JPH06319132A - Picture signal coder and decoder - Google Patents

Abstract

PURPOSE:To allow the coder/decoder to cope with controlling a code quantity by preparing a code book with respect to scaling factors and a quantization matrix set as a reference in advance. CONSTITUTION:A data block generated by a block processing circuit 2 is given to an orthogonal transformation device 3, in which 2-dimension orthogonal transformation such as discrete cosine transformation is applied to each block to obtain a transformation coefficient. A multiplier 6 multiplies each quantization step and each scaling factor fed from a quantization table 5 and a quantizer 7 quantizes the transformation coefficient based on the obtained coefficient. A variable length coder A8 generates a variable length code from an output of the quantizer T. The scaling factor and the variable length code are sent to a transmission line 10 via a bit stream mixer 9. A variable length decoder 12 decodes a quantization coefficient from the variable length code separated by a bit stream separator 11 and a multiplier 13 multiplies the quantization coefficient and the scaling factor, a table retrieval device A retrieves a relating code book to obtain a code vector.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image signal coding / decoding device for recording and transmitting a digitized image signal at a constant bit rate.

[0002]

2. Description of the Related Art As a method of dividing a digitized image signal into a plurality of data blocks in one screen and decoding image data encoded using orthogonal transformation and quantization, there are inverse quantization and inverse quantization. Decoding method using orthogonal transform (linear operation) (for example, J using discrete cosine transform as orthogonal transform)
The PEG method) is generally used. When decoding a data block encoded using orthogonal transform and quantization in this way, instead of using the inverse transform as described above, prepare a codebook for each coefficient of the data block,
There has been proposed a method for improving the image quality of a decoded image by decoding a data block by obtaining a vector sum of code vectors obtained by a table search for each coefficient.

FIG. 6 shows, for example, Proc. SPIE Conference on
Visual Communication and Image Processing ■ 91, Vo
l.1605, pp.487-498, Nov.1991, “Enhancemennt of Tranc
eform Coding by Nonlinear Interpolation ”, Siu-Wai
It is a block diagram of the conventional image signal encoding / decoding apparatus proposed by Wu, et.al. In the following, 1 is an image input terminal, 2 is a blocking device that divides an input image signal into 8 × 8 blocks, 29 is a cosine transformer that performs a two-dimensional discrete cosine transform on a data block obtained from the blocking device 2, 4 is a block scan for rearranging the output of the cosine transformer 29 in a predetermined order, 7 is a quantizer,
30 is a two-dimensional Huffman encoder for two-dimensional Huffman coding the quantized DCT coefficient, 10 is a transmission line, 31 is a two-dimensional Huffman decoder for decoding the variable length code reproduced from the transmission line 10, 33, Reference numeral 35 is a pre-designed codebook for each quantized coefficient, 32 and 34 are table search devices A, and 20 is a vector for calculating the vector sum by adding the code vectors output from the table searcher for each element. Contrary to block scan 4, adder 21 is
Reverse block scan for rearranging the output of the vector adder 20 and reproducing the data block, 22 is reverse block scan 2
The deblocking device reconstructs a decoded image signal by collecting the data blocks obtained in 1 for one screen.

Next, the operation of FIG. 6 will be described. The image signal input from the image input terminal 1 is divided by the block formation device 2 into data blocks in units of 8 pixels in the vertical direction and 8 pixels in the horizontal direction. The cosine transform device 29 performs a two-dimensional discrete cosine transform on the data block read out from the blocking device 2 in the form of 8 × 8 blocks. The transform coefficients obtained from the cosine transform device 29 are rearranged in the order shown in FIG. 7 and quantized by the quantizer 7. The output of the quantizer 7 is encoded by the two-dimensional Huffman encoder 30 and sent to the transmission line 10.

The variable length code reproduced from the transmission line 10 is 2
It is decoded by the three-dimensional Huffman decoder 30. The table search device A32 extracts a code vector for reproducing the data block from the corresponding code book 33 according to the decoded quantized coefficient. The operations of the table searcher A34 and the codebook 35 are the same as the operations of the table searcher A31 and the codebook 32. There are 64 sets of the table searcher A and the codebook, and the table searcher A32 is
For the first quantized coefficient I ₁ , the table searcher A34
It operates only on the _Nth quantized coefficient I _N. Then, the code vector X for each quantized coefficient
Outputs ₁ and X _N. Considering this for the k-th quantized coefficient I _k , the following equation is obtained.

[0006]

[Equation 1]

The search of the codebook is performed for the quantized coefficients for one block, and the output, that is, the code vector is all supplied to the vector adder 20, and the addition is performed for each element of the vector. This can be expressed by the following equation.

[0008]

[Equation 2]

The outputs of the vector adder 20 are rearranged in the order shown in FIG. 7 to reproduce the data block. The data blocks thus obtained are combined by the deblocking device 22 for one screen to reconstruct a decoded image.

Next, a codebook design method will be described. Letting X be the original data block obtained from the blocking device 2 and X be the reproduced data block, the average is 2
From the equations (1) and (2), the multiplication error d is

[0011]

[Equation 3]

[0012] Here, when the codebooks other than CB _j are fixed and C _j (i) that minimizes the above equation is obtained,

[0013]

[Equation 4]

[0014] Assuming that the training vector used for designing the codebook is V = {v}, C in equation (4)
_j (i) is calculated by the following equation.

[0015]

[Equation 5]

However, in R (j) i, the j-th quantization coefficient is i
Is a subset of the training vector
Represents the number of the elements. The vector regenerated using the codebook obtained by the equation (5) is v
Then, the mean square error d is expressed by the following equation.

[0017]

[Equation 6]

From the above, the codebook design algorithm is as follows. (1) Encode all training vector sets and store their coefficients. (2) Generate a set of initial codebooks. (3) The codebook is repeatedly updated until the convergence condition is satisfied.

[0019]

[Equation 7]

[0020]

An image for retrieving a data block by extracting a code vector from a code book designed in advance for a code encoded using orthogonal transformation and quantization as in the above example. In the signal encoding / decoding device, if the quantization matrix used on the encoding side is changed, the codebook for the matrix must be redesigned, and the code amount cannot be easily controlled. .

The present invention has been made in order to solve the above-mentioned problems, and by only designing a set of codebooks in advance, it is possible to control the amount of coded image signal coding. It is an object to provide a decoding device.

[0022]

In order to quantize a transform coefficient obtained by performing orthogonal transform on a data block divided into a certain size, the image signal coding / decoding device according to the invention of claim 1 It is provided with a quantization table that stores a quantization matrix set as a reference in advance.
The codebook is pre-designed for this reference quantization matrix. Then, at the time of encoding, quantization is performed using the coefficient obtained by multiplying the quantization step supplied from the quantization table and the scaling factor. The scaling factor is sent to the decoding side together with the image information. The decoding side is configured to perform decoding by searching the corresponding codebook from the multiplication result of the quantization coefficient and the scaling factor obtained by decoding the variable length code reproduced from the transmission path. is there.

An image signal coding / decoding apparatus according to a second aspect of the present invention comprises a scaling factor control apparatus for dividing the data block into smaller areas and changing the scaling factor for each area. Be prepared.

An image signal coding / decoding apparatus according to the invention of claim 3 directly, without decoding the variable length code,
It is equipped with a table searcher that retrieves code vectors.

An image signal coding / decoding apparatus according to a fourth aspect of the present invention further comprises a scaling factor control apparatus for dividing the data block into finer areas and changing the scaling factor for each area. A table search device for directly extracting a code vector without decoding the variable length code is provided.

[0026]

According to the invention of claim 1, quantization is performed by using a multiplication result of a quantization matrix set as a reference and a scaling factor, and the decoding side decodes the variable length code reproduced from the transmission line. From the multiplication result of the obtained quantization coefficient and the scaling factor, the corresponding codebook is searched.
Only the reference quantization matrix needs to be prepared, and the code amount can be controlled.

According to the second aspect of the invention, quantization is performed using the multiplication result of the quantization matrix set as a reference in advance and the scaling factor set for each area to which each quantization coefficient belongs, and the quantization is performed on the decoding side. Since the corresponding codebook is searched from the multiplication result of the quantization coefficient and the scaling factor obtained by decoding the variable length code reproduced from the path, the codebook is the one for the reference quantization matrix. It suffices to prepare only this, and it becomes possible to control the code amount. Further, since the data block can be divided into smaller areas (areas) and the scaling factor can be changed for each area, improvement in image quality of the decoded image can be expected.

According to the third aspect of the invention, quantization is performed using the multiplication result of the quantization matrix set as a reference and the scaling factor, and on the decoding side, the variable length code and the scaling factor reproduced from the transmission line are used. Since the corresponding codebook is directly searched, only the codebook for the reference quantization matrix needs to be prepared, and the code amount can be controlled.

According to the invention of claim 4, quantization is performed by using a multiplication result of a quantization matrix set as a reference in advance and a scaling factor set for each area to which each quantization coefficient belongs, and on the decoding side, transmission is performed. Since the corresponding codebook is searched directly from the variable length code and the scaling factor reproduced from the path, only the codebook for the reference quantization matrix needs to be prepared as a codebook. Correspondence becomes possible. Further, since the data block can be divided into smaller areas (areas) and the scaling factor can be changed for each area, improvement in image quality of the decoded image can be expected.

[0030]

【Example】

Example 1. 1 is a block diagram showing an image signal coding / decoding apparatus according to an embodiment of the present invention. In the figure, 1 is an image input terminal, 2 is a first blocking circuit for forming a data block from an input image signal, 3 is an orthogonal transformer, and 4 is an output of the orthogonal transformer 3 arranged in a predetermined order. The first block scan, 5 is a quantization table that stores a quantization matrix set as a reference in advance, 6 is a multiplier that multiplies the quantization step supplied from the quantization table 5 and a scaling factor, and 7 is a quantum , 8 is a variable length encoder, 9 is a bitstream mixer, 10 is a transmission line, 11 is a bitstream separator, 12 is a variable length decoder, 13 is a multiplier, 1
5, 17, and 19 are codebooks, 14 and 16 and 18 are table search devices A for searching codebooks from the codebooks, 20 is a vector adder, and 21 is a rearrangement of the outputs of the vector adder 20, and the data blocks are The second block scan for reproduction, 22 is an inverse block device for reconstructing a decoded image signal by collectively collecting data blocks for one screen.

Next, the operation of FIG. 1 will be described. The image signal input from the image input terminal 1 is divided into small blocks having a fixed size by the blocking circuit 2. The size of this small block is, for example, 8 pixels both in the vertical and horizontal directions. The data block formed by the blocking circuit 2 is input to the orthogonal transformer 3. The orthogonal transformer 3 performs a two-dimensional orthogonal transform such as a discrete cosine transform on each block to obtain a transform coefficient. The transform coefficients output from the orthogonal transformer 3 are rearranged in a predetermined order in the first block scan 4 and supplied to the quantizer 7. The multiplier 6 multiplies each quantization step supplied from the quantization table 5 by the scaling factor, and supplies the result to the quantizer 7. The quantizer 7 quantizes using the quantization step supplied from the multiplier 6. The variable length encoder A8 generates a variable length code for the output of the quantizer 7. The coding method used by the variable length encoder A may be two-dimensional coding as used in the JPEG method or a method of assigning one variable length code to each coefficient. The scaling factor used in the multiplier 6 and the variable length code generated by the variable length encoder 8 are mixed in the bit stream mixer 9 and sent to the transmission line 10.

The bit stream reproduced from the transmission line 10 is separated into a variable length code and a scaling factor by a bit stream separator 11. The variable length decoder 12
The quantized coefficient is decoded from the variable length code and supplied to the multiplier 13. The multiplier 13 multiplies the quantized coefficient by the scaling factor. Table search device A14
For the coefficient output from the multiplier 13, the code vector is extracted from the corresponding code book 15 and supplied to the vector adder 20. Table searcher A16 and codebook 17, table searcher A18 and codebook 1
The operation of 9 is the same as the operation of the table searcher 14 and the codebook 15. The table searcher A14 operates only for the first coefficient, the table searcher A16 operates only for the second coefficient, and the table searcher A18 operates only for the third coefficient. Although FIG. 1 shows the case of decoding using three coefficients, generally (when the data block is 8 × 8 pixels), the number of table searcher A and codebook pairs is six.
There are four, and one table searcher operates for one coefficient. The vector adder 20 is the table search unit 1
The vector sums of the code vectors output from 4, 16, and 18 are obtained. The output of the vector adder 20 is rearranged by the inverse vector scan 21 to reproduce the data block. The data blocks reproduced by the second vector scan 21 are combined by the deblocking device 22 for one screen to reconstruct a decoded image signal.

The scaling factor can be fixed to a constant value for one screen, or can be changed for each data block. When changing the scaling factor for each block, it is necessary to transmit the information for each block. However, when using a fixed value for one screen, it is only necessary to transmit the information once at the beginning of the screen. .

Example 2. FIG. 2 is a block diagram showing an image signal coding / decoding device according to an embodiment of the present invention. In the figure, 1 is an image input terminal, 2 is a first blocking circuit for forming a data block from an input image signal, 3 is an orthogonal transformer, and 4 is an output which rearranges the outputs of the orthogonal transformer in a predetermined order. 1 block scan, 5
Is a quantization table that stores a quantization matrix set as a reference in advance, 24a is a first scaling factor control device that divides the data block into smaller areas, and controls the scaling factor for each area. Is a multiplier for multiplying the quantization step supplied from the quantization table 5 and the scaling factor supplied from the scaling factor control device, 7 is a quantizer, 8 is a variable length encoder, and 9 is a bit stream. A mixer, 10 is a transmission line, 11 is a bit stream separator, 12 is a variable length decoder, and 24b is a scaling factor supplied to a multiplier based on the combination information of the scaling factors sent from the bit stream separator. 2 scaling factor controller, 13 multipliers, 15, 1 , 19 code book, 14,1
6 and 18 are a table searcher A for searching a code vector from a codebook, 20 is a vector adder, 21 is a reverse block scan for rearranging the outputs of the vector adder 20 and reproducing a data block, and 21 is one screen. Is a deblocking device for reconstructing a decoded image signal by collectively collecting the data blocks of.

Next, the operation of FIG. 2 will be described. The image signal input from the image input terminal 1 is subjected to blocking, orthogonal transformation, and block scanning as in the embodiment shown in FIG. 1, and is supplied to the quantizer 7. Scaling factor controller 2
4a is an area in which the data block is made finer
And the scaling factor set for each area is sent to the multiplier 6. Also, the scaling factor combination information is output. An example of the above area division is shown in FIG.
Shown in. The multiplier 6 is a quantization step and scaling factor control device 24a supplied from the quantization table 5.
Multiply with the scaling factor given by.
The quantizer 7 quantizes in the quantization step supplied from the multiplier 6. The variable length encoder A8 generates a variable length code for the output of the quantizer 7. The encoding method used in the variable length encoder A is such as that used in the JPEG method.
A two-dimensional coding method or a method of assigning one variable length code to each coefficient may be used. The combination information of scaling factors output from the scaling factor control device 24a and the variable length code generated by the variable length encoder 8 are mixed by the bit stream mixer 9 and sent to the transmission line 10.

The bit stream reproduced from the transmission path 10 is separated by a bit stream separator 11 into variable length code and scaling factor combination information. The variable length decoder 12 decodes the quantized coefficient from the variable length code,
It is supplied to the multiplier 13. The scaling factor control device 25b multiplies the scaling factor for each coefficient according to the scaling factor combination information.
Supply to. The multiplier 13 multiplies the quantized coefficient by the scaling factor. Table searcher 14
Extracts a code vector from the corresponding code book 15 according to the output of the multiplier 13,
Supply to 0. Table searcher A16 and codebook 1
7. The operations of the table searcher A18 and the codebook 19 are the same as the operations of the table searcher A14 and the codebook 15. The table searcher 14 uses the first coefficient
The table searcher 16 operates only on the second coefficient, and the table searcher 18 operates only on the third coefficient. FIG. 2 shows the case of decoding using three coefficients, but generally (when the data block is 8 × 8 pixels), there are 64 sets of the table searcher A and the codebook, and one coefficient is used. On the other hand, one table searcher operates. The vector adder 20 obtains the vector sum of the code vectors output from the vector searchers A14, 16, and 18. The output of the vector adder 20 is the inverse vector scan 2
The data blocks are re-arranged by 1 and the data block is reproduced.
The data blocks reproduced by the inverse vector scan 21 are combined by the inverse block device 22 for one screen to reconstruct a decoded image signal.

Example 3. FIG. 3 is a block diagram showing an image signal coding / decoding apparatus according to an embodiment of the present invention. In the figure, 1 is an image input terminal, 2 is a blocking circuit that forms a data block from an input image signal, 2 is an orthogonal transformer, and 3 is a first block that rearranges the outputs of the orthogonal transformer in a predetermined order. Scan 5 is a quantization table that stores a quantization matrix set as a reference in advance, 6 is a multiplier that multiplies the quantization step supplied from the quantization table 5 and a scaling factor, 7 is a quantizer, and 8 Is a variable length encoder, 9 is a bitstream mixer, 10 is a transmission line, 11 is a bitstream separator, and 26, 27 and 28 are codes corresponding to variable length codes separated by the bitstream separator and scaling factors. Table search device B for outputting a vector, 20 is a vector adder, 21 is a vector adder 20
Is an inverse block scanning device that rearranges the outputs of the above, and reproduces a data block, and 22 is an inverse blocking device that collectively reconstructs a decoded image signal by combining data blocks for one screen.

Next, the operation of FIG. 3 will be described. The image signal input from the image input terminal 1 is subjected to blocking, orthogonal transformation, and block scanning as in the embodiment shown in FIG. 1, and is supplied to the quantizer 7. The multiplier 6 multiplies each quantization step supplied from the quantization table 5 by the scaling factor, and supplies the result to the quantizer 7. The quantizer 7 quantizes using the quantization step supplied from the multiplier 6. The variable length encoder B8 is
A variable length code is generated for the output of the quantizer 7. The coding method used by the variable length encoder B is limited to the method of assigning one variable length code to each quantized coefficient. The scaling factor used in the multiplier 6 and the variable length code generated by the variable length encoder 8 are the bitstream mixer 9
Are mixed and sent to the transmission line 10.

The bit stream reproduced from the transmission line 10 is separated into a variable length code and a scaling factor by the bit stream separator 11. Table search device B26
Outputs a corresponding code vector from the variable length code and the scaling factor, and supplies the code vector to the vector adder 20. The operations of the table searchers B27 and 28 are the same as the operations of the table searcher B26. The table search unit B26 adds the first variable length code to the table search unit B2.
7 is the second variable length code in the table search unit B28
Operates only for the third variable length code. FIG. 3 shows a case of decoding using three variable length codes.
Generally (when the data block is 8 × 8 pixels), there are 64 table searchers B, and for one variable length code,
One table searcher works. Vector adder 20
Calculates the vector sum of the code vectors output from the table searchers B26, 27, 28. The outputs of the vector adder 20 are reordered in the second vector scan 17 to recreate the data block. The data blocks reproduced by the inverse block scanning 21 are combined by the inverse block forming device 22 for one screen and the decoded image signal is reproduced.

FIG. 4 is a block diagram showing a concrete example of the table search device B in FIG. In the figure, 101 is an input terminal to which a variable length code is input, 102 is an input terminal to which a scaling factor is input, 104 is a register for storing a variable length code, a shift register for storing a scaling factor, and 103 is A changeover switch for controlling the variable length code so that it is always stored in order from the beginning of the register 104, and 106 outputs each element of the corresponding code vector according to the data stored in the register 104 and the shift register 105. R
OM, 107 is an output terminal for outputting the elements of the code vector, 108 is an output terminal for outputting a flag bit for identifying whether the data output from the output terminal 107 is valid or invalid, and 109 is an output terminal 108. It shows a register that stores data only when the data output from is valid.

Next, the operation of FIG. 4 will be described. Input terminal 1
The scaling factor input from 02 is stored in the shift register 105. The variable length code input from the input terminal 101 is stored in the register 104. At this time, the changeover switch 103 is
The inputs are switched so that they are always stored in order from the beginning of the register 104. The ROM 106 is prepared in the same number as the number of elements of the data block (64 when the data block is 8 × 8), and the variable length code register 104.
And the data stored in the scaling factor register 105 as an address, and the corresponding data is output terminal 1
It outputs from 07. At this time, the ROM outputs data each time 1 bit is input to the register 104, but the data is output with the code completely stored in the register for the code of 1 word (valid Data)
Or the data is stored in advance by adding a flag bit for identifying whether the code is output (invalid data) while the code is still being stored. Then, it is stored in the register 109 only when the data is valid.

Example 4. FIG. 5 is a block diagram showing an image signal coding / decoding apparatus according to an embodiment of the invention of claim 4. In the figure, 1 is an image input terminal, 2 is a first blocking circuit that forms a data block from an input image signal, is an orthogonal transformer, and 4 is a first that rearranges the outputs of the orthogonal transformer in a predetermined order. Block scan, 24
a is a first scaling factor control device that divides a data block into smaller areas (areas) and controls the scaling factor for each area, 5 is a quantization table that stores a quantization matrix set as a reference in advance, 6 Is the quantization step supplied from the quantization table 5 and the first scaling factor controller 2
4a is a multiplier for multiplying with a scaling factor, 7 is a quantizer, 25 is a variable length encoder B, 9 is a bitstream mixer, 10 is a transmission line, 11 is a bitstream separator, and 24b is Second scaling factor control device for outputting a scaling factor for each coefficient based on the combination information output from the scaling factor control device, 26, 27, 28 are table search devices B, 20 is a vector adder, 21 is a vector addition Bowl 2
Inverse block scanning for rearranging the outputs of 0 and reproducing the data blocks, and 21 is an inverse blocking device for collectively reconstructing the decoded image signal for the data blocks of one screen. The configuration of the table search device B is the same as in FIG.

Next, the operation of FIG. 5 will be described. The image signal input from the image input terminal 1 is subjected to blocking, orthogonal transformation, and block scanning as in the embodiment shown in FIG. 1, and is supplied to the quantizer 7. Scaling factor controller 4
5a is an area in which the data block is made finer
And the scaling factor set for each area is sent to the multiplier 6. Further, the scaling factor combination information is supplied to the bitstream mixer 9. The multiplier 6 multiplies the quantization step supplied from the quantization table 5 and the scaling factor supplied from the scaling factor controller 24a. Quantizer 7
Performs quantization with the quantization step supplied from the multiplier 6. The variable length encoder B25 generates a variable length code for the output of the quantizer 7. The coding method used by the variable length encoder B is limited to a method of assigning one variable length code to each quantized coefficient. The scaling factor combination information output from the scaling factor control device 24a and the variable length code generated by the variable length encoder B25 are mixed by the bit stream mixer 9 and sent to the transmission line 10.

The bit stream reproduced from the transmission line 10 is separated by the bit stream separator 11 into variable length code and scaling factor combination information. The scaling factor control device 24b supplies the scaling factors to the table searchers B26, 27, 28 according to the above-mentioned scaling factor combination information. The table searcher B26 outputs a corresponding code vector from the variable length code and the scaling factor, and the vector adder 2
Supply to 0. The operations of the table searchers B27 and B28 are the same as the operations of the table searcher B26. The table search unit B26 is the scaling factor for the first variable-length code and its code, the table search unit B27 is the scaling factor for the second variable-length code and its code, and the table search unit B28 is the third variable. Only works for long codes and scaling factors for that code. FIG. 5 shows a case of decoding using three variable length codes, but generally (when the data block is 8 × 8 pixels), there are 64 table searchers B, and one variable length code is used. On the other hand, one table searcher operates. The vector adder 20 is a table searcher B26, 27, 2
For the code vector output from 8, the vector sum is obtained. The output of the vector adder 20 is rearranged by the inverse vector scan 21 to reproduce the data block. The data blocks reproduced by the inverse vector scanning 20 are combined by the inverse blocking device 22 for one screen to reconstruct a decoded image signal.

[0045]

As described above, according to the present invention, the decoding is performed by the codebook search as in the prior art.
A quantization table that stores a reference quantization matrix is provided, and quantization is performed using the coefficient obtained by multiplying the quantization step supplied from this quantization table and the scaling factor, and the variable length code is decoded. Since the corresponding codebook is searched from the quantized coefficient obtained as above and the above scaling factor, it suffices to prepare a codebook for the quantization matrix set as a reference in advance, and it is possible to control the code amount. It will be possible.

Further, by providing the scaling factor control device, the data block can be divided into finer areas (areas) and the scaling factor can be changed for each area, so that improvement in the image quality of the decoded image can be expected.

[Brief description of drawings]

FIG. 1 is a block diagram showing an image signal coding / decoding device according to an embodiment of the present invention.

FIG. 2 is a block diagram showing an image signal coding / decoding device according to an embodiment of the present invention.

FIG. 3 is a block diagram showing an image signal coding / decoding device according to an embodiment of the present invention.

FIG. 4 is a block diagram showing a specific example of the table search device in FIG.

FIG. 5 is a block diagram showing an image signal encoding / decoding device according to an embodiment of the invention of claim 4;

FIG. 6 is a block diagram of a conventional image signal encoding / decoding device.

FIG. 7 is a diagram showing a scanning order when scanning data in one block.

FIG. 8 is a diagram showing an example of a quantization matrix.

FIG. 9 is a diagram showing an example of area division.

[Explanation of symbols]

 5 quantization table 6 multiplier 9 bit stream mixer 11 bit stream separator 14, 16, 18 table searcher A 15 codebook (CB1) 17 codebook (CB2) 19 codebook (CB3) 20 vector adder 24a, 24b Scaling factor control device 26, 27, 28 Table searcher B

Claims (4)

[Claims] 1. A coding unit that divides a digital image signal into a plurality of data blocks in one screen, performs orthogonal transformation and quantization, and then performs variable length coding, and the coding unit. In an image signal encoding / decoding device configured with a decoding unit that reproduces a data block from the obtained code,
A quantization table that stores a quantization matrix set as a reference in advance to quantize the transform coefficient obtained by the orthogonal transformation; a first multiplier that multiplies the quantization matrix and a scaling factor; A bitstream mixer for mixing a variable length code generated by long coding and a scaling factor; a bitstream separator for separating the bitstream output from the bitstream mixer into a variable length code and a scaling factor; A second multiplier for multiplying the quantized coefficient obtained by decoding the variable length code separated by the bit stream separator and a scaling factor separated in the same manner; With the multiplication result as input, the corresponding code vector is prepared from the prepared code book. A table search unit to issue an image signal coding and decoding apparatus characterized by comprising a vector adder for obtaining the vector sum of the resulting code vector from the table search unit. 2. The image signal coding / decoding apparatus according to claim 1, wherein the data block obtained by dividing the image signal is further divided into smaller areas, and a scaling factor is set for each area. An image signal encoding / decoding device comprising a scaling factor control device for changing. 3. A digital image signal is divided into a plurality of data blocks in one screen, subjected to orthogonal transformation and quantization, and then subjected to variable length coding. In an image signal encoding / decoding device configured with a decoding unit that reproduces a data block from the obtained code,
A quantization table that stores a quantization matrix set as a reference in advance to quantize the transform coefficient obtained by the orthogonal transformation, a multiplier that multiplies the quantization matrix and a scaling factor, and variable-length coding A bitstream mixer that mixes the variable length code generated by the above with a scaling factor, a bitstream separator that separates the bitstream output from the bitstream mixer into a variable length code and a scaling factor, and the bitstream A table search device that extracts a corresponding code vector from a codebook prepared in advance by using a variable length code separated by a separator and a scaling factor as input, and a vector that calculates the vector sum of the code vectors obtained from the table search device. With adder Picture signal coding and decoding apparatus characterized by comprising. 4. The image signal coding / decoding apparatus according to claim 3, wherein the data block obtained by dividing the image signal is further divided into smaller areas, and a scaling factor is set for each area. An image signal encoding / decoding device comprising a scaling factor control device for changing.