1” corresponding to the case where LV2 is a binary number in “level 2” of the record
The value of the decimal number (≧ 0) is stored, and step 250 in FIG.
Proceed to 9 (C).

According to the procedure described above, FIGS.
Is generated as shown in FIG.

A table TF5 shown in FIG.
This is a table used in a state SF5, that is, a state where decoding is performed from the fifth bit in the initial state. The record of record number 8 in the table TF5 has the bit string “0000”.
In response to the appearance of "1000", decoding from the fifth bit gives a code "10" (run = 0, level = 1),
Since “00” is left, the normal state SN5 indicating (0, 1) in “data” and the surplus bit string “00” in “next state”
Is stored.

The table TF8 shown in FIG.
This is a table used in a state SF8, that is, a state where decoding is performed from the eighth bit in the initial state. The record with the record number 255 in the table TF8 corresponds to the case where the bit string “11111111” appears. When decoding from the eighth bit, there is no corresponding codeword that can be decoded, and “1” is left. This “1” may be the code “1s” used at the beginning of the variable length code. Therefore, nothing is stored in "data", and the remaining bit string is "1" in "next state".
Stores the normal state SN1 for decoding the code “1s”.

The table TN1 shown in FIG.
This is a table used in a state SN1, that is, a state where there is no extra bit string in a normal state. The record of the record number 247 of the table TN1 has the bit string “1111011
In response to the appearance of "1", decoding this will result in the code "1"
11 "(run = 0, level = -1), code" 10 "(EO
B) is obtained. Therefore, "data" contains (0,-
1) is stored, and the EOB is up to 5 bits and the next bit string starts from the 6th bit.
6) is stored. In the “next state”, the EOB detection state SEOB is stored.

The table TN2 shown in FIG.
This table is used in a state SN2, that is, a state where the remaining bit string is “0” in the normal state. The record of record number 0 in the table TN2 corresponds to the case where the bit string “00000000” appears. In this case, there is no corresponding code word that can be decoded, and the remaining bit string cannot be decoded and the remaining bit string is “000000000”. Therefore, nothing is stored in the "data", and the normal state SN57 is stored in the "next state".

Table TN57 shown in FIG.
Is a table used in a state SN57, that is, a state where the remaining bit string is “000000000” in the normal state. The record of the record number 46 in the table TN57 corresponds to the case where the bit string “00101110” appears, and when this is decoded, the code “00000000000101110” (run =
14, level = 2) is obtained, and there is no remaining bit string. Therefore, (14, 2)
The "next state" stores the normal state SN1.

The table TN3 shown in FIG.
State SN3, that is, a table used in a state where the remaining bit string is “1” in the normal state and the code “1s” is decoded first. The record with the record number 152 in the table TN3 corresponds to the case where the bit string “10011000” appears. When this is decoded, the code “11” (run = 0, level = −1) and the code “001100” (run = 4, level = 1), and the remaining bit string becomes “0”.
Therefore, (0, −1) and (4, 1) are stored in “data”, and the normal state SN2 is stored in “next state”.

The table TN4 shown in FIG.
This is a table used in a state SN4, that is, a state where the remaining bit string is “1” in the normal state. In this state, the code “1s” is not decoded because it is not the first decoding unlike the case of (F). The record of the record number 152 in the table TN4 corresponds to the case where the bit string “10011000” appears, and when this is decoded, the code “110” (run = 0, level = 1) and the code “0110” (run = 1) , Level = 1), and the remaining bit string becomes “00”. Therefore, (0, 1) and (1, 1) are stored in the “data”, and the normal state SN5 is stored in the “next state”.

The table TN5 shown in FIG.
In the state SN5, that is, in the normal state, the surplus bit string is “0”.
This is a table used in a state of "0". The record of record number 23 in the table TN5 corresponds to the case where the bit string “00010111” appears, and when this is decoded, the code “000001” (escape) is obtained. This is followed by a 6-bit run and 8-bit or 16-bit levels.
The data processed here is the first 4 bits of the run "0111"
And the value of the run corresponding to this is the value 28 of the bit string “011100” when the remaining two bits are set to 0. At this time, since no level is included, the value of the level is set to 0. The “next state” is SRRUN5 because decoding starts from the fifth bit of the run. Therefore, (28, 0) is stored in the “data”, and the run decoding state SRUN5 is stored in the “next state”. Similarly, record number 24
Record includes the first four bits "1000" of the run, so the value of the run is the value 32 corresponding to the bit string "100000". Therefore, (32,0) is stored in "data",
The “next state” stores the run decoding state SRRUN5.

Table TN18 shown in FIG.
Is a table used in a state SN18, that is, a state where the remaining bit string is “00000” in the normal state. The record with the record number 136 of the table TN18 corresponds to the case where the bit string “10001000” appears. When this is decoded, the code “000001” (escape) is obtained, and the bit string “000100” is obtained as the run of the subsequent 6 bits. From this, the value of the run becomes 4. The subsequent one bit “0” is the first bit of the 8-bit or 16-bit level. The level is a bit string “00000000” in which the remaining 7 bits are “0” as an 8-bit level, so that the level value is 0. The “next state” is SLVL2 because decoding starts from the second bit of the level. Therefore, "data"
And (4, 0) are stored in the “next state”, and the level decoding state SLVL2 is stored in the “next state”.

Table TRUN5 shown in FIG.
Is a table used in a state SRRUN5, that is, a state in which a run is decoded from the fifth bit. The record with the record number 160 of the table TRUN5 corresponds to the case where the bit string “10100000” appears. The first two bits “10” are the fifth to sixth bits of the run.
Get a value of 2 for the run. The remaining 6 bits “100000” are the first to sixth bits of the level, and the level value 12 corresponding to the bit string “10000000” is set as an 8-bit level.
Get 8. The "next state" is SLVL7 since decoding starts from the seventh bit of the level. Therefore, (2,128) is stored in “data”, and the level decoding state SLVL7 is stored in “next state”. Similarly, the record with the record number 190 corresponds to the case where the bit string “10111110” appears, the run value is also 2, and the level value is the value 248 of the bit string “11111000”. Therefore,
"Data" stores (2,248) and "Next state"
Stores the level decoding state SLVL7.

Table TLVL2 shown in FIG.
Is a table used in a state SLVL2, that is, a state in which the level is decoded from the second bit. The record of record number 1 in the table TLVL2 corresponds to the case where the bit string “00000001” appears, and the first 7 bits “0000000” obtain the level value 0 from the second bit to the eighth bit of the level. Therefore, the first decoded data stored in “data” is run = 0, level = 0, and (0,
0). The remaining one bit “1” performs two interpretations. One is a case where the level is 8 bits, and decoding of a normal variable length code is performed. In this case, no codeword is obtained for the bit string “1”, and the “next state” is the normal state SN4 in which the remaining bit string is “1”. Therefore, only (0, 0) is stored in “data”, and the normal state SN4 is stored in “next state”. The other is level 1
This is the case of 6 bits, and the remaining 1 bit is regarded as the ninth bit of the level. And "level 2" has level 9
A bit string “10” in which the value from the bit to the 16th bit, that is, the ninth bit is “1” and the subsequent bits are “0”
The value 128 of “00000” is stored. Also, “Next state 2”
In order to be able to interpret the 10th to 16th bits of the level, the decoding state is substituted from the second bit of the level, that is, stored in the “next state 2” as the level decoding state SLVL2.

Similarly, in the record of record number 6 in the table TLVL2, since the first 7 bits are “0000011”, a level value 3 is obtained, and “0”,
3) is stored. "Next state" is the surplus bit string "0"
Is left, the normal state SN2 is set. “Level 2” is a value 0 corresponding to a bit string “00000000” obtained by adding seven “0” s to the remaining bit string “0”, and the “next state 2”
Becomes the level decoding state SLVL2 by substituting the decoding state from the second bit of the level.

A table TLVL7 shown in FIG.
Is a table used in a state SLVL7, that is, a state in which the level is decoded from the seventh bit. The record of record number 1 in the table TLVL7 corresponds to the case where the bit string “00000001” appears. The first two bits “00” are the seventh to eighth bits of the level.
Get level value 0. Therefore, the first decoded data stored in the “data” is (0, 0) at run = 0 and level = 0. The remaining bit string “000001” is interpreted in two ways. One is a case where the level is 8 bits, and decoding of a normal variable length code is performed. In this case, the bit string "000001"
Is an escape code, so the “next state” is SRRUN1 from the first bit of run decoding. Also, (0, 0) is entered as the run and level values. This allows
The processing flow is simplified. Therefore, "data"
(0, 0) and (0, 0) are stored, and the run decoding state SRRUN1 is stored in the “next state”. The other case is when the level is 16 bits, and the remaining bit strings are regarded as the ninth to 14th bits of the level. In the “level 2”, a value from the ninth bit to the 16th bit of the level, that is, a value 4 of an 8-bit bit string “00000100” obtained by adding “0” after the remaining “000001” is stored. The “next state 2” is a level 7 in order to be able to interpret the 15th to 16th bits of the level.
The decoding state is substituted from the bit number, that is, it is stored in the “next state 2” as the level decoding state SLVL7.

Similarly, the record of record number 112 in table TLVL7 has the first two bits “0” as the level.
Since “1” is stored, (0, 1) is first stored in “data”, and subsequent “110” obtains (0, 1), and the bit string “000” remains. Therefore, (0, 1), (0, 1)
And the normal state SN8 is stored in the “next state”. “Level 2” is a value 192 corresponding to an 8-bit bit string “11000000” obtained by adding “0” after “110000” from the third bit. Similarly, record number 1
50 records have the first two bits "10" as the level
Therefore, first, (0, 2) is stored in the “data”, and the subsequent “01011” obtains (2, −1), and the bit string “0” remains. Therefore, "data" includes (0, 2), (2, -1)
Is stored, and the normal state SN2 is stored in the “next state”. “Level 2” is a value 88 corresponding to an 8-bit bit string “01011000” obtained by adding “0” after “010110” from the third bit.

Next, a specific example of a record acquisition procedure for a table will be described with reference to the processing flows of FIGS. 25 to 29, taking six of the records shown in FIGS. 21 to 23 as examples.

In the case of table TF5, record 8 (state S = SF5, RN = 8) in FIG.
“00001000” is substituted (step 2501), and Code
= “1000” is substituted (step 2502). And
Since the codeword "1s" is found at the beginning of the code (step 2503), steps 2504 to 2505 are performed.
To step 25 since the codeword is not escape.
Proceed to 06. In this case, since s = 0 in the code word “1s”, (0, 1) is added to “data” (step 2506). Next, the codeword “1s” from the beginning of the code
And set Code = “00” (step 2507),
In this case, since the length of the Code is 2, step 250
8 to step 2509, where "1"
Find codewords other than "s". Here, since the codeword is not found, the process proceeds from step 2510 to step 2521 to find the head of the codeword other than “1s” in Code. In this case, Code “00” is replaced with codeword “00101s”
Since it is the first part such as "00100110s", step 2
The process proceeds from step 522 to step 2523, where SN5 indicating "normal state: surplus bit string: 00" is stored in the "next state", and the processing ends.

In the case of table TN4 and record 152 (state S = SN4, RN = 152) in FIG. 22G, BIT = “10011000” is substituted first (step 2701),
Since REM = “1”, Code = “110011000” is substituted (step 2702). And state S is SN3
Therefore, the process proceeds from step 2703 to step 2509. In this case, since the code word “11s” is found at the head of the code (step 2509), the process proceeds from step 2510 to step 2511. And the code word is EOB or
Since it is not escape, steps 2511 to 25
Then, the process proceeds to step 2513 via "12", and "11"
A value (0, 1) obtained by decoding "0" is added. Next, Cod
Code = “001100” excluding the code word “11” from the beginning of e
0 ”(step 2514), and the flow proceeds to step 2508.

At this time, since the length of the code is 7, the flow advances from step 2508 to step 2509 again. In the second step 2509, “00110” is added to the head of the Code.
s "is found, so steps 2510 to 2
Proceed to 511. Then, since the code word is not EOB or escape, the process proceeds from step 2511 to step 2513 via step 2512, and a value (1, 1) obtained by decoding "001100" is added to "data". Next, code = “00” is set by removing the code word “00110” from the head of the code (step 2514), and the process returns to step 2508. At this time, since the length of the code is 2, the process proceeds from step 2508 to step 2509 again. Third step 2509
Since no code word other than "1s" is found at the beginning of the code, the process advances from step 2510 to step 2521 to find the head of the code word other than "1s" in the code. In this case, Code “00” is replaced with codewords “00101s” and “00”.
100110s ”or the like, so step 252
From 2, the process proceeds to step 2523, where SN5 of “normal state: surplus bit string: 00” is stored in the “next state”, and the processing ends.

In the case of table TN5 and record 23 (state S = SN5, RN = 23) in FIG.
T = “00010111” is substituted (step 2701), and RE
Since M = “00”, Code = “0000010111” is substituted (step 2702). Since the state S is not SN3, the process proceeds from step 2703 to step 2509. In this case, since the code word “000001” is found at the head of the code (step 2509), the process proceeds from step 2510 to step 2511. And this codeword is esca
Since it is pe, steps 2511 to 2512
Then, the process proceeds to step 2601.

Next, run = 0 and level = 0 are substituted (step 2601), and escape is removed from the head of the code to set Code = “0111” (step 2602). Subsequently, Len = 4 (step 2603), and RU =
As "0111" (step 2604), "0" is added after this RU until it becomes 6 bits, and RU = "011100" (step 2605). Then, a decimal value corresponding to the case where this RU is a binary number is substituted into run and
= 28 (step 2606). In this case, Len
= 4 <6, so steps 2607 to 2
Proceed to 608, and run and level at this time
The value (28, 0) of l) is added to “data”. Then, since n = Len + 1 → 5, S is changed to the “next state”.
RUN5 is stored and the process is terminated (step 2609).

In the case of table TN18 and record 136 (state S = SN18, RN = 136) in FIG.
First, BIT = “10001000” is substituted (step 270
1) Since REM = “00000”, Code = “000”
0010001000 ”(step 2702). Since the state S is not SN3, the process proceeds from step 2703 to step 2509. In this case, since the code word “000001” is found at the head of the code (step 2509), the process proceeds from step 2510 to step 2511. And
Since this code word is escape, the process proceeds from step 2511 to step 2601 via step 2512. Next, run = 0 and level = 0 are substituted (step 260
1) Remove the escape from the beginning of the Code
= "0001000" (step 2602). continue,
Assuming that Len = 7 (step 2603), RU = “00010”
"0" (steps 2604 and 2605). In this case, since the RU has 6 bits, there is no need to add “0”. Then, a decimal value corresponding to the case where this RU is a binary number is substituted into run to make run = 4 (step 260).
6).

In this case, since Len = 7, the value becomes 6 or more, so that the process proceeds from step 2607 to step 2610, where C
Code = “0” by removing the first 6 bits of mode. Then, Len = 1 (step 2611),
Assuming that LV = “0” (step 2612), “0” is added after this LV until it becomes 8 bits, and LV = “0000000”
"0" (step 2613). Then, a decimal value corresponding to the case where this LV is a binary number is substituted into level to make level = 0 (step 2614). At this time, the value (4, 0) of the run and the level is added to “data” (step 2615), and n = Len
Since + 1 → 2, SLVL2 is stored in the “next state” and the processing is terminated (step 2616).

In the case of table TRUN5, record 160 (state S = SRRUN5, RN = 160) in FIG. 23J, BIT = “10100000” is substituted first (step 28).
01), and assign run = 0 and level = 0 (step 28)
02). In this case, since n = 5, up to the second bit of the BIT is substituted for RU, RU = “10”, and B
Code = “100000” by substituting the third and subsequent bits of IT
(Step 2803). Then, a decimal value corresponding to a case where the RU is a binary number is substituted into run, and run =
2 (step 2804), and the process proceeds to step 2611. Next, Len = 6 (step 2611), and LV
= “100,000” (step 2612), and after this LV, “0” is added until it becomes 8 bits, and LV = “1000000”
"0" (step 2613). Then, a decimal value corresponding to the case where this LV is a binary number is substituted into level to make level = 128 (step 2614). The values of the run and the level at this time (2, 12
8) is added to “data” (step 2615), and n =
Since Len + 1 → 7, SLVL7 changes to the “next state”.
Is stored (step 2616).

In the case of the table TLVL7, record 150 (state S = SLVL7, RN = 150) in FIG. 23L, BIT = “10010110” is first substituted (step 29).
01), and assign run = 0 and level = 0 (step 29)
02). In this case, since n = 7, “next state 2”
Is stored, and “10” from the first bit of the BIT is substituted for LV, and “010110” for the third and subsequent bits of the BIT is substituted for Code (step 290).
3). Then, a decimal value corresponding to the case where this LV is a binary number is substituted into level to make level = 2 (step 2904). Run and level (lev
The value (0, 2) of el) is added to “data” (step 2905). Next, LV2 = “010110” (step 2906), and “0” is added after this LV2 until the number of bits becomes 8 to make LV2 = “01011000” (step 2907). Then, a decimal value 88 corresponding to the case where LV2 is a binary number is stored in “level 2” (step 2908), and the flow proceeds to step 2509.

In this case, since "0101s" is found at the beginning of the code (step 2509), step 2510
To 2511. And the code word is EOB
Because it is not escape or step 2 from step 2511
The process proceeds to step 2513 via 512, and the value (2, −1) obtained by decoding “01011” is added to “data”. Next, remove Codeword “01011” from the beginning of Code
de = "0" (step 2514) and step 25
Proceed to 08. At this time, since the length of the code is 1, the process proceeds from step 2508 to step 2509 again. In the second step 2509, since a codeword other than "1s" is not found at the head of the code, the process proceeds from step 2510 to step 2521, and the head of the codeword other than "1s" is found in the code. In this case, since Code “0” is the leading part of the code words “00101s” and “00100110s”, the process proceeds from step 2522 to step 2523,
The SN2 that is “normal state: surplus bit string: 0” is stored in the “next state”, and the processing ends.

Next, the operation of the variable length code decoding means 102 according to the second embodiment will be described. As in the first embodiment, the variable-length code decoding unit 102 determines that the position of the bit at which decoding is started is given by the address (the number of bytes from the head of the memory) and the number of the bit in the address. By designating a read address to the variable-length code storage unit 101, a 1-byte code at that address is extracted, the variable-length code is decoded with reference to the decoding table 103, and the decoded data is stored in the decoded data storage unit. The variable length code is decoded by repeating the process of outputting the variable length code to the variable length code.

FIGS. 30 and 31 show the variable length code decoding means 1.
02 shows an example of the processing flow of the operation No. 02, and the decoding procedure of the second embodiment will be described below based on these processing flows.

(Processing in Initial State and Normal State) First, in step 3001, the address of the bit position at which decoding is started (decoding start address on the memory) is stored in a variable ReadP indicating the read address. For,
State number SF corresponding to the bit position at which decoding starts in state S
Substitute n. As described above, the state numbers of the initial state are SF1 for the decoding start bit positions 1 to 8, respectively.
To SF8.

Next, in step 3002, an address ReadP is instructed to the variable-length code storage means 101, and 1-byte data of the corresponding address is acquired from the codes stored in the variable-length code storage means 101. Assign to variable D. Then, in step 3003, by referring to the D-th record of the table corresponding to the state S in the decoding table 103, the run and level decoded data data stored in the "data" and the "next state" Snext are obtained. . Next, in step 3004, the decoded data data is output to the decoded data storage unit 104. However, if the state Snext is SRUN
In the case of n, SLVLn, SEOB, and SERR, the last data of the decoded data data is not output.

Thereafter, at step 3005, Sne is set to state S.
Set xt and transition state. At this time, the state S is S
If it is not EOB or SERR, add 1 to ReadP,
The read position of the data to be decoded is advanced by one byte. In step 3006, it is determined whether or not the state S is the run decoding state SRUNn. If the state S is SRRUNn, the process proceeds to step 3101 for performing run decoding processing (A). In step 3007, the state S is set to the level decoding state SLVLn.
Then, if the state S is SLVLn, the process proceeds to step 3110 for performing level decoding processing (B).
In step 3008, the state S is changed to the EOB detection state SEO.
It is determined whether the state is B or the error detection state SERR. If the state S is SEOB or SERR, the process is terminated. If not, the process returns to step 3002 to repeat the same process. At this time, the last data of the decoded data data has a run of -1, and the level indicates the position of the bit after the EOB in the case of SEOB, and indicates the position of the bit in which an error is detected in the case of SERR. It is.

(Run Decoding Process) In step 3101, the run and level of the last data of the decoded data
Assign to run and level respectively. And step 310
In step 2, the address is read from the variable-length code storage unit 101.
P is designated, and 1-byte data at the corresponding address is obtained and assigned to a variable D. Next, in step 3103, the D-th record of the table corresponding to the state S in the decoding table 103 is referred to, and the run and level decoded data data stored in the “data” and the “next state” Snext
And get Then, in step 3104, the decrypted data data
Run and level of the first data in
Add to evel. Further, in Step 3105, 1 is set to ReadP.
Is added, and the read position of the data to be decoded is advanced by one byte. Also, Snext is set in the state S, and the state is shifted.

Next, in step 3106, it is determined whether or not the state S is the level decoding state SLVLn.
In the case of LVLn, step 31 for performing level decoding processing
Proceed to 11. In step 3107, the value of level is determined. If level is 0 or 128, the flow advances to step 3115 for performing 16-bit level decoding. Then, in step 3108, the run and level are output to the decoded data storage unit 104. At this time, the run outputs the value of run so that the fixed-length code as shown in FIG. 16 is decoded, and the level outputs the level when level <128, and outputs the level when level> 128. Outputs level-256. Then, in step 3109, the second and subsequent data of the decoded data data are output to the decoded data storage unit 104,
Proceed to step 3008.

(Process of Level Decoding) In step 3110, the run and level of the last data of the decoded data data are
Assign to run and level respectively. And step 311
1, the address is read from the variable-length code storage
P is designated, and 1-byte data at the corresponding address is obtained and assigned to a variable D. Next, in step 3112, the D-th record of the table corresponding to the state S in the decoding table 103 is referred to, and the run and level decoding data data stored in the “data”, the “next state” Snext,
“Next state 2” Snext2 and “level 2” level2 are obtained.
Then, in step 3113, the level of the first data in the decoded data data is added to level. In step 3114, 1 is added to ReadP, and the read position of the data to be decoded is advanced by one byte. Also, Snext is set in the state S, and the state is shifted. Then, the process proceeds to step 3107.

(Processing of 16-bit level decoding) In step 3115, if level = 128, the code flag F
Is substituted for −1, and 1 is substituted for the sign flag F when level = 0. Next, in step 3116, level2 is substituted for level, and in step 3117, the state S is set to Snext2. Then, in step 3118, the variable length code storage unit 1
01, an address ReadP is instructed, 1-byte data of the corresponding address is obtained, and is substituted for a variable D. Next, in step 3119, the D-th record of the table corresponding to the state S in the decoding table 103 is referred to,
Run and level decrypted data stored in "Data"
Data and the “next state” Snext are obtained. Then, in step 3120, the level of the first data in the decoded data data is added to level. Further, in step 3121, Rea
One is added to dP, and the read position of the data to be decoded is advanced by one byte. Also, Snext is set in the state S, and the state is shifted. Then, in step 3122, the run and the level are output to the decoded data storage unit 104. At this time, FIG.
The run outputs the value of run so that the fixed-length code as shown in FIG. 6 is decoded, and the level is level when F = 1.
Is output, and when F = -1, level-256 is output. Thereafter, the flow advances to step 3109.

Next, a specific example of the operation of the variable length code decoding means 102 will be described in detail with reference to FIGS. 30 and 31, taking the case of decoding the variable length code shown in FIG. FIG. 32 shows an example of the variable-length code stored in the variable-length code storage means 101. Code contents of each byte are shown in a square frame, and the address is decoded above the frame and decoded at the bottom of the frame. Shows the data. Two numerical values in parentheses of the decoded data are a run and a level. For the contents of each byte,
The upper part in the square frame shows a numerical value in decimal number, and the lower part shows a bit string in binary number. Here, the bit position to start decoding is from the fifth bit of the address 1001. When this variable length code is decoded, the set of run and level is (0, 1), (30, -6), (2, -1), (1)
4, 2) and (0, -1). Tables and records referred to in the following description are all shown in FIGS.

First, the start address 1001 is substituted for ReadP,
In the state S, SF5 is assigned because decoding is performed from the fifth bit in the initial state (step 3001). Then, the address of ReadP, that is, 1-byte data “8 (00001000)” of the address 1001 is obtained from the variable-length code storage unit 101 and substituted into D (step 3002). Next, S becomes S
Since F5 and D are 8, reference is made to record 8 of table TF5 in decoding table 103 (step 30).
03). In this case, decoded data data = (0, 1) and “next state” Snext = SN5 are obtained. And Snext = SN
5, all the decoded data data = (0, 1) is output to the decoded data storage unit 104 (step 300).
4). Subsequently, Snext = SN5 is set in the state S, and Re
Add 1 to adP to set ReadP = 1002 (step 30)
05).

In this case, the state S is the run decoding state SRU
Nn, level decoding state SLVLn, EOB detection state SE
Since it is neither OB nor error detection state SERR,
After steps 3006, 3007, and 3008, the process returns to step 3002.

In the second step 3002, the address
The 1-byte data “23 (00010111)” of 1002 is obtained and substituted for D. Next, since S is SN5 and D is 23, record 2 of table TN5 in decoding table 103
3, the decoded data data = (28, 0) and the “next state” Snext = SRUN5 are obtained (step 3003).
In this case, since Snext = SRUN5, the decoded data
Do not output the last data of data. That is, data is not output (step 3004). Then, Snext = SRUN5 is set in the state S, 1 is added to ReadP, and ReadP = 1003 is set (step 3005). Since the state S is the run decoding state SRUN5, the process proceeds from step 3006 to step 3101.

Next, the last data (2
8,0) is assigned to run and level, and run = 28, level = 0
(Step 3101). And the address ReadP
1-byte data “190 (10111110)” of “1003” is obtained and substituted into D (step 3102). And S is SR
Since UN5 and D are 190, the record 190 of the table TRUN5 in the decoding table 103 is referred to, and
Decrypted data data = (2,248), “Next state” Snext
= SLVL7 is obtained (step 3103). Subsequently, a run of the first data of the decoded data data = (2,248),
The level is added to run = 28 and level = 0, respectively, and run =
30, level = 248 (step 3104). Further, Snext = SLVL7 is set in the state S, and ReadP is set.
Is added to 1 so that ReadP = 1004 (step 310)
5). Here, since the state S is the level decoding state SLVL7, the process proceeds from step 3106 to step 3111.

Next, 1-byte data "150 (10010110)" of address ReadP = 1004 is obtained and substituted for D (step 3111). S is SLVL7 and D is 150
Therefore, the table TL in the decoding table 103
With reference to the record 150 of the VL7, the decrypted data data =
(0, 2), (2, −1), “next state” Snext = SN
2. “Next state 2” Snext2 = SLVL7 and “level 2” level2 = 88 are obtained (step 3112). Subsequently, the level of the first data (0, 2) of the decoded data data is added to level = 248, and level = 250 (step 3113). Further, Snext = SN2 is set in the state S, 1 is added to ReadP, and ReadP = 1005 is set (Step 3114).

Next, the value of level is determined (step 31).
07), since the level is not 0 or 128, the process proceeds to the next step 3108. Then, the run and level values are output to the decoded data storage unit 104 (step 310).
8). In this case, run = 30 is output as the run,
Since the level is level = 250 and level> 128, the value of level-256, that is, -6 is output.
After that, the second and subsequent data of the decoded data, that is, (2, -1), are output to the decoded data storage unit 104 (step 3109).

Then, the state S is determined (step 300).
8) In this case, since the state S is not the EOB detection state SEOB or the error detection state SERR, step 300
Return to 2. In the third step 3002, the address 10
The 1-byte data “0 (00000000)” of 05 is obtained and substituted into D. Next, since S is SN2 and D is 0, reference is made to record 0 of table TN2 in decoding table 103, and decoding data data = none, “next state” Snext = SN
57 are obtained (step 3003). In this case, there is no decoded data, so no data is output (step 3004). Then, Snext = SN57 is set in the state S, 1 is added to ReadP, and ReadP = 1006 is set (step 3005). Here, the state S is the run decoding state SR
UNn, level decoding state SLVLn, EOB detection state S
Since it is neither EOB nor error detection state SERR, the process returns to step 3002 via steps 3006, 3007 and 3008.

In the fourth step 3002, the address
The 1-byte data "46 (00101110)" of 1006 is obtained and substituted for D. Next, since S is SN57 and D is 46,
With reference to the record 46 of the table TN 57 in the decoding table 103, the decoding data data = (14, 2) and the “next state” Snext = SN1 are obtained (step 3003).
Then, all the decoded data data = (14, 2) is output to the decoded data storage unit 104 (step 3004).
Further, Snext = SN1 is set in the state S, and 1 is added to ReadP to make ReadP = 1007 (step 300).
5). Here, since the state S is not any of the run decoding state SRUNn, the level decoding state SLVLn, the EOB detection state SEOB, and the error detection state SERR, the state S goes through the steps 3006, 3007, and 3008 to the step 3002.
Return to

At the fifth step 3002, the address
Obtain 1 byte data "247 (11110111)" of 1007 and D
Substitute for Next, since S is SN1 and D is 247, the record 247 of the table TN1 in the decoding table 103 is referred to, and the decoded data data = (0, −1),
(−1, 6), “Next state” Snext = SEOB is obtained (step 3003). In this case, since Snext = SEOB, the last data of the decoded data data is removed, that is, only (0, -1) is output to the decoded data storage unit 104 (step 3004). Then, Sne changes to state S.
xt = SEOB is set, and ReadP is not added because it is SEOB (step 3005). And the state S is
Since it is neither the run decoding state SRUNn nor the level decoding state SLVLn, the process proceeds from steps 3006 and 3007 to step 3008, and the process ends because it is in the EOB detection state SEOB. Here, in the bit string after EOB, the address is indicated by ReadP = 1007, and the bit position is:
The level of the last data of the decoded data "data" is indicated by "6".

The bit string shown in FIG. 32 is for decoding a fixed-length code following escape, where there is a byte break in the middle of the run and the code length of the level is 8 bits. It was explained in. FIG. 33 shows an example of another code having a different bit string. FIG. 33A shows an example of a bit string in the case where there is a byte break in the middle of the run and the level code length is 16 bits, and FIG. 33B shows no bit break in the middle of the run and the level code length is 8 bits. (C) shows an example of a bit string having a 16-bit level code length without a byte break in the middle of a run.

As a second specific example of the operation of the variable length code decoding means 102, the operation for decoding the variable length code shown in FIG. 33A will be described in detail with reference to FIGS.

When decoding the address 2002, the state S
= SN5 (“normal state: surplus bit string: 00”) and address ReadP = 2002. In this case, 1 of address 2002
The byte data "24 (00011000)" is obtained and substituted for D (step 3002). Next, since S is SN5 and D is 24, the table TN5 in the decoding table 103 is set.
, The decrypted data data = (32,
0), “Next state” Snext = SRUN5 is obtained (step 3003). Here, since Snext = SRUN5, the last data of the decoded data data is not output. That is, data is not output (step 3004). Then, Snext = SRRUN5 is set in the state S, and ReadP is set.
Is added to 1 so that ReadP = 2003 (Step 300)
5). Since the state S is the run decoding state SRUN5, the process proceeds from step 3006 to step 3101.

Next, the last data (3
(2,0) is assigned to run and level, and run = 32, level = 0
(Step 3101). And the address ReadP
= 1-byte data "160 (10100000)" of 2003 is substituted into D (step 3102). At this time, S becomes S
Since RUN5 and D are 160, the decoding table 103
, The decrypted data data = (2, 128) and the “next state” S
Next = SLVL7 is obtained (step 3103). Subsequently, the run and level of the first data of the decoded data data = (2,128) are added to run = 32 and level = 0, respectively.
run = 34, level = 128 (step 310)
4). Further, Snext = SLVL7 is set in the state S, 1 is added to ReadP, and ReadP = 2004 is set (step 3105). Here, the state S is the level decoding state SLV
Since it is L7, steps 3106 to 311
Proceed to 1.

Next, 1-byte data "1 (00000001)" of address ReadP = 2004 is obtained and substituted into D (step 3111). Then, since S is SLVL7 and D is 1, reference is made to the record 1 of the table TLVL7 in the decoding table 103, and the decoded data data = (0, 0),
(0,0), "next state" Snext = SRUN1, "next state 2" Snext2 = SLVL7, "level 2" level2 =
4 is obtained (step 3112). Then, decrypted data da
The level of the first data (0,0) of ta is level = 128
Is set to level = 128 (step 3113).
Further, Snext = SRUN1 is set in the state S, and Read
Add 1 to P to make ReadP = 2005 (step 311
4).

Next, the value of level is determined (step 31).
07) In this case, since the level is 128, the process proceeds to step 3115. Then, since level = 128, -1 is substituted for the code F (step 3115). Also, le
vel is substituted with level2 = 4 to make level = 4 (step 3116). Next, Snext2 = SLVL7 is set in the state S (step 3117), and 1-byte data “112 (01110000)” of the address ReadP = 2005 is obtained and substituted into D (step 3118). And S is SLVL7,
Since D is 112, reference is made to the record 112 of the table TLVL7 in the decoding table 103, and the decoded data data = (0, 1), (0, 1), "next state" Snext
= SN8 is obtained (step 3119). Subsequently, the level of the first data (0, 1) of the decoded data data is set to level =
In addition to level 4, level = 5 is set (step 3120). Further, Snext = SN8 is set in the state S, and 1 is set in ReadP.
Is added to make ReadP = 2006 (Step 312)
1). Then, the run and level values are output to the decoded data storage unit 104 (step 3122). In this case, run = 34 is output as the run, and the code of the level is F = −1.
-256 = -251 is output. Then decrypted data
The second and subsequent data of data, that is, (0, 1) is output to the decoded data storage unit 104 (step 310).
9).

As a third specific example of the operation of the variable-length code decoding means 102, the operation for decoding the variable-length code shown in FIG. 33B will be described in detail with reference to FIGS.

When decoding the address 2002, the state S
= SN18 (“normal state: surplus bit string: 0000
0 ”) and the address ReadP = 2002. In this case, 1-byte data "136 (10001000)" of the address 2002 is obtained and substituted into D (step 3002). Next, S is SN1
8, since D is 136, the record 136 of the table TN18 in the decoding table 103 is referred to, and the decoded data data = (4, 0) and the “next state” Snext = SLVL
2 is obtained (step 3003). Here, Snext = SL
Since it is VL2, the last data of the decoded data data is not output. That is, data is not output (step 3004). Then, Snext = SLVL2 is set in the state S, 1 is added to ReadP, and ReadP = 2003 is set (step 3005). Here, the state S is the level decoding state S
Since it is LVL2, the process proceeds to step 3110 via steps 3006 and 3007.

Next, the last data (4, 0) of the decoded data data is substituted into run and level, and run = 4, level = 0
(Step 3110). And the address ReadP
= 1-byte data "6 (00000110)" of 2003 is substituted for D (step 3111). At this time, S is SLV
Since L2 and D are 6, reference is made to record 6 of table TLVL2 in decoding table 103, and the
data = (0,3), “next state” Snext = SN2, “next state 2” Snext2 = SLVL2, “level 2” level2
= 0 is obtained (step 3112). Then, decrypted data
The level of the first data (0, 3) of data is added to level = 0 to set level = 3 (step 3113). Further, Snext = SN2 is set in the state S, 1 is added to ReadP, and ReadP = 2004 is set (step 3114).

Next, the value of level is determined (step 31).
07) In this case, since the level is not 0 or 128, the process proceeds to the next step 3108. Then, the run and level values are output to the decoded data storage unit 104 (step 3108). In this case, run = 4 is output as a run, and level = 3 and level <128 for the level. Therefore, the value of level, that is, 3 is output. And
Since there is no second or subsequent data in the decoded data data, no data is output (step 3109).

As a fourth specific example of the operation of the variable-length code decoding means 102, the operation for decoding the variable-length code shown in FIG. 33C will be described in detail with reference to FIGS.

When decoding the address 2002, the state S
= SN18 (“normal state: surplus bit string: 0000
0 ”) and the address ReadP = 2002. In this case, 1-byte data "136 (10001000)" of the address 2002 is obtained and substituted into D (step 3002). Next, S is SN1
8, since D is 136, the record 136 of the table TN18 in the decoding table 103 is referred to, and the decoded data data = (4, 0) and the “next state” Snext = SLVL
2 is obtained (step 3003). Here, Snext = SL
Since it is VL2, the last data of the decoded data data is not output. That is, data is not output (step 3004). Then, Snext = SLVL2 is set in the state S, 1 is added to ReadP, and ReadP = 2003 is set (step 3005). Here, the state S is the level decoding state S
Since it is LVL2, the process proceeds to step 3110 via steps 3006 and 3007.

Next, the last data (4, 0) of the decoded data data is substituted into run and level, and run = 4, level = 0
(Step 3110). And the address ReadP
= 1-byte data "1 (00000001)" of 2003 is obtained and substituted into D (step 3111). At this time, S is SLV
Since L2 and D are 1, reference is made to record 1 of table TLVL2 in decoding table 103, and the
data = (0,0), “next state” Snext = SN4, “next state 2” Snext2 = SLVL2, “level 2” level2
= 128 (step 3112). Subsequently, the level of the first data (0, 0) of the decoded data data is set to level =
In addition to 0, level = 0 is set (step 3113). Further, Snext = SN4 is set in the state S, and 1 is set in ReadP.
Is added to make ReadP = 2004 (Step 311)
4).

Next, the value of level is determined (step 31).
07) In this case, since level is 0, step 31
Proceed to 15. Since level = 0, the code F is 1
Is substituted (step 3115). Also, level to level
2 = 128, and level = 128 (step 3
116). Next, Snext2 = SLVL2 is set in the state S (step 3117), and 1 of the address ReadP = 2004 is set.
The byte data "6 (00000110)" is obtained and substituted for D (step 3118). Since S is SLVL2 and D is 6, the table T in the decoding table 103
Referring to record 6 of LVL2, decrypted data data =
(0, 3), “next state” Snext = SN2 is obtained (step 3119). Subsequently, the level of the first data (0, 3) of the decoded data data is added to level = 128, and level =
131 (step 3120). Further, Snext = SN2 is set in the state S, 1 is added to ReadP, and
dP is set to 2005 (step 3121). And run,
The value of the level is output to the decoded data storage unit 104 (Step 3122). In this case, run = 4 is output as the run and the code F = 1 for the level.
The value of vel, ie, 131, is output. Then, since there is no second or subsequent data in the decoded data data, no data is output (step 3109).

Thus, the DC1 in MPEG1
When decoding a fixed-length code starting with a run of T coefficients, a variable-length code of a level, and an escape code, data can be handled in units of bytes, and decoding can be performed without performing a bit operation.

In the second embodiment, FIGS. 15 and 1
Although the configuration example of decoding the run of the DCT coefficient of MPEG, the variable length code of the level as shown in FIG. 6, and the fixed length code has been shown, any variable length code or the variable length code including the fixed length code can be used. It is applicable, and a decoding table and decoding means can be configured according to the content of the variable length code to be decoded.

In the first and second embodiments described above, the code is extracted and decoded every one byte, that is, every 8 bits. However, the code is extracted and decoded with an arbitrary number of bits. It may be configured. For example, if a code is taken out and decoded every two bytes, each table requires records from 0 to 65535, and a large table is required. Since it is reduced by a factor of 1, the decoding processing time can be reduced. In the case where a code is extracted and decoded every four bits, each table requires records from No. 0 to No. 15, and the size of the table can be reduced. In this case, the number of times that the table is referred to is doubled, and a bit operation is required although it is a fixed process of extracting every four bits.

In this embodiment, the decoding process is terminated by detecting the EOB code. However, the decoding process is terminated under conditions other than the detection of the EOB, such as terminating the process after decoding to a specified address. Such a configuration may be adopted.

As described in detail above, in the present embodiment, a plurality of tables corresponding to the current state S during the decoding process are provided as the decoding table, and the bit string indicated by the RN in that state is recorded in the record RN of each table. By storing the decoded "data" and the "next state", and also for decoding the fixed length part, a value storing the value evaluated for the part included in the record RN is created, and by referring to this decoding table, A variable-length code and a fixed-length portion included in the variable-length code are decoded while extracting a bit string for each byte. As a result, a variable-length code including a fixed-length code can be decoded without performing a bit operation such as a bit shift, so that the decoding process can be executed at a high speed and the processing time required for decoding can be reduced. , Its practical effect is great.

[0209]

As described above, according to the present invention, when decoding only a variable-length code or a code in which a variable-length code and a fixed-length code are mixed, a bit length is There is no need to perform bit operations such as shift,
The decoding can be performed only by taking out the bit string in byte units and referring to the table, so that the decoding processing can be performed at high speed.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of a variable-length code decoding device according to an embodiment of the present invention.

FIG. 2 is a diagram illustrating an example of a variable length code used in the first embodiment.

FIG. 3 is an explanatory diagram showing a code pattern to be decoded in an initial state.

FIG. 4 is an explanatory diagram showing a code pattern and a surplus bit string in a normal state.

FIG. 5 is a diagram showing combinations of remaining bit strings.

FIG. 6 is an explanatory diagram showing eight types of patterns in an EOB detection state.

FIG. 7 is an explanatory diagram showing eight types of patterns in an escape detection state.

FIG. 8 is a diagram showing the correspondence between state numbers of states that can be taken at the time of decoding in the first embodiment, their meanings, and tables to be referred to.

FIG. 9 is a flowchart illustrating a processing procedure for listing “remaining bit strings” that can be taken in a normal state.

FIG. 10 is a diagram illustrating an example of a decoding table used in the first embodiment.

FIG. 11 is a flowchart illustrating a processing procedure for creating a decoding table.

FIG. 12 is a diagram showing a part of a decoding table used in the first embodiment.

FIG. 13 is a flowchart showing a decoding procedure based on the operation of the variable length code decoding means in the first embodiment.

FIG. 14 is an explanatory diagram showing an example of a bit string of a variable length code and a specific example of a decoding process thereof.

FIG. 15 is a diagram illustrating a variable length code of a DCT coefficient run and level of MPEG as an example of a variable length code used in the second embodiment.

FIG. 16 is a diagram showing a run and level fixed-length code by an escape code of a DCT coefficient of MPEG as an example of a variable-length code used in the second embodiment.

FIG. 17 is a diagram showing the correspondence between state numbers of states that can be taken at the time of decoding in the second embodiment, their meanings, and tables to be referred to.

FIG. 18 is an explanatory diagram showing all states of a run decoding state and a level decoding state.

FIG. 19 is an explanatory diagram showing an example of state SN1.

FIG. 20 is an explanatory diagram showing an example of states SN3 and SN4.

FIG. 21 is a diagram illustrating an example of a decoding table used in the second embodiment.

FIG. 22 is a diagram illustrating an example of a decoding table used in the second embodiment.

FIG. 23 is a diagram illustrating an example of a decoding table used in the second embodiment.

FIG. 24 is an explanatory diagram showing an example of a bit string of a variable-length code including an escape code.

FIG. 25 is a flowchart illustrating a processing procedure for creating a table referred to in an initial state.

FIG. 26 is a flowchart illustrating a processing procedure for creating a table referred to in an initial state.

FIG. 27 is a flowchart illustrating a processing procedure for creating a table referred to in a normal state.

FIG. 28 is a flowchart illustrating a processing procedure for creating a table to be referred to in the run decoding state.

FIG. 29 is a flowchart showing a processing procedure for creating a table referred to in the level decoding state.

FIG. 30 is a flowchart showing a decoding procedure by the operation of the variable length code decoding means in the second embodiment.

FIG. 31 is a flowchart showing a decoding procedure based on the operation of the variable length code decoding means in the second embodiment.

FIG. 32 is an explanatory diagram illustrating an example of a bit string of a variable-length code including an escape code and a specific example of a decoding process thereof.

FIG. 33 is an explanatory diagram showing an example of a bit string of a variable-length code including an escape code and another specific example of the decoding process.

FIG. 34 is a diagram illustrating an example of a variable length code.

FIG. 35 is an explanatory diagram showing a bit string by the variable length code of FIG. 34 and its decoded data.

FIG. 36 is a diagram illustrating an example of a decoding table corresponding to the variable-length code in FIG. 34.

FIG. 37 is a flowchart showing a procedure for decoding a variable length code by a conventional decoding method.

FIG. 38 is a flowchart showing a processing procedure for extracting a bit string at a predetermined position.

FIG. 39 is a flowchart illustrating a processing procedure for updating a bit string extraction start position.

FIG. 40 is a diagram illustrating an example of a decoding table used in a conventional decoding method.

[Explanation of symbols]

 Reference Signs List 101 variable length code storage means 102 variable length code decoding means 103 decoding table 104 decoded data storage means

Claims (13)

[Claims] 1. A procedure for extracting a bit string B of N bits (N is an integer equal to or greater than 1) from a variable-length code string to be decoded, and a table corresponding to a current state S, wherein a record corresponding to the extracted bit string B is stored. And obtaining a decoded data sequence V and the next state S 'to transition from this record, and obtaining decoded data of a variable-length code sequence by repeating these procedures. Code decoding method. 2. A plurality of tables respectively corresponding to a state S at the time of decoding, and a plurality of records respectively corresponding to a bit string B extracted from a variable length code string in byte units in each table are provided. A decoding table configured to store at least a decoded data string V indicating data obtained when the bit string B is decoded and a next state S ′ indicating a next transition state after referring to the record A step of extracting a bit string B of N bytes (N is an integer of 1 or more) from a variable-length code string to be decoded; and a step of extracting the bit string B corresponding to the extracted bit string B in a table corresponding to a current state S in the decoding table. A procedure of referring to a record and obtaining a decoded data string V and the next state S 'from the record, and sequentially changing these procedures by repeating these procedures. A decoding method for a variable-length code, comprising obtaining decoded data of a long code string. 3. The state S includes a plurality of initial states SFi in which a state is decoded from the i-th bit (i is an integer of 1 or more) of the extracted bit string B, and the preceding bits are not to be decoded. A plurality of normal states SNx in a state in which a surplus bit string that cannot be decoded occurs as a result of decoding
The decoding method for a variable-length code according to claim 1, comprising: 4. A decoding method for extracting and decoding a bit string for each N bits (N is an integer of 1 or more) from a variable-length code string to be decoded, wherein the i-th bit (i = 1 to N), a plurality of initial states SFi in a state of decoding, and a plurality of normal states SNx in which there is a surplus bit string R that cannot be decoded in the previous decoding result and the surplus bit string R can take. Have a number of records that can be represented by an N-bit binary number, and the records indicated by the N-bit bit string b correspond to the initial state SFi. In the table, a decoded data string V obtained by decoding the bit string b from the i-th bit and a next state S 'corresponding to the next remaining bit string r are stored. The table corresponding to the state SNx stores a decoded data string V obtained by decoding a bit string obtained by combining the remaining bit string R and the bit string b, and a next state S 'corresponding to the next remaining bit string r. A procedure for extracting an N-bit bit string B from a variable-length code string to be decoded by using a decoding table configured as described above, and referring to a record corresponding to the bit string B in the table corresponding to the current state S. A method for obtaining a decoded data sequence V and a next state S 'from a record, and sequentially obtaining decoded data of a variable length code sequence by repeating these procedures. 5. A decoding method for extracting and decoding a bit string every N bytes (N is an integer of 1 or more) from a variable-length code string to be decoded, wherein an i-th bit (i = 1 to N × 8)
, And a plurality of normal states SNx in which the remaining bit strings R that cannot be decoded occur as a result of the previous decoding and the remaining bit strings R can take. These tables have a number of records that can be represented by an N × 8-bit binary number, and records indicated by an N × 8-bit bit string b correspond to the initial state SFi. The table stores a decoded data string V obtained by decoding the bit string b from the i-th bit and a next state S 'corresponding to the next remaining bit string r,
In the table corresponding to the normal state SNx, a decoded data string V obtained by decoding a bit string obtained by combining the remaining bit string R and the bit string b, and a next state S ′ corresponding to the next remaining bit string r Using a decoding table configured to store a bit string B of N bytes in units of bytes from a variable-length code string to be decoded; and a record corresponding to the bit string B in the table corresponding to the current state S. And a procedure for obtaining a decoded data sequence V and the next state S 'from the record, and sequentially obtaining decoded data of a variable-length code sequence by repeating these procedures. Decryption method. 6. When a variable-length code and a fixed-length code are mixed in the variable-length code string to be decoded and a fixed-length identification code CS is provided as a code indicating the start of the fixed-length code, each bit string b in the table is set. In the record corresponding to the above, the record in which the bit sequence b includes at least a part of the fixed-length identification code CS is obtained by decoding the code sequence up to the fixed-length identification code CS as the decoded data sequence V. And the state corresponding to the remaining bits is stored as the next state S ', and decoding of the fixed-length code portion is performed using this record. The variable length code decoding method according to any one of claims 1 to 5, wherein: 7. When a variable length code and a fixed length code are mixed in the variable length code string to be decoded and a fixed length identification code CS is provided as a code indicating the start of the fixed length code, the state S is as follows.
It includes a plurality of fixed-length decoding states SCy in which decoding of fixed-length codes is performed, and refers to a table corresponding to the fixed-length decoding states SCy to decode fixed-length codes after the fixed-length identification code CS. The variable length code decoding method according to any one of claims 1 to 5, wherein: 8. When a variable length code and a fixed length code are mixed in the variable length code string to be decoded, and a fixed length identification code CS is provided as a code indicating the start of the fixed length code, the variable length code string corresponds to the normal state SNx. In the table, in a record indicated by each bit string b, in which the bit string b includes at least a part of the fixed-length identification code CS, the code string up to the fixed-length identification code CS is decoded. The decoded data sequence V obtained by the above, the numerical value v1 determined by the fixed length code part after the fixed length identification code CS in the bit sequence b, and the number of bits of the fixed length code to be decoded next time And the fixed-length decoding state SCy indicating the decoding from the y-th bit of the fixed-length code is indicated by the bit string b in the table corresponding to the fixed-length decoding state SCy. The records, a numerical value v2 obtained by decoding the bit sequence b as y-th bit of the fixed-length code, the decoded data sequence V which is obtained by decoding the code sequence of the fixed length code after
And a next state S ′ indicating the state after decoding, a decoding table is configured. When decoding the fixed-length code, after decoding the numerical value v1 and the numerical value v2, 8. The method for decoding a variable-length code according to claim 4, wherein the fixed-length code is decoded using v1, v1 + v2. 9. The variable-length code and the fixed-length code in the variable-length code string to be decoded are a run and a level of a DCT coefficient in the MPEG image compression method, and a run of the MPEG variable-length coded DCT coefficient. The decoding method for a variable-length code according to any one of claims 6 to 8, wherein the level is decoded. 10. A plurality of tables each corresponding to a state S at the time of decoding, and a plurality of records respectively corresponding to a bit string B extracted from a variable length code string in byte units are provided in each table. Characterized in that at least a decoded data string V indicating data obtained when the bit string B is decoded and a next state S 'indicating a next transition state after referring to the record are stored. Code decoding table. 11. A plurality of initial states SFi in which decoding is started from the i-th bit (i = 1 to N) of a bit string B of N bits (N is an integer of 1 or more) extracted from a variable-length code string to be decoded. And a plurality of tables respectively corresponding to a plurality of normal states SNx that can be taken by the surplus bit string R in a state where there is a surplus bit string R that cannot be decoded as a result of the previous decoding. , A record represented by an N-bit bit string b in the table corresponding to the initial state SFi, the bit string b is decoded from the i-th bit in a table corresponding to the initial state SFi. And the next state S 'corresponding to the next remaining bit string r, and the remaining state is stored in the table corresponding to the normal state SNx. A decoded data sequence V obtained by decoding a bit sequence obtained by combining the bit sequence R and the bit sequence b, and a next state S 'corresponding to the next remaining bit sequence r. Decryption table. 12. A plurality of initial states in which decoding is performed from the i-th bit (i = 1 to N × 8) of a bit string B of N bytes (N is an integer of 1 or more) extracted from a variable-length code string to be decoded. It has a plurality of tables respectively corresponding to a state SFi and a plurality of normal states SNx in which there is a surplus bit string R that cannot be decoded as a result of the previous decoding and the surplus bit string R can take. The table has a number of records that can be represented by an N × 8-bit binary number, and a record indicated by an N × 8-bit bit string b includes i-bits in the table corresponding to the initial state SFi. In the table corresponding to the normal state SNx, a decoded data string V obtained by decoding from the eyes and a next state S ′ corresponding to the next remaining bit string r are stored. A variable length storing a decoded data string V obtained by decoding a bit string obtained by combining the remainder bit string R and the bit string b, and a next state S 'corresponding to the next remainder bit string r. Code decoding table. 13. A variable-length code storage unit for storing a variable-length code string to be decoded, comprising the decoding table according to claim 10 and storing a data string obtained by decoding. Decoding data storage means, extracting a bit string in a predetermined number of bits or a predetermined byte unit from the variable length code string stored in the variable length code storage means, and sequentially decoding the bit string with reference to the decoding table. And a variable-length code decoding means for storing the obtained decoded data in the decoded data storage means.