JP2001007706A - Decoding device for variable length code - Google Patents

Abstract

PROBLEM TO BE SOLVED: To quicken a decoding speed by reducing the capacity of a lock-up table. SOLUTION: This device is provided with a non-decoded bit stream segmenting part 1 for segmenting a non-decoded variable length code from an input bit stream, a bit stream direct input type lockup table 2, an address/ symbol value one-to-one corresponding type look-up table 3, an address generating part 4 for generating an address for the lookup table 3, and a symbol value code length selecting part 5 for selecting a symbol value and code length outputted from the lookup tables 2 and 3. When the code length of the non- decoded variable length code is short, decoding to the symbol value is excuted by referring to the lookup table 2, and when the code length is long, decoding to the symbol value is executed by referring to the look-up table 3. Thus, the capacity of the look-up table can be reduced, and the decoding speed can be quickened.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a variable-length code decoding device for decoding a variable-length code such as a Huffman code.

[0002]

2. Description of the Related Art When decoding a variable length code such as a Huffman code, a lookup table is generally used. Well-known lookup tables of this type include bitstream direct input lookup tables and address-to-symbol value one-to-one lookup tables.

FIG. 7 is a block diagram showing a schematic configuration of a conventional variable length code decoding apparatus having a bit stream direct input type lookup table. In the apparatus of FIG. 7, the undecoded bit stream cutout unit 51 extracts a bit string of the maximum code length of the variable length code from the first bit of the undecoded bit stream, and uses the extracted bit string as an input address in the lookup table 52. The symbol value and the code length of are read out.

In the case of the apparatus shown in FIG. 7, even if the code length of the symbol value read from the lookup table 52 is shorter than the maximum code length, the lookup table 52 always uses a bit string of the maximum code length as an input address. Must give. Therefore, when the actual code length is shorter than the maximum code length, only a part of the bit string given as an input address to the lookup table 52 is valid.

As described above, the bit stream direct input type lookup table 52 of FIG. 7 can be said to be a many-to-one correspondence type lookup table in which one symbol value corresponds to a plurality of input addresses. For this reason, when the lookup table 52 of FIG. 7 is used, the decoding speed can be increased because the decoding process can be performed in one step, but the capacity of the lookup table 52 is an exponent with respect to the maximum code length of the variable length code. There is a problem that it increases functionally.

FIG. 8 is a block diagram showing a schematic configuration of a conventional variable-length coding apparatus having an address-symbol value one-to-one correspondence type lookup table 53. As shown in FIG. The lookup table 53 in FIG. 8 stores symbol values according to a certain rule, and when a certain address is input,
The corresponding symbol value and code length are output, and the address and symbol value of the lookup table 53 correspond one-to-one.

In the apparatus shown in FIG. 8, first, as a first step, an undecoded bit stream cutout section 51 cuts out a bit string of the maximum code length of a variable length code from the first bit of the undecoded bit stream. Next, as a second step, an address for referencing the lookup table 53 is generated by the address generation unit 54 based on the cut-out bit sequence, and a corresponding symbol value and a corresponding symbol value are generated from the lookup table 53 based on the address. Read the code length.

In the case of the apparatus shown in FIG. 8, since the decoding process is divided into two steps, the decoding speed is slower than that of the apparatus shown in FIG. 7, but the symbol value entries can be eliminated. That is, the number of table entries (table addresses) can be reduced, and the capacity (size) of the lookup table can be reduced.

As described above, the two types of look-up tables have advantages and disadvantages. Therefore, conventionally, one of the look-up tables is used depending on the type of encoded data. .

For example, MPEG2 (Moving Picture coding Ex)
In the case of data compressed (encoded) according to the standard of perts Group 2), since the correspondence between variable length codes and symbol values is defined in advance, the above-described bit stream direct input type lookup table may be used. Was.

In the case of MPEG2, since the maximum code length of the variable length code is 17 bits, a lookup table of 128 k words is required.
Although it is not realistic to implement it inside the EG2 decoder and generally comprises a gate circuit, there is a problem that the number of gates increases. However, in a circuit operating at 27 MHz, one variable-length code could be decoded in one cycle.

On the other hand, JPEG (Joint Photographic image co
In the case where the correspondence between the variable length code and the symbol value can be defined by the user, such as data compressed (encoded) according to the standard of the Ding Experts Group, the above-described address-symbol value one-to-one correspondence type lookup table In many cases, variable-length decoding processing is performed by using.

In the case of using an address-symbol value one-to-one correspondence type lookup table, first, variable length codes are arranged in order of code length, and in the case of the same code length, arranged in ascending order of code value and correspond to the arranged order. Store the symbol value in a lookup table. Next, the I-bit code extracted from the head of the extracted undecoded bit stream is compared with the maximum code value or the minimum code value of the I-bit code, and the code length is detected. Next, the I-bit code extracted from the beginning and a correction value depending on the code length of the code are added, and the addition result is used as an address,
Read the symbol value from the lookup table.

When the correspondence between the variable length code and the symbol value can be defined by the user as in JPEG, it is impossible to form a look-up table with a gate circuit, and usually a storage circuit such as a RAM is used. Is used.

[0015]

However, when performing JPEG baseline processing, one look-up table having 162 words of entries is required for the AC coefficient of one component. It is realistic to do.

However, since the decoding process consists of a plurality of steps as described above, the decoding speed is inevitably lower than in the case of using a bit stream direct input type lookup table.

On the other hand, the maximum code length of the variable length code of the AC coefficient of JPEG is 16 bits, and when a bit stream direct input type look-up table is used, the number of entries is 64 k words. Unrealistic.

The present invention has been made in view of such a point, and an object of the present invention is to provide a variable length code decoding device capable of reducing the capacity of a lookup table and decoding at high speed. To provide.

[0019]

According to a first aspect of the present invention, there is provided a variable length code decoding apparatus for decoding a variable length code included in an input bit stream into symbol values. A first look-up table for storing a symbol value corresponding to each of the address signals using the variable length code as an address signal, address conversion means for converting the variable length code into an address signal for table reference, A second look-up table for storing a symbol value corresponding to the address signal converted by the means; and a first look-up table when the code length of the variable length code included in the input bit stream is equal to or less than a predetermined bit. And decodes the symbol value into a symbol value. If the code length of the variable length code is larger than the predetermined bit, the second rule And a table control means for decoding the symbol values by using the click-up table.

According to the first aspect of the present invention, when the code length of the variable length code is short, decoding is performed using the first look-up table (bit stream direct input type look-up table). Can be decoded at high speed. When the code length of the variable length code is long, decoding is performed using the second look-up table (address-symbol value one-to-one correspondence type look-up table). The capacity can be reduced.

According to the second and seventh aspects of the present invention, variable-length codes are sequentially extracted from an undecoded bit stream and decoded, so that a large number of variable-length codes included in the bit stream can be decoded continuously. .

According to the third aspect of the present invention, the first look-up table, the address conversion means and the second look-up table are operated in parallel, so that the conversion into the symbol value can be performed at high speed. .

According to a fourth aspect of the present invention, in a variable length code decoding apparatus for decoding a variable length code included in an input bit stream into symbol values, a plurality of hierarchized look-up tables and reading from these look-up tables are provided. Table control means for decoding a variable-length code into a symbol value based on the output data, wherein each of the look-up tables uses a different bit string in the variable-length code as an address signal, and Stores the corresponding symbol value.

According to the fourth aspect of the present invention, since a plurality of hierarchized look-up tables are provided, the capacity of the entire look-up table can be greatly reduced as compared with the case where decoding is performed using only one look-up table.

According to the fifth aspect of the present invention, a bit string constituting a part of the variable length code is supplied to each of the hierarchized look-up tables, so that decoding can be performed with a simple configuration.

According to the sixth aspect of the present invention, since the hierarchized look-up tables are operated in parallel, a decoding speed equivalent to the case where the look-up tables are not hierarchized can be obtained.

[0027]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a variable length code decoding apparatus according to the present invention will be specifically described with reference to the drawings. Hereinafter, an example will be described in which the present invention is applied to variable-length code decoding of AC coefficients of a component (for example, a luminance component) in JPEG baseline processing.

(First Embodiment) The first embodiment is characterized in that decoding is performed by combining the above-described bit stream direct input type lookup table and the address-symbol value one-to-one correspondence type lookup table. There is.

FIG. 1 is a block diagram of a first embodiment of a variable-length code decoding apparatus according to the present invention. The device of FIG.
An undecoded bit stream extracting unit (variable length code extracting means) 1 for extracting an undecoded variable length code from an input bit stream; a bit stream direct input type lookup table (first lookup table) 2; A symbol-value one-to-one correspondence type look-up table (second look-up table) 3 and an address generator 4 for generating an address for the look-up table 3
And a symbol value / code length selection unit 5 for selecting a symbol value and a code length output from the lookup tables 2 and 3.

The maximum code length of the JPEG format variable length code is 16
Is a bit. In the present embodiment, code data having a code length equal to or smaller than a predetermined bit (for example, 4 bits) is decoded using the bit stream direct input type look-up table 2, and the code length is changed from the predetermined bit (for example, 4 bits). Is decoded using the address-symbol value one-to-one correspondence type lookup table 3.

FIG. 2 is a diagram showing an example of a variable length code table for AC coefficients of the JPEG baseline processing. Usually, a short code length is assigned to a symbol value having a high frequency of occurrence. The variable length code shown in FIG. 2 has a code length from 2 bits to 16 bits.

In the first embodiment, a variable length code having a code length of 4 bits or less (data from the first data to the sixth data in FIG. 2) is decoded using the bit stream direct input type lookup table 2. For other variable-length codes, decoding is performed using the address-symbol value one-to-one correspondence type lookup table 3.

If the code length is 4 bits or less, 4 bits are required as an input address, and the number of entries in the lookup table 2 is 2 ⁴ = 16 words. On the other hand, the number of entries in the lookup table 3 is less than 162 words. As described above, since the look-up tables 2 and 3 do not have a large capacity,
And can be easily mounted.

Next, the operation of the variable length code decoding apparatus shown in FIG. 1 will be described. The undecoded bit stream cutout unit 1 outputs an undecoded bit string of the maximum code length (16 bits) of the variable length code from the input bit stream. The undecoded bit string is input to both the bit stream direct input type lookup table 2 and the address generator 4.

The lookup table 2 outputs a symbol value and a code length corresponding to a variable length code. Further, the address generator 4 outputs an address for lookup table reference corresponding to the variable length code. The lookup table 3 outputs a symbol value and a code length corresponding to the address from the address generator 4.

Thus, look-up table 2 and
The address generator 4 and the lookup table 3
Operating in parallel, the look-up tables 2 and 3 each output a symbol value and a code length separately.

The symbol value / code length selection unit 5 determines whether one of the symbol value and the code value is output from the lookup tables 2 and 3 depending on whether or not the code length exceeds a threshold value (for example, 4 bits). Select the length and output. Specifically, the output of the lookup table 2 is selected when the code length of the variable length code is 4 bits or less, and the output of the lookup table 3 is selected when the code length is 5 bits or more.

The code length output from the symbol value / code length selection unit 5 is input to the undecoded bit stream cutout unit 1. The undecoded bit stream cutout unit 1 generates a bit string deleted from the input bit stream by the code length output from the symbol value / code length selection unit 5, and again generates the maximum code length of the variable length code from the bit string. Extract and output only.

FIG. 3 incorporates the variable length decoding device of FIG.
It is a block diagram of a JPEG stream reproducing device. JPEG in Figure 3
A JPEG stream is input to the stream player.

The JPEG stream player shown in FIG. 3 is a RAM 11 for temporarily storing a JPEG stream, and a JPEG stream stored in the RAM 11 is decoded into symbol values in FIG.
A variable-length code decoding device 12 having the same configuration as the above, an inverse quantization circuit 13 for inversely quantizing the symbol values, and an IDCT for performing inverse discrete cosine transform (IDCT) of the inversely quantized symbol values to image data Circuit 14 and RAM 1 for storing image data
5 and an output circuit 16 for outputting image data.
The variable-length code decoding device 12, the inverse quantization device 13, and the IDCT circuit 14 in FIG.

As described above, the variable-length code decoding apparatus according to the first embodiment includes the bit stream direct input type lookup table 2 and the address-symbol value one-to-one correspondence type lookup table 3. If the code length of the undecoded variable-length code is short, decoding to a symbol value is performed with reference to the look-up table 2, and if the code length is long, decoding to a symbol value is performed with reference to the lookup table 3. Therefore, a variable-length code having a high frequency and a short code length can be decoded into a symbol value at a high speed, and a variable-length code having a long code length is subjected to address conversion and then to a lookup table. 3, the capacity of the lookup table 3 can be reduced.

The lookup table 3, which requires a longer decoding speed than the lookup table 2, is used only for code data with a low frequency of occurrence, so that the overall decoding speed is not so slow.

In the embodiment described above, when the number of bits is equal to or smaller than a predetermined bit (for example, 4 bits), the lookup table 2
However, the number of bits of the variable length code and the switching of the look-up tables 2 and 3 may be programmably changed.

In the above-described embodiment, the lookup table 2, the address generation unit 4, and the lookup table 3 are operated in parallel. However, as shown in FIG. 1, a table selection unit 6 is provided, and the code length of the variable length code is determined by the table selection unit 6. According to the code length, the variable length code is assigned to only one of the lookup table 2 and the address generation unit 4. May be sent to perform the decoding process.

In the case of the configuration of FIG. 4, only one of the lookup tables 2 and 4 operates, so that the power consumption can be reduced as compared with the configuration of FIG. However, since the table selection unit 6 is additionally required, the number of gates increases and the signal propagation delay time also increases. For this reason,
The apparatus shown in FIG. 4 is effective when the table selection unit 6 is small.

(Second Embodiment) In the second embodiment, the look-up table has a hierarchical structure. FIG.
FIG. 3 is a block diagram showing a schematic configuration of a second embodiment of the variable-length code decoding device according to the present invention. The device of FIG. 5 uses a bit stream direct input type lookup table,
The feature is that it is divided into three layers.

The apparatus shown in FIG. 5 includes an undecoded bit stream extracting section 1, a lookup table 31 of the highest hierarchy (first hierarchy), and a lookup table 32 of the second hierarchy.
, A third-level lookup table 33, and a symbol value / code length selection unit 34.

FIG. 6 is a diagram showing an example of the look-up tables 31 to 33. FIG. 6 (a) corresponds to the look-up table 31, and FIG.
(C) corresponds to the look-up table 33.

As shown in FIG. 5, lookup tables 31 to 33 store not only the symbol values and code lengths corresponding to the variable length codes, but also look up tables 31 to 31.
Stores a first flag for symbol value search, and the lookup table 32 stores a second flag for symbol value search. The first and second flags indicate whether the final symbol value is output from any of the three types of lookup tables 31 to 33.

Assuming that a bit string of a variable length code when the lookup table is not divided is a [15: 0], a bit string a [15: 9] is input to the lookup table 31 of the first hierarchy as an input address. The bit string a [10: 4] is input as an input address to the second hierarchy lookup table 32, and the bit string a [6: 0] is input to the third hierarchy lookup table 33 as an input address.

The lookup table 31 of the first hierarchy is
When the bit string a [15: 9] is (0000000) to (1111011), the corresponding symbol value and code length are output, and Table 1 is set as the first flag. Since the second flag is determined by the value of a [10: 4], the second flag cannot be determined only by a [15: 9], so that the state is undefined. Here, Table 1 is a lookup table 31 of the first hierarchy.
To refer to.

The bit string a [15: 9] is (1111100) to (1111
11111), the symbol value is undecided and the first
Set Table2 as a flag of. Here, Table 2 indicates that the lookup table 32 of the second hierarchy is referred to.

On the other hand, the lookup table 3 of the second hierarchy
2 outputs a symbol value and a code length when the bit string a [15:11] is (11111) when the bit string a [10: 4] is (0000000) to (1110111). At this time, Table 2 is set as the second flag. Also, bit string a [10: 4] is (1
In the case of (111000) to (1111111), the symbol value is undecided, and Table 3 is set as the second flag. Here, Table3 is a lookup table 3 of the third hierarchy.
3 is referred to.

On the other hand, the lookup table 3 of the third hierarchy
3 outputs a symbol value when a [15: 7] is assumed to be (111111111) when the bit string a [6: 0] is (0000000) to (1111110).

The symbol value / code length selector 34 shown in FIG.
When the first flag is Table1, the symbol value of the first-level lookup table 31 and the code length of the variable-length code corresponding to the symbol value are output as the final result.
When the first and second flags are both Table2, the symbol value and the code length of the lookup table 32 of the second layer are output. Also, the first flag is Table2,
When the second flag is Table 3, the symbol value and the code length of the lookup table 33 of the third hierarchy are output.

As described above, in the second embodiment, the bit stream input type look-up tables 31 to 33 including a plurality of layers are provided, and the respective look-up tables 31 to 33 are provided.
33, the bit strings constituting a part of the undecoded variable-length code are given as addresses, and the lookup tables 31 to 33 are operated in parallel. Therefore, the total number of entries in these lookup tables can be significantly reduced. Thus, the circuit size of the look-up table can be reduced.

Specifically, the total number of entries in the lookup table before division is 2 ¹⁶ = 65536 words, whereas the total number of entries in the hierarchical lookup tables 31 to 33 in FIG. 384 words.
Therefore, the look-up tables 31 to 33 are assumed to be R
Even with AM configuration, the number of gates does not increase so much.
It can be easily realized.

The symbol value / code length selection unit 3 shown in FIG.
4 is not so complicated, the same decoding speed as before the lookup table is divided can be obtained.

In the above-described second embodiment, an example in which the look-up table is divided into three layers has been described, but the number of layers is not particularly limited. Also, it is possible to arbitrarily change the setting of which layer of the variable length code is assigned to which hierarchy.

Further, the first and second embodiments may be implemented in combination. That is, the bit stream input type lookup table of FIG. 1 may be divided into a plurality of layers.

In each of the embodiments described above, the JPEG format and the MPEG
Although the decoding of data encoded in two formats has been described, the present invention is equally applicable to any type of variable length code.

[0062]

As described above in detail, according to the present invention, two types of look-up tables are provided, and either one of the look-up tables is selected according to the code length of the variable length code, and the symbol value is selected. In this case, the decoding speed can be increased, and the capacity of the entire look-up table can be reduced.

Further, according to the present invention, since a plurality of hierarchized look-up tables are provided and decoding into symbol values is performed, the capacity of the look-up table can be greatly reduced and the number of divisions can be reduced to an appropriate number. By setting to, it is possible to prevent a decrease in decoding speed.

[Brief description of the drawings]

FIG. 1 is a block diagram of a first embodiment of a variable-length code decoding apparatus according to the present invention.

FIG. 2 is a diagram showing an example of a variable-length code table for AC coefficients of JPEG baseline processing.

FIG. 3 is a block diagram of a JPEG stream reproducer incorporating the variable length decoding device of FIG. 1;

FIG. 4 is a block diagram obtained by adding a table selection unit to FIG. 1;

FIG. 5 is a block diagram showing a schematic configuration of a second embodiment of the variable-length code decoding apparatus according to the present invention.

FIG. 6 is a diagram showing an example of lookup tables 31 to 33.

FIG. 7 is a block diagram showing a schematic configuration of a conventional variable-length coding device having a bit stream direct input type lookup table.

FIG. 8 is a block diagram showing a schematic configuration of a conventional variable length encoding device having an address-symbol value one-to-one correspondence type lookup table 53;

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 Undecoded bit stream extraction part 2 Bit stream direct input type lookup table 3 Address-symbol value one-to-one correspondence type lookup table 4 Address generation part 5 Symbol value / code length selection part 6 Table selection part

Claims (7)

[Claims] 1. A variable length code decoding device for decoding a variable length code included in an input bit stream into a symbol value, wherein a symbol value corresponding to each of the address signals is stored using an undecoded variable length code as an address signal. A first look-up table, an address converting means for converting a variable length code into an address signal for table reference, and a second look-up for storing a symbol value corresponding to the address signal converted by the address converting means. If the code length of the variable length code included in the input bit stream is equal to or smaller than a predetermined bit, decoding into a symbol value is performed using the first lookup table, and the code length of the variable length code is set to the predetermined length. Table control for decoding to symbol values using the second look-up table if larger than bits Means for decoding a variable-length code. 2. A variable length code extracting means for extracting a bit string of a maximum code length of a variable length code from an undecoded bit stream, wherein the variable length code extracting means, when extracting a bit string next time, From the first bit of the previously extracted bit string, the bit string of the code length of the variable length code corresponding to the symbol value read from the first or second look-up table is removed, and the remaining maximum length of the variable length code is newly obtained from the rest. The variable-length code decoding apparatus according to claim 1, wherein a bit string corresponding to the code length is cut out. 3. The apparatus according to claim 1, wherein said first look-up table, said address conversion means and said second look-up table are operated in parallel.
Or the variable-length code decoding device according to 2. 4. A variable length code decoding apparatus for decoding a variable length code included in an input bit stream into a symbol value, comprising: a plurality of hierarchical look-up tables; and data read from these look-up tables. Table control means for decoding a variable-length code into a symbol value based on each of the look-up tables. A variable-length code decoding device for storing. 5. The plurality of lookup tables,
The uppermost look-up table stores bit values corresponding to predetermined bits from the first bit of the variable length code as address signals, and stores symbol values corresponding to the respective address signals. In the look-up table, symbol values corresponding to the respective address signals are stored by using a bit string of a predetermined bit from the last bit of the variable-length code as an address signal. 5. The variable length code decoding apparatus according to claim 4, wherein a bit string close to the first bit is supplied as an address signal. 6. The variable length code decoding apparatus according to claim 4, wherein reading from the plurality of lookup tables is performed in parallel. 7. A variable-length code extracting means for extracting a bit string of the maximum code length of a variable-length code from an undecoded bit stream, wherein the variable-length code extracting means, when extracting the next bit string, From the first bit of the previously extracted bit string, a bit string corresponding to the code length of the variable length code corresponding to the symbol value output from any of the plurality of look-up tables is removed, and the maximum of the variable length code is newly added from the rest. 7. A method according to claim 4, wherein a bit string corresponding to the code length is cut out.
The variable-length code decoding device according to any one of the above.