JPH05191652A - Decoder for variable length code of picture data - Google Patents

Abstract

PURPOSE:To decode a variable length code at a high speed with respect to the decoder for a variable length code of picture data. CONSTITUTION:A code string for a prescribed length outputted from a code shifter 181 is outputted to a decoding table 185 and a group index segmentation section 186. A decoding table 185 decodes a group number IDX and a zero run length RM1 to which an effective coefficient of a DC difference value or an AC coefficient belongs, and outputs the total code length CLEN between the decoded code and the additional bits with respect to the code to the code shifter 181 and the group index segmentation section 186. The group index segmentation section 186 extracts the additional bits based on the IDX and the CLEN and the code shifter 181 outputs the code string of a prescribed length in succession to the code decoded based on the CLEN to a decoding table 185 and the group index segmentation section 186.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image data restoration device for decoding a variable length code of a multi-valued image coded with high efficiency using orthogonal exchange.

[0002]

2. Description of the Related Art In the fields of video conferencing / telephone calls, color still images, color moving images, etc., image data having an order of magnitude greater than numerical data, particularly halftone images and color image data, is stored. Furthermore, since it is necessary to transmit those image data at high speed and high quality, it is essential to perform a process of highly efficiently encoding the gradation values of pixels and the like.

An adaptive discrete cosine transform coding method is known as a highly efficient compression method for such image data. This adaptive discrete cosine transform coding system (Adaptive Discrete Cosine Transform, hereinafter abbreviated as ADCT) divides an image into a plurality of blocks each including, for example, 8 × 8 pixels, and the image signal of each block is a two-dimensional discrete cosine. Coefficient of spatial frequency distribution (two-dimensional DCT coefficient) by transformation (hereinafter referred to as two-dimensional DCT)
To a Huffman table in which the two-dimensional DCT coefficient in each block obtained by the conversion is quantized by a threshold value adapted to the visual sense, and the quantized coefficient obtained by the quantization is statistically obtained. Is encoded by.

Next, the ADCT encoding operation will be described in more detail with reference to the block diagram of the ADCT conversion circuit shown in FIG. First, the original image is 8 × as shown in FIG.
The image signal of each block is divided into a plurality of blocks each having 8 pixels and sequentially input to the DCT conversion unit 24. In the DCT transform unit 24, the input image signal of one block is converted into a two-dimensional DCT and subjected to orthogonal transform to transform into 8 × 8 two-dimensional DCT coefficients of the spatial frequency distribution shown in FIG. Output to.

FIG. 9 shows an example of the block configuration of the above-mentioned two-dimensional DCT conversion section. In the two-dimensional DCT conversion unit having the configuration shown in FIG. 9, an input image signal of one block (8 × 8 pixels) is subjected to one-dimensional DCT conversion in the one-dimensional DCT conversion unit 30 and then in the transposition unit 31 in the block. After the rows and columns of the coefficients obtained by the above one-dimensional DCT transform are transposed (transposed), they are output to the one-dimensional DCT transform unit 32. Next, in the one-dimensional DCT conversion unit 32, the one-dimensional DCT conversion unit 30
The same one-dimensional DCT transform is performed and output to the transposing unit 33. Then, the transposing unit 33 further performs the same transposing process as the transposing unit 31 and outputs the one-block two-dimensional DCT coefficient obtained by the transposing process to the linear quantization 25 shown in FIG. By applying such processing to all the blocks of the image data, the image signals of all the blocks become 2
It is converted into a dimensional DCT coefficient.

The linear quantizing unit 25 uses the quantizing matrix 28 configured by a threshold value as shown in FIG. 10, for example, to determine the two-dimensional DCT coefficient for each block input as described above by visual experiments. Linearly quantize. The result obtained by this linear quantization is shown in FIG. As shown in the figure, the two-dimensional DCT coefficient having a value smaller than the threshold is 0.
Therefore, a quantized coefficient having values of only the DC component (= 5) and a few AC components is generated.

Next, the quantized coefficients arranged two-dimensionally as shown in FIG. 11 are a one-dimensional numerical sequence by a scan called a zigzag scan for scanning the quantized coefficients in the order of the numbers shown in FIG. And is input to the variable length encoding unit 26. The variable length coding unit 26 uses D at the beginning of each block.
C component (DC coefficient) and DC component (DC
Coefficient) is subjected to variable length coding. For the AC component, the value of the effective coefficient (coefficient whose value is not 0) (hereinafter,
The index and the invalid coefficient up to that point (value is 0)
Variable length coding for each block in combination with the run length (hereinafter referred to as the zero run length) of the coefficient. In this variable length coding, each component of DC and AC is coded using a code table 27 composed of a Huffman table created based on the statistic of each image, and the code obtained by the coding The data is sequentially output to the outside.

On the other hand, the coded data coded as described above is restored to an image by the ADCT restoration circuit shown in FIG. Next, an image restoration method performed by the ADCT restoration circuit will be described.

Code data input in block units is
First, it is input to the variable length coding unit 41. Variable length decoding unit 41
Then, using the decoding table 42 that is a table that is the inverse of the Huffman table of the code table 27, the input code data is decoded into fixed-length data composed of the index and the zero run length, and the dequantization unit Output to 43. The inverse quantizer 43 inversely quantizes the input quantized coefficient by multiplying each quantized coefficient of the input fixed-length data by the corresponding threshold value of the quantization matrix 46 to generate a two-dimensional D
Decode to CT coefficient, and the decoded two-dimensional DCT coefficient to 2
The data is output to the dimensional inverse DCT conversion unit 44. The two-dimensional inverse DCT transform unit 44 uses the input two-dimensional DCT coefficient to perform the inverse 2
Dimensional DCT conversion is performed to restore the image signal.

FIG. 14 shows an example of the configuration of the two-dimensional inverse DCT conversion section. In the two-dimensional inverse DCT transform unit, the input two-dimensional DCT coefficient is converted into the one-dimensional inverse DCT transform unit 51.
Is one-dimensional inverse DCT-transformed and output to the transposing unit 52.
The transposing unit 52 interchanges the rows and columns of the coefficients in one block and outputs the transposed coefficients to the one-dimensional inverse DCT transform unit 53. 1-dimensional inverse DC
The T conversion unit 53 performs one-dimensional inverse DCT conversion on the input transposed coefficient again, and outputs it to the transposition unit 54. Transposed part 54
Outputs a signal obtained by replacing the rows and columns of the coefficients in one block again like the transposing unit 52, and thereby the image signal is restored.

As described above, in the conventional ADCT method, the quantized coefficient at the time of encoding the image signal is obtained by quantizing the two-dimensional DCT coefficient with the quantized threshold value. FIG. 15 shows a configuration example of a conventional linear quantization circuit.

In the figure, 2 input from the terminal 60
The dimensional DCT coefficient is temporarily held in the DCT coefficient input unit 63. The DCT coefficient input unit 63 sequentially outputs the input DCT coefficients for each pixel to the division unit 69 according to the data read signal (RED) from the timing control unit 61. The quantization threshold holding unit 62 also sequentially outputs the quantization threshold corresponding to the pixel held in the DCT coefficient input unit to the division unit 69 according to the data read signal (RED) from the timing control 61.

The division unit 69 receives the DCT of each pixel input.
The coefficient is quantized by dividing it by the input quantization threshold, and the result of the division is output to the latch unit 64 as a quantized coefficient (QUD). The timing controller 61 calculates the access time of the divider 69 and generates a latch signal (LAT) for data latch in the latch 64. The quantized coefficient is latched in the latch section 64 by this latch signal (LAT), and the quantized coefficient is output from the terminal 70 to the variable length coding section 26 shown in FIG. Timing control unit 6
When the quantization of the two-dimensional DCT coefficient for one pixel is completed, 1 instructs the DCT coefficient input unit 63 and the quantization threshold holding unit 62 to read the two-dimensional DCT coefficient and the quantization threshold of the next pixel. , Quantizes the two-dimensional DCT coefficient of the next pixel.

In this way, the two-dimensional DCT coefficient held in the DCT coefficient input unit 63 is read out in pixel units,
The read two-dimensional DCT coefficient is stored in the quantization threshold holding unit 6
The process of dividing by the quantization threshold value held in 2 and outputting the division result as the quantization coefficient of the target pixel is repeated for each screen for each pixel for one screen.
The two-dimensional DCT coefficient for the screen is quantized.

Then, the obtained quantized coefficient is subjected to a zigzag scan (see FIG. 12), and for the DC component,
The DC component of the previous block and the difference value are obtained, and the difference value (D
According to the table shown in FIG. 16, the C difference value is divided into a group number (SSSS) and an additional bit (group index: GDL) indicating the position of the difference value in the group indicated by the group number (SSSS). Then, the group number (SSSS) is variable-length coded using a one-dimensional Huffman code table, and the additional bit (the number of bits is the SSSS is added after the Huffman code.
(Equal to the value of) is generated. On the other hand, the AC component is divided into a plurality of sets consisting of a combination of the number of consecutive “0” s (zero run length: NNNN) and a non-zero coefficient (effective coefficient), and further divided into effective coefficients (index) of each set. On the other hand, using the table shown in FIG. 17, the same grouping as that applied to the difference value of the (DC coefficient) of the DC component is performed, and the group number (SSSS) to which the effective coefficient belongs and the group thereof. A group index (GI) indicating a position within the group is obtained. Then, two-dimensional Huffman coding is performed with a combination of the zero run length (NNNN) and the group number (SSSS) to generate a variable length code, and the variable length code indicates the value of the group index (GI). A variable length code is generated by adding additional bits.

As described above, in the variable length coding of multi-valued image data using the two-dimensional ADCT, the DC component (DC
Coefficient), the DC component (DC coefficient) of the previous block
And the difference value is encoded by a one-dimensional Huffman coding into a variable-length code in which a group index [GDL] is added to a code [CGDL] indicating a group number (SSSS), and the AC component (AC coefficient) is 2 Variable length coding is performed by the combination of the zero run length (NNNN) and the group number (SSSS) by the dimension Huffman coding, and the variable length coding [CRMi /
A variable length code in which the group index [CGI _i ] is added to CIi] is generated.

FIG. 18 shows the configuration of a variable length code string of one block generated by the variable length coding as described above. Next, the variable length decoding unit 41 of the ADCT restoration circuit shown in FIG.
FIG. 19 shows the configuration of a conventional variable-length decoding circuit including a decoding table (decoding table) 42 and a decoding table.

In this variable length decoding circuit, the code shifter 81 starts with the variable length code string of the first block having the configuration shown in FIG. The code string [CDL] is cut out and output to the latch unit 84 and the reverse grouping table 86. Then, the above code [CDL] latched by the latch section 84.
The variable-length code string including "1" is decoded by the decoding table 85 to the group number [IDX] of the group to which the DC difference value DL belongs, and then is held in the latch unit 88 and latched by the latch unit 88.
Is output to the code shifter 81 via. At this time, the decoding table 85 outputs the code length (bit length) CLEN of the code [CGDL] to the code shifter 81.

The code shifter 81 is based on the group number [IDX] and the code length CLEN.
Code of group index of C difference value DL [CGD
L] is output to the reverse grouping table 86 (group number [IDX] = group index [CGD]
L] bit length).

The reverse grouping table 86 decodes and outputs the DC difference value [DL] based on the code [CGDL] and the group number [IDX] input via the latch unit 88.

Next, the code shifter 81 controls the variable length code [CRM] of the first index (effective coefficient) of the AC component.
The variable length code string having a predetermined length including ₁ / CI ₁ ] is output to the decoding table 85 via the latch unit 84. The decoding table 85 uses the variable length code [C
RM ₁ / CI ₁ ], a variable length code string including a zero run length R
While decoding M (NNNN), the code [CRM ₁
/ CI ₁ ], and outputs the code length (bit length) CLEN to the code shifter 81. At the same time, the decoding table 85 also decodes the number (SSSS) IDX of the group to which the first index (effective coefficient) I _{1 of} the AC component belongs, and codes the group number (SSSS) IDX via the latch unit 88. Shifter 81 and reverse grouping table 8
Output to 6.

The code shifter 81 then determines the code length CL.
On the basis of EN and the group number IDX, the first code [CRM ₁
/ CI ₁ ] additional bit [CGI ₁ ] is output to the inverse grouping table 86 (value of group number IDX = bit length of code bit [CGI ₁ ]).

The inverse grouping table 86 decodes and outputs the first index (effective coefficient) I ₁ of the AC component based on the additional bit [CGI ₁ ] and the group number IDX.

Thereafter, the variable length decoding circuit similarly performs the remaining code string of one block, that is, the code [CRM ₂ / CI ₂ ].
And code [CGI ₂ ], code [CRM ₃ / CI ₃ ], code [CGI ₃ ], ... Code [CRM _N / C
I _N ], and the variable length code sequence of the code [CGI _N ], sequentially [RM ₂ , I ₂ ], [RM ₃ , I ₃ ], ... [RM
_N , I _N ], and the first one-block variable-length code string is decoded.

Then, the variable-length code circuit repeats the above-described decoding processing of the variable-length code string in block units for the remaining blocks of one screen, and completes the decoding of the variable-length code string of one screen. To do.

[0026]

As described above, in the past, when decoding a variable-length code string encoded by two-dimensional ADCT, the difference value D of the DC component (DC coefficient)
For L, the group number DL and its additional bits (group index) GDL are used as AC components (AC coefficients).
For the effective coefficient (index) group number I _i (i = 1, 2, ..., N) and zero run length (NN
NN) RM _i , and the group number I _i (i = 1, 1)
, ..., N) additional bits (group index) GI _i (i = 1, 2, ..., N) are serially decoded.

Therefore, there is a problem that the decoding processing speed is slow. The present invention, for each block obtained by dividing the original image into a plurality of blocks composed of a plurality of pixels,
The transform coefficients obtained by orthogonally transforming the gradation values of multiple pixels in the block are quantized, the quantized coefficients obtained as a result are grouped, and then the encoded variable-length code is decoded. The purpose is to enable high-speed operation.

[0028]

FIG. 1 is a block diagram showing the principle of the present invention. The present invention, for each block obtained by dividing the original image into a plurality of blocks composed of a plurality of pixels, the transform coefficient obtained by orthogonally transforming the gradation values of the plurality of pixels in the block is quantized, It is premised on a decoding device for a variable length code of image data, which performs grouping of the obtained quantized coefficients and then decodes the encoded variable length code.

The first aspect of the present invention comprises the following means. The code string extracting means 1 cuts out and outputs the code string of the variable length code of each of the input blocks in units of a predetermined length.

When the code string of the variable-length code is a code string coded by the adaptive discrete cosine transform coding method using Huffman coding, the extracted code string has a DC coefficient (DC coefficient). Is the sign of the difference value of the DC coefficient (DC difference value), its additional bit, and the AC coefficient (A
The C coefficient includes a code of a combination of the number of the group to which the AC coefficient belongs (SSSS) and the zero run length (NNNN) and its additional bit.

The decoding means 2 decodes, from the code string output from the code string extracting means 1, the decoding of the group to which the difference value of the DC coefficient belongs or the decoding of the number of the group to which the effective coefficient of the AC component belongs and the effective coefficient thereof. Decoding the run length of the invalid coefficient of, the total code length of the code of the number of the group to which the differential value of the decoded DC coefficient belongs and the additional bit following the code, and the bit length of the additional bit, or the decoded bit The total code length of the code of the combination of the group number to which the effective coefficient of the AC coefficient belongs and the run length of the decoded invalid coefficient and the additional bit following the code, and the bit length of the additional bit are output.

The additional bit cutout means 3 extracts the code string output from the code string extraction means 1 from the decoding means 2
Based on the total code length and the bit length of the additional bit output from, the number of the additional bit following the group number to which the differential value of the DC coefficient belongs or the group number to which the effective coefficient of the AC coefficient belongs and the invalid coefficient The additional bit following the code of the combination with the run length of is cut out.

Then, the code extracting means 1 determines the head of the code string to be output next, based on the total code length output from the decoding means 2 after the second extraction of the code string.

Next, the second invention according to claim 2 further comprises the following restoring means 4 in addition to the code string extracting means 1, the decoding means 2 and the additional bit extracting means 3. The restoring means 4 is based on the number of the group to which the difference value of the DC coefficient output from the decoding means 2 belongs or the number of the group to which the effective coefficient of the AC coefficient belongs and the additional bit output from the additional bit extracting means 3. The difference value of the DC coefficient and the effective coefficient of the AC component are decoded.

Further, the variable length code of the image data decoded by the decoding device of the first invention and the second invention is obtained by adaptive discrete cosine transform coding as described in claim 3, for example. May be

[0036]

The operation of the first invention is as follows. When the variable length code string from the outside is input, the code string cutout unit 1 extracts a code having a predetermined bit length from the head of the variable length code string and outputs it to the decoding unit 2 and the additional bit cutout unit 3.

The decoding means 2 decodes the number (SSSS) of the group to which the difference value of the DC coefficient belongs from the input code string, and obtains the bit length of the additional bit following the decoded code from the decoded value. , And outputs the bit length to the additional bit extracting means 3. Further, the decoding means 2 outputs the total code length of the code of the number of the group to which the differential value of the DC coefficient subjected to the above decoding and the additional bit and the additional bit length to the code string extracting means 1 at the same time as the above operation. To do.

When the total code length (total bit length) and the additional bit length are input, the additional bit cutting-out means 3 extracts the additional bits from the code string input from the code string extracting means 1 based on them. (A value indicating the position of the difference value of the DC coefficient in the group) is cut out.

As described above, the code extracting means 1 outputs the code string having the predetermined bit length once from the beginning of the input code string, and the decoding means 2 decodes the number of the group to which the differential value of the DC coefficient belongs. At the same time, the additional bit cutting-out means 3 cuts out an additional bit which is information indicating the position of the difference value of the DC coefficient in the group.

Then, the code string extracting means 1 is based on the total code length input from the decoding means 2 and has a predetermined bit length following the decoded code from the input variable length code string. A code of a combination of a code of a group, that is, a group number (SSSS) to which the first effective coefficient of the AC coefficient belongs and a zero run length (NNNN) preceding the string and an additional bit following the code (the first coefficient of the AC coefficient). A code string including a value indicating the position of the effective coefficient in the group) is output to the decoding unit 2 and the additional bit cutting unit 3.

The decoding means 2 decodes the number (SSSS) of the group to which the first effective coefficient of the AC coefficient belongs and the zero run length from the input code string, and from the value of the group number (SSSS). The bit length of the additional bit following the code of the group number (SSSS) is obtained, and the additional bit length is output to the additional bit extracting means 3. In addition, the decoding means 2 performs the above decoding at the same time as the above operation (the group number to which the first effective coefficient of the AC coefficient belongs (SSSS) and the zero run length (NNNN)).
The total code length of the code (combined with the code) and the additional bit is output to the code string extracting unit 1 and the additional bit extracting unit 3.

When the total code length and the additional bit length are input, the additional bit cutting means 3 extracts the additional bit from the code string input from the code string cutting means 1 based on them.

In this way, the code extracting means 1 is based on the total code length input from the decoding means 2 and is the number of the group to which the first effective coefficient of the AC coefficient to be decoded next from the input variable length code string belongs. (SSSS) and a code of a combination of a zero run length (NNNN) preceding the effective coefficient and a code string including additional bits for the code can be extracted and output to the decoding unit 2 and the additional bit extracting unit 3, and the decoding unit 2 and the additional bit cut-out means 3 decode the group number (SSSS) and the zero run length (NNNN) of the group to which the first effective coefficient of the AC coefficient belongs from the variable length code string and the group of the first effective coefficient. It is possible to cut out an additional bit that indicates the position within the area. Thereafter, the same operation is repeated, and the AC coefficient is 2
For the second and subsequent effective coefficients, the number of the group to which it belongs (SSSS) and the corresponding zero run length (NNNN)
Decoding and clipping of the corresponding additional bits are performed by the decoding means 2 and the additional bit clipping means 3 each time the code string is output from the code clipping means 1.

Therefore, the decoding of the group number to which the difference value of the DC coefficient belongs, the extraction of the additional bit indicating the position in the group, and the zero run length corresponding to the group number (SSSS) to which the effective coefficient of the AC coefficient belongs (NNNN)
Can be decoded and additional bits indicating the position within the group can be cut out in parallel.

Further, in the second invention, in addition to the operation of the first invention described above, the restoring means 4 adds the group number (SSSS) of the difference value of the DC coefficient input from the decoding means 2. The difference value of the DC coefficient is restored based on the additional bits input from the bit extracting means 3. In addition, the restoration unit 4 sets the effective coefficient of the AC coefficient to a variable length code based on the group number (SSSS) to which the effective coefficient of the AC coefficient input from the decoding unit 2 belongs and the additional bit input from the additional bit cutout unit 3. Restores sequentially in the order of the columns.

Therefore, the variable-length code can be decoded at high speed as in the above-mentioned first invention, and the effective coefficient of the AC coefficient can be restored at high speed.

[0047]

Embodiments of the present invention will be described below with reference to the drawings. As the code of the image data handled in this embodiment, as described above, the original image is divided into a plurality of blocks composed of 8 × 8 pixels, and two-dimensional adaptive discrete cosine transform (two-dimensional ADCT) is performed for each block. It is a variable length code obtained by further performing quantization, grouping, and Huffman coding on the transform coefficient obtained as a result.

16 and 17, DC difference values (difference values of DC coefficients) before the Huffman coding are performed, respectively.
Tables 100 and 2 used to group AC coefficient values
It is a figure which shows 00.

The DC difference value of each block is shown in Table 100 above.
Group numbers (1 to 11) (SS
SS). In addition, the effective coefficient (non-zero AC coefficient) of the AC coefficient of each block is classified into groups having group numbers of "1" to "16" using the above table 200.

Then, the DC difference value is variable-length coded into a Huffman code [CDL] having a minimum of 2 bits and a maximum of 11 bits according to the number (SSSS) of the group to which the DC difference value belongs, and further the DC difference with respect to the Huffman code [CDL]. A group index [CGDL] indicating position information of the value “L” in the belonging group is added, and variable length coding of the DC difference value is performed (see FIG. 18).

Then, in the variable length encoding of the DC difference value, the relationship of the group number (SSSS) value = the bit length of the group index [CGDL] (1) is established (see FIG. 16).

Therefore, the range of the bit length of the variable length code consisting of the variable length code [CDL] and the group index [CGDL] of the group number SSSS to which the DC difference value belongs is from 2 bits to 22 (= 11 + 11) bits. (When the group number SSSS = "0", [CD
L] = “00” is variable-length coded and [CGDL] is not added).

The AC coefficient (AC component) is variable-length coded using the run length (NNNN) of the invalid coefficient and the value of the effective coefficient (coefficient whose value is not "0"). That is,
Run length of invalid coefficient (NNNN) RM _i (i = 1, 2, ...
.., N) and the number of the group to which the effective coefficient belongs (SSS
S) I _i (i = 1, 2, ..., N) in combination,
Huffman code of minimum 2 bits and maximum 16 bits [CRM
_i / CI _i ] is generated and its Huffman code [CRM _i
/ I _i ], the group index [CGI _i ] (i = 1, 2,
, N) are added and variable length coding of the AC component is performed (see FIG. 17).

In the variable length code of the AC component, the relationship of the group number (SSSS) value = the bit length of the group index [CGI _i ] (i = 1, 2, ..., N) is established.

Therefore, the code [CRM _i / CI _i ] (i = 1, 1) generated for each effective coefficient of each of the N AC coefficients.
2, ..., N) and code [CGI _i ] (i = 1, 2, ...
.., N) and the bit length range of the variable length code is 3
Bits to 32 bits.

Next, FIG. 2 shows the configuration of a variable length decoding circuit which is an embodiment of the present invention. The code shifter 181 inputs the variable-length coded code as described above transmitted from an external circuit (not shown), for example, in 1-block units, and stores it in an internal buffer (not shown). Store. Then, the latch portion 182 has the code [G
DL], [CRM ₁ ], ... 22 bits starting with [CRM _N ] (when starting with code (GDL)) or 32
The variable-length code string CODE ₂ of bits (starting with code [CRM ₁ ], ... [CRM _N ]) is sequentially output. In addition, the code shifter 181 applies codes [GDL], [CRM ₁ ], ... [CRM] to the latch unit 184.
_A 16-bit variable length code string CODE ₁ starting with _N ] is output.

The latch unit 182 latches the 22-bit or 32-bit variable length code string output from the code shifter 181 by a latch signal applied from a timing control unit (not shown), and a group index cutout unit. Output to 186.

The latch section 184 outputs the 16-bit variable length code string CODE ₁ output from the code shifter 181.
Is latched by the latch signal applied from the timing control unit and output to the decoding table 185.

The decoding table 185 shows the DC difference value (DC
Coefficient difference value) Group number to which "L" belongs (SSS
The code [CDL] of S) is the group number (SSSS)
A decoding table for a DC difference value for decoding into a value DL indicating [0 run length RM _i (NNNN) / AC coefficient I _i ].
Codes (i = 1, 2, ..., N) [CRM _i / C
I _i ], the original zero run length RM _i (i = 1, 2, ...,
N) and the group number to which the effective coefficient of the AC coefficient belongs (S
SSS) I _i (i = 1, 2, ..., N) has a built-in decoding table for AC coefficients.

The decoding table for the DC difference value is the group number to which the decoded DC difference value "L" belongs (S
A value obtained by adding the bit length of the group index [CGDL] of the DC difference value “L” to the bit length of the code [CDL] of the SSS) DL, that is, the code [CDL] and the code [CG
Information for outputting the total code length CLEN of [DL] is also stored.

In the total code length CLEN, the code length (bit length) of the group index CGDL, which is an additional bit, is equal to the value of the group number (SSSS), that is, the decoded value (SSSS) of [CDL] = [CGDL]. ] The bit length of [] is obtained by using the fact that (3) holds.

FIG. 3 shows the difference in structure between the conventional decoding table for DC difference values and the decoding table for DC difference values used in this embodiment. The decoding table for the DC difference value is composed of a ROM (leat only memory) to which a 16-bit address signal is input via the latch circuit 184. Therefore, in the decoding table for the DC difference value, as shown in FIG. 3, “00”, “010”, ...
., DC difference value "L" of minimum 2 bits and maximum 9 bits that can take 12 types of values up to "111111110"
Of the group to which the belongs (SSSS) DL code CDL
16-bit variable-length code (CGD
L, GDL, CRM ₁ etc. may be included) 16
It is input as a bit address signal. Note that, in the figure, the bits indicated by "x" are bits other than the code [CDL], and their values may be undefined. In the figure, the code "0" included in the 16-bit address signal
0 ”is the group number (SSSS) with the highest appearance probability
Is the code [CDL] of "3", and the code "010" is the code [CDL] of which the group number (SSSS) with the next highest occurrence probability is "0". In the same manner, the group numbers (SSSS) are "1", "2", "4", ...
.., the value [01] of the code [CDL] of "10" or "11"
1 ”,“ 100 ”,“ 110 ”...,“ 111111 ”
"10" and "111111110" are shown on the lower bit side of the 16-bit address signal.

Since the Huffman code is a compact code that can be instantaneously decoded (uniquely decodable) as is well known, a 16-bit variable length code having the code [CDL] as the lower bit is sent from the code shifter 181 to the address signal. By inputting as, the decoded data IDX, which is a binary value equal to the value of the group number (SSSS) DL to which the DC difference value “L” belongs, is converted into the group index cutout unit 186 and the reverse grouping table as shown in FIG. It is output to 187.

By the way, in the decoding table for the DC difference value of the present embodiment, the fact that the relationship of the above expression (3) is established is utilized, as shown in FIG. The code length of the CDL] is not output as the code length CLEN as it is, but based on the group number (SSSS) to which the DC difference value DL obtained by the decoding belongs, a code CDL equal to the value of the group number (SSSS) is obtained. A value obtained by adding the bit lengths of the subsequent additional bits CGDL is output as the total code length CLEN to the code shifter 181 and the group index cutout unit 186.

Next, FIG. 4 shows the difference between the decoding table for the effective coefficient of the AC coefficient (AC component) of the present embodiment, which is built in the decoding table 185, and the conventional decoding table for the effective coefficient of the AC coefficient. Show.

Decoding table for effective coefficients of this AC coefficient
Also from the ROM to which the 16-bit address signal is input
Variable length code [CRM_i/ CI_i] (I =
1, 2, ..., N) on the lower bit side or 16
All bits are variable length code [CRM _i/ CI_i] (I =
1, 2, ..., N) 16-bit address signal
Is input via the latch unit 184.

This variable length [CRM_i/ CI_i] (I =
1, 2, ..., N) were also obtained by Huffman coding
The decoding table for the effective coefficient of the AC coefficient
From the code shifter 181 to the variable length code [CRM_i/ C
I_i] Of 16 bits including (i = 1, 2, ..., N)
When the address signal is input via the latch unit 184,
Immediately the above variable length code [CRM_i/ CI_i]
Zero run length (NNNN) RM_iAnd the effective coefficient of the AC coefficient is
Value IDX (I of the group number to which it belongs (SSSS) _i: I
, 1, 2, ..., N) are immediately decoded, and IDX is
Loop index cutout portion 186 and reverse glue pin
It is output to the table 187. In addition, the zero run length (NN
NN) RM_iIs an invalid coefficient restoration circuit not specifically shown.
Is output.

The decoding table for the effective coefficient of the AC coefficient is also different from the conventional decoding table for the effective coefficient of the AC coefficient, like the above-mentioned decoding table for the DC difference value.
Based on the value I _i of the group number (SSSS) to which the effective coefficient of the decoded AC coefficient belongs, instead of the code length (bit length) of the decoded code [CRM _i / CI _i ], the code [C
RM _i / CI _i ], the group index [CGI
_i ]] code length (bit length) is added to the total code length CL
The EN is output to the code shifter 181 and the group index cutout unit 186.

That is, the total code length CLEN is obtained and output by utilizing the fact that the bit length of the decoded value (SSSS) of [CI _i ] = [CGI _i ] (4) holds.

In FIG. 4, as an address signal
The indicated "00" is [RM_i(NNNN) = 0,
I_i(SSSS) = 1] code [CRM_i/ CI_i〕so
Yes, "100" means [RM_i(NNNN) = 0, I_i
(SSSS) = 1], that is, the code “CRM” of [EOB]
_i/ CI_i]. Also, "11011" is [RM _i
(NNNN) = 4, I_i(SSSS) = 1] code [C
RM_i/ CI_i].

The EOB (End of Block) is added only when the last AC component in the block is "0". That is, it is not added when the last AC component in the block is the effective coefficient (not 0).

Next, the group index cutout unit 1
Moving on to the description of 86. Group index cutout part 1
86, based on the total code length CLEN input from the decoding table 185 and the decoded group number (SSSS) value IDX from the variable length code data input from the latch unit 186, DC The group index [CGDL] of the difference value “L” and the group index [CGI _i ] (i of the effective coefficient of the AC coefficient)
, 1, 2, ..., N) are cut out (extracted) and output to the reverse grouping table 187.

The inverse grouping table 187 is the IDX (DC difference value "L") input from the decoding table 185.
The group number (SSSS) to which the group belongs or the group number (SSSS) to which the effective coefficient of the AC coefficient belongs and the DC input from the group index cutout unit 186.
Group index [CGD] indicating the position information of the difference value “L” or the effective coefficient of the AC coefficient in the group
L] or [CGI _i ] (i = 1, 2, ..., N), the group number (SSSS) to which the DC difference value “L” belongs and the group number (SSSS) to which the effective coefficient of the AC coefficient belongs ( SSSS) I _i (i = 1, 2, ..., N) is decoded.

Next, the operation of the variable length decoding circuit having the above configuration will be described. When the code shifter 181 inputs a variable length code string of an arbitrary block having the configuration shown in FIG.
16-bit variable length code data C starting from the code [CDL] of the group number (SSSS) to which the difference value "L" belongs
ODE ₁ is input to the latch unit 184 and the DC difference value “L” is also set.
The 22-bit variable-length code data CODE ₂ starting from the code [CDL] of the group number (SSSS) to which is belongs to is output to the latch unit 182.

When the latch section 184 latches the 16-bit variable length code data CODE ₁ starting from the code [CDL], it outputs it to the decoding table 185 as a 16-bit address signal.

The decoding table 185 outputs the 16-bit variable-length code data CODE ₁ starting from the input code [CDL] as an address signal to the DC component decoding table 300 including a ROM. As a result, the decoding table 300 for the DC component has the code [C
The group number (SSSS) IDX to which the DC difference value “L” corresponding to DL] belongs is output to the group index cutout unit 186 and the reverse grouping table 187, and the code [CDL] and the following code [CGD] are output.
The total code length CLEN of L] is output to the code shifter 181 and the group index cutout unit 186.

While the above operation is being performed, the 22-bit variable length code data (CODE ₂ ) 500 starting from the code [CDL] shown in FIG. Is entered.

Group index cutout unit 186
Is the 22-bit variable length code data (CODE ₂ )
DC from 500 based on the total code length CLEN and the decoded data IDX input from the decoding table 185.
The group index [CGDL] of the difference value “L” is cut out and output to the reverse grouping table 187. That is, the group index cutout unit 186
First, the variable length code data 501 equal to the value of the total code length CLEN is cut out from the beginning of the 22-bit variable length code data (CODE ₂ ) 500 input from the latch unit 182 shown in FIG. Then, a group index [CGDL] having the number of bits equal to the value of the decoded data IDEX behind the variable length code data 501 is cut out and is output to the reverse grouping table 187.

The reverse grouping table 187 has the group index CGDL and the group number (S) to which the DC difference value “L” input from the decoding table 185 belongs.
SSS) IDX and Table 10 in the format shown in FIG.
With reference to 0, the DC difference value “L” is decoded and output.

Next, the code shifter 181 receives the code [CRM ₁ / CI ₁ ] following the code [CGDL] from the variable length code string of one block shown in FIG. 18 based on the CLEN input from the decoding table 185. The 16-bit variable-length code string CODE ₁ starting from
A 32-bit variable-length code string starting from the code [CRM ₁ / CI ₁ ] shown in is output to the latch unit 182.

The latch section 184 is a 16-bit variable
Long code string CODE₁Type in the decryption table
Output to 185. The decryption table 185 is input
The above code [CRM₁/ CI₁] 16-bit
Variable length code data CODE₁Is the ROM
Output as an address signal to the decoding table for numbers. This
Therefore, the decoding table for the AC coefficient is the first A
Group number (SSSS) I to which the effective coefficient of C coefficient belongs
₁Value (equal to the value of IDX) is decoded and grouped in
Dex cutting section 186 and reverse grouping table
To the decoded code [CRM ₁
/ CI₁] And the following group index [CGI
₁] The total code length CLEN of
It is output to the loop index cutout unit 186. Well
Further, the decoding table of the AC coefficient is the effective coefficient I₁Before
Zero run length RM equal to the number of invalid coefficients following₁Also decrypts
Output.

While the above operation is being performed, the latch section 18
2 latches the 32-bit variable length code data shown in FIG. 5 (b) starting from the code [CRM ₁ / CI ₁ ] and outputs it to the group index cutout unit 186.

Group index cutout unit 186
When the 32-bit variable-length code data 600 shown in FIG. 5B is input, the 32-bit variable-length code string is generated based on the decoded data IDX and the total code length CLEN input from the decoding table 185. The group index [CGI ₁ ] is cut out from the group index and output to the reverse grouping table 187. That is, the group index slicing unit 186 first has a variable length code string for the number of bits equal to the value of the total code length CLEN from the head of the 32-bit variable length code string, that is, a variable length code [CRMI ₁ / CGI ₁ ]. And the group index [CGI ₁ ] are cut out, and then the group index [CGI ₁ ] having a bit length equal to the value of the IDX located behind the cut-out data is cut out based on the value of the IDX.

The inverse grouping table 187 is a decoding table.
First effective coefficient I of AC coefficient input from table 185
₁Group ID (SSSS) value IDX and
The above input from the loop index cutout unit 186
First effective coefficient I of AC coefficient ₁Group index of
[CGI₁], Decoding the first effective coefficient of the AC coefficient
And output.

Thereafter, similarly, the code shifter 181
Is the total code length CLE input from the decoding table 185.
A code [CRM based on the value of N₂/ CI₂], [CRM₃
/ CI ₃], ... 16-bit variable length code
Data and 32-bit variable length code data
The data is sequentially output to the latch unit 184 and the latch unit 182. That
Then, the same operation as above is repeated, and [zero run length R
M₂, Second effective coefficient of AC coefficient], [zero run length RM
₃, Third effective coefficient of AC coefficient], [Zero run length
RM_N, The Nth effective coefficient of the AC coefficient] are sequentially decoded.
The decoding of the variable length code string of the first 1 block is completed.
It

Then, the decoding process for the variable length code string of the first one block is repeated for the remaining blocks of one screen, whereby the decoding of the variable length code string for one screen is completed, and each block is decoded. All the effective coefficients of the AC coefficient are restored to.

The zero run length RM _i (i = 1, 2, ..., N) output from the decoding table 185 is set by an invalid coefficient restoring circuit for AC coefficients (not shown). It is restored to the column of "0" AC coefficient).

As a result, the variable length code for one screen is restored to the quantization coefficient for one screen. As described above, in the variable length coding circuit of the present embodiment, the code shifter 18
1 to the decoding table 18
5 and the group index cutout unit 186 respectively indicate the number of the group to which the DC difference value belongs (SSSS).
Group number (SSSS) I _i (i = 1, 2, ..., N) to which the effective coefficient of the DL or AC coefficient belongs and zero run length (NNNN) RM _i (i = 1, 2, ..., N) )
, And additional bits (CGDL or CGI _i (i = 1, 2, ...
.., N)) are extracted in parallel, so that variable-length code decoding can be performed faster than in the past.

The grouping of the differential value of the DC coefficient and the effective coefficient of the AC coefficient is not limited to the method of the above embodiment, and various other methods can be applied. Further, in the above embodiment, the variable-length code subjected to the orthogonal transform using the discrete cosine transform (DCT) is decoded, but the present invention is the discrete cosine transform (DCT), the discrete Lendend transform, and the Hadamard. It is also applicable to the decoding of a variable-length code in which image data is encoded using other orthogonal transforms such as (Hadamard) transform and Harr transform.

Further, in the above embodiment, the variable length code which has been Huffman coded as the entropy coding is decoded, but it is applicable to other variable length codes as the entropy coding.

[0091]

As described above, according to the decoding device of the variable length code of the image data of the first aspect of the invention, the decoding of the number of the group to which the differential value of the DC coefficient belongs and the decoding thereof are performed. Of the additional bit for the code, the decoding of the group number to which the effective coefficient of the AC coefficient belongs and the zero run length (the number of consecutive invalid coefficients), and the extraction of the additional bit for the decoded code are performed once for a variable length. Since the code strings are output in parallel, the variable-length codes encoded by grouping can be decoded at a higher speed than in the past.

According to the second aspect of the present invention, the variable length code decoding device for image data has the effect of the first aspect of the invention, in addition to the restoration of the difference value of the DC coefficient and the effectiveness of the AC coefficient. The coefficient can be restored faster than before.

[Brief description of drawings]

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

FIG. 2 is a block configuration diagram of a variable length decoding circuit according to an embodiment of the present invention.

FIG. 3 is a diagram illustrating a difference in configuration between a DC difference value decoding table used in the present embodiment and a conventionally used DC difference value decoding table.

FIG. 4 is a diagram illustrating a difference in configuration between a decoding table for effective coefficients of AC coefficients used in the present embodiment and a decoding table for effective coefficients of AC coefficients that has been conventionally used.

FIG. 5 is a diagram showing a configuration of a code string output from a code shifter to a group index cutout unit via a latch unit.

FIG. 6 is a block diagram of an ADCT conversion circuit.

FIG. 7 is a diagram showing an example of an original image signal of one block.

FIG. 8 is a two-dimensional DC for the original image signal of the block.
It is a figure which shows the two-dimensional DCT coefficient obtained by performing T conversion.

FIG. 9 is a block diagram of a two-dimensional DCT transform unit.

FIG. 10 is a diagram showing an example of a quantization threshold for linear quantization.

11 is a diagram showing a quantized coefficient obtained by linearly quantizing the two-dimensional DCT coefficient shown in FIG. 8 with the quantization threshold shown in FIG.

FIG. 12 is a diagram showing a scanning order when a quantized coefficient is subjected to zigzag scanning.

FIG. 13 is a block diagram of an ADCT restoration circuit.

FIG. 14 is a block diagram of a two-dimensional DCT transform unit.

FIG. 15 is a block diagram of a conventional linear quantization circuit.

FIG. 16 is a diagram showing a configuration of a table used for grouping DC difference values.

FIG. 17 is a diagram showing a configuration of a table used for grouping effective coefficients of AC coefficients.

FIG. 18: 1 encoded by adaptive discrete cosine transform
It is a figure which shows the structure of the variable length code sequence of a block.

FIG. 19 is a block diagram showing a configuration of a conventional variable length decoding circuit.

[Explanation of symbols]

 1 Code Sequence Extracting Means 2 Decoding Means 3 Additional Bit Extracting Means 4 Restoring Means

Claims (3)

[Claims] 1. A conversion coefficient obtained by orthogonally transforming tone values of a plurality of pixels in each block obtained by dividing an original image into a plurality of blocks composed of a plurality of images. In the decoding device of the variable length code of the image data, which is quantized, and the quantized coefficients obtained as a result are grouped, and the encoded variable length code is decoded, the variable length code of each block is input. From the code string extracting means (1) for extracting and outputting the code string in units of a predetermined length, and the code string output from the code string extracting means (1), the decoding or AC component of the group to which the difference value of the DC coefficient belongs Decoding the number of the group to which the effective coefficient belongs, decoding the run length of the invalid coefficient before the effective coefficient, and following the code of the group number to which the difference value of the decoded DC coefficient belongs and its code. The total code length of the additional bits and the bit length of the additional bits, or the code of the combination of the number of the group to which the effective coefficient of the decoded AC coefficient belongs and the run length of the decoded invalid coefficient, and the following code Decoding means (2) that outputs the total code length of the additional bits and the bit length of the additional bits, and the code string output from the code string extracting means (1) are output from the decoding means (2). Based on the total code length and the additional bit length, a combination of the additional bit following the number of the group to which the difference value of the DC coefficient belongs or the group number to which the effective coefficient of the AC coefficient belongs and the run length of the invalid coefficient An additional bit cutout means (3) for cutting out an additional bit following the code of, and the code string extracting means (1) cuts the second code string. A decoding device for a variable length code of image data, characterized in that the head of a code string to be output next is determined based on the total code length output from the decoding means (2) after output. 2. The number of the group to which the difference value of the DC coefficient output from the decoding means (2) or the group number to which the effective coefficient of the AC component belongs and the additional bit extraction means (3) are output. Decoding the variable length code of the image data according to claim 1, further comprising: a restoring unit (4) for decoding the differential value of the DC coefficient and the effective coefficient of the AC coefficient based on the additional bits. apparatus. 3. The variable-length code of the image data according to claim 2, wherein the variable-length code of the image data to be decoded is obtained by adaptive discrete cosine transform coding. apparatus.