JPH03177163A - Picture data decoding system - Google Patents

Abstract

PURPOSE:To quicken the inverse quantization processing by selectively applying inverse quantization only to a valid component included in a quantized coefficient. CONSTITUTION:A valid component fed from a selection means 121 and not zero in the quantization coefficients is fed to a multiplier means 123. Moreover, a quantized threshold level corresponding to the valid component in the quantized threshold levels stored in advance in a storage means 122 is outputted and fed to the multiplier means 123. Thus, the component of the quantization coefficient fed from the multiplier means 123 and the component of the quantized threshold level are multiplied, then only the valid component in the quantization coefficient is selectively subject to inverse quantization. Thus, the inverse quantization processing as to the valid component being zero included in the quantization coefficient is omitted and the quantization processing is quickened.

Description

[Detailed description of the invention] [Table of contents] Overview Industrial field of application Conventional technology The invention attempts to solve! l! Means-effect embodiment for solving the problem ■, Correspondence relationship between the embodiment and FIG. In order to reduce the time required for the restoration process, the image data restoration method was developed by dividing the image into multiple blocks each consisting of multiple pixels, and by orthogonally transforming the pixel data of these blocks. The encoded data generated by quantizing the component and encoding the quantized coefficients generated by this quantization process is introduced, and the quantized coefficients obtained by decoding the encoded data by the decoding means are dequantized. Dequantize by
In the image data restoration method in which image data is generated by orthogonally transforming the transform coefficients generated by the dequantization means by the orthogonal transformation means, the dequantization means converts the values in the quantized coefficients supplied from the decoding means. A selection means for selecting an effective component for which is not zero, and a quantized darkness value corresponding to each component of the conversion coefficient are stored, and the quantized darkness value corresponding to the effective component selected by the selection means is outputted. storage means;
Multiplying the effective component selected by the selection means and the quantized dark value component corresponding to the effective component output by the storage means, and outputting this multiplication result as a component of the conversion coefficient corresponding to the effective component. and a multiplication means.

[Industrial application field]

The present invention relates to an image data restoration method for multivalued images such as halftone images and color images, and in particular, the present invention relates to an image data restoration method for multivalued images such as halftone images and color images. The present invention relates to an image data restoration method that restores encoded data that is encoded later.

Since the amount of information in image data representing multivalued images such as halftone images and color images is enormous, it is necessary to compress the amount of information when storing and transmitting this image data. .

As a coding method that compresses the amount of information without impairing the characteristics of the image data of such a multivalued image, an adaptive discrete cosine transform coding method (Adaptive
e Discrete Co51neTransfo
rm (hereinafter referred to as ADCT method) is widely used.

[Conventional technology]

FIG. 5 shows the configuration of an encoding device to which the ADCT method is applied.

The image to be encoded is divided into blocks (see FIG. 6) each consisting of, for example, 8×8 pixels, and the pixel data of each of these blocks is input to the DCT transform unit 511.

This DCT transformation unit 511 performs two-dimensional discrete cosine transformation (hereinafter referred to as DCT) on input pixel data.
By performing this, the pixel data of each block is orthogonally transformed. As a result, DCT coefficients representing the spatial frequency distribution of the image of each block are generated as a matrix of 8 rows and 8 columns as shown in FIG. 7, and are supplied to the linear quantization unit 520.

In the linear quantization unit 520, processing is performed to quantize the above-mentioned DCT coefficients using a vision adaptive darkness value adapted to vision.

This vision adaptive darkness value was determined based on the results of a vision experiment that investigated the sensitivity of vision to each spatial frequency. This quantization matrix VyH is given in the form of a matrix of 8 rows and 8 columns as shown in FIG. 8, and by dividing each component of the DCT coefficient by each component of this quantization matrix VTN, a linear Quantization processing is performed by the quantization unit 520.

FIG. 9 shows an example of DCT coefficients (hereinafter referred to as quantization coefficient D0°) quantized by the above-mentioned quantization matrices Vy and I. As shown in FIG. Since the component corresponding to the spatial frequency has a smaller absolute value than the value of the corresponding component of the quantization threshold,
The components of these DCT coefficients are quantized to rQJ.

In this way, the component in the upper left corner of the matrix indicating the DC component and the AC
Only a very small number of components representing low spatial frequency components among the components are often quantized to values other than rQ1.

The quantized coefficient [)IILI generated as described above
are outputted in the order shown in FIG. 10 and inputted to the encoding unit 531 as one-dimensional data. Furthermore, outputting matrix components in the order shown in FIG. 10 and converting them into one-dimensional data in this manner is called zigzag scanning.

In the encoding unit 531, first, a quantized coefficient U component (hereinafter referred to as an index) having a value other than rQJ is detected, and a quantized coefficient I following this index is detected.
) the number of times rQl is continuously supplied as a component of ou (
(hereinafter referred to as runs) are counted. Hereinafter, the component of the quantization coefficient u having a value other than rQ1 will be referred to as an effective coefficient, and the component of the quantization coefficient I)00 having a value rQJ will be referred to as an invalid coefficient.

For example, when the quantization coefficient I) IILI shown in FIG. 9 is converted into one-dimensional data by the above-mentioned zigzag scan and expressed as a combination of index and run, the result is as shown in FIG.

Here, the first index data of each block corresponds to the DC component of this block, and in the figure, the index data D is m in the image of one screen.
The DC component of the block is shown. Also, index data h.

..., 1. each of (1) of the quantization coefficient I)ou
. . , the (4,1) component, and the run data R1 indicates that the run is "ri". In addition, the run data Rmob is
The quantization coefficients obtained for the block indicate that all remaining components of u are invalid coefficients.

The combination of this index and run is code table 53.
Based on the Huffman table stored in 2, each combination is converted into a code with a length corresponding to the frequency of appearance.

Here, in order to reduce the amount of code data when encoding an image, the Huffman table is determined so that short codes are assigned in descending order of the frequency of appearance of combinations of index and run.

In this way, one block of images is encoded,
By repeating the above operations for all blocks included in one screen, the image data of one screen is encoded, and this encoded data is transmitted via a transmission path etc.
Alternatively, it is stored in a disk device or the like.

The encoded data generated in this manner is restored to image data by a restoration device as shown in FIG.

The decoding table 612 stores combinations of indexes and runs corresponding to variable length codes in the form of a table, and based on this decoding table 612, the decoding unit 611 decodes the supplied encoded data. Processing is performed. Through this decoding process, index data indicating the value of the effective coefficient,
Run data indicating the number of invalid coefficients following the valid coefficients are alternately supplied to the inverse quantization unit 620.

The quantized dark value holding unit 626 of this inverse quantization unit 620 has
The above-mentioned quantization matrix V? ll is stored.

Further, the decoded data from the decoding unit 611 is transferred to the dequantization unit 6
20 demultiplexers 621.

Further, the address counter 627 is connected to the tie control section 6.
A counting operation is performed in accordance with the pixel update signal supplied from 22, and the quantization coefficient holding unit 6 uses this counted value as an address.
28.

The supplied decoded data is sent to the demultiplexer 621 according to the selection signal supplied from the tie selection control section 622.
The index data is supplied to a multiplexer 623 and the run data is supplied to a run length determination section 624.

The run length determining unit 624 holds the number of invalid coefficients indicated by the supplied run data, and maintains the logic "0'°" until pulses of the above-mentioned pixel update signal are supplied by the number of invalid coefficients. A switching signal is supplied to the multiplexer 623.

This multiplexer 623 outputs the index data in response to the index data input, and also outputs "01" output from the zero generator 625 in response to the above-mentioned switching signal, thereby controlling the encoding process. The invalid coefficients that were omitted during the process are inserted, and the quantized coefficients I)
ou converted into one-dimensional data is reproduced.

The output of this multiplexer 623 is sent to the quantization coefficient holding unit 6 in accordance with the output of the address counter 627 mentioned above.
28, the quantized coefficient I)ou in a matrix format of 8 rows and 8 columns is restored.

Next, the timing control section 622 switches the quantization coefficient holding section 628 and the quantization rilj (I! 1)ou
Each component of the quantization matrix VTH and the quantization matrix VTH are sequentially read out.

By multiplying each component of the quantization coefficient D1 by each component of the quantization matrix TH by the multiplier 629,
Quantization coefficient I) Inverse quantization processing of QU is performed, and DCT
The coefficients are restored.

The inverse DCT transform unit 631 performs inverse DCT transform on the DCT coefficients, thereby orthogonally transforming the DCT coefficients of each block and restoring the image data of each block.

[Problem to be solved by the invention]

By the way, in the conventional method described above, the multiplier 629
For all components of Ll, the quantization coefficients are restored by the quantization matrix ■. . Multiplication with the corresponding component of is performed. Therefore, in order to dequantize the quantization coefficient D (ILI) corresponding to one block, it is necessary to
This requires multiple multiplication operations.

Therefore, when restoring an image of one screen, multiplication processing is performed an enormous number of times, resulting in the problem that the time required for the restoration processing is long.

However, since it is clear that the result of multiplication with zero is zero, the multiplication process for the component whose value is rQJ (that is, the invalid coefficient) included in the quantization coefficient I)l+11 is a wasteful process.

On the other hand, as mentioned above, most of the components of the quantization coefficient [ If only the multiplication with the component of the quantization matrix VtH is performed, the speed of the inverse quantization process can be increased.
It is considered possible to shorten the time required for image restoration processing.

The present invention was created in view of these points, and an object of the present invention is to provide an image data restoration method that reduces the time required for restoration processing of encoded data.

[Means to solve the problem]

FIG. 1 is a principle block diagram of the image data restoration method of the present invention.

In the figure, an image is divided into multiple blocks each consisting of multiple pixels, each component of the transform coefficient obtained by orthogonally transforming the pixel data of these blocks is quantized, and the quantization generated by this quantization process is performed. Encoded data generated by encoding the coefficients is introduced, the decoding means 111 decodes the encoded data, the quantized coefficients obtained are dequantized by the dequantization means 120, and the dequantization means 120 generates the quantized coefficients. The transformed coefficients are converted into orthogonal transform means 1.
The inverse quantization means 120 in the image data restoration method that generates image data through orthogonal transformation using the decoding means 1
a selection means 121 for selecting an effective component whose value is not zero among the quantization coefficients supplied from the transform coefficient;
a storage means 122 for outputting a quantized darkness value corresponding to the effective component selected by the selection means 121 and a quantized darkness value corresponding to the effective component selected by the selection means 121 and the effective component output by the storage means 122; , and outputs the multiplication result as a component of the conversion coefficient corresponding to the effective component.

[For production]

The decoding means 111 decodes the introduced encoded data, and the generated quantization coefficients are supplied to the selection means 121 of the dequantization means 120.

This selection means 121 selects an effective component whose value is not zero among the supplied quantization coefficients, and the multiplication means 123
is supplied to At this time, the storage means 122 outputs a quantized dark value corresponding to the above-mentioned effective component among the quantized dark values stored in advance, and the multiplication means 123
is supplied to

Therefore, by multiplying the supplied quantization coefficient component and the quantized dark value component by the multiplication means 123, only the above-mentioned effective components among the quantization coefficients are selectively dequantized. .

In the present invention, by selectively supplying the effective component in the quantized coefficient and the component of the quantized dark value corresponding to this effective component to the multiplication means 123, only the above-mentioned effective component is is selectively dequantized, so the dequantization process for invalid components whose quantization coefficients have zero values can be omitted, and the number of multiplication processes can be reduced. Processing can be performed at high speed.

〔Example〕

Hereinafter, embodiments of the present invention will be described in detail based on the drawings.

FIG. 2 shows the configuration of a restoration device to which the image data restoration method in one embodiment of the present invention is applied.

I and 1゛... Here, the correspondence between the embodiment of the present invention and FIG. 1 will be shown.

The decoding means 111 includes a decoding section 211. This corresponds to the decoding table 212.

The dequantization section 120 corresponds to the dequantization section 220.

The selection means 121 includes an address calculation section 222. Address holding unit 223. This corresponds to the quantization coefficient holding section 226.

The storage unit 122 includes an address calculation unit 222. Address holding unit 223. This corresponds to the i quantization threshold holding unit 227.

Multiplying means 123 corresponds to multiplier 228.

The orthogonal transform means 141 corresponds to the inverse DCT transform section 241.

Examples of the present invention will be described below assuming that the correspondence relationship as described above exists.

2, 2 and 2, 211 indicates a decoding section, 212 a decoding table, 220 an inverse quantization section, and 241 an inverse DCT transformation section, respectively.

In the decoding table 212, combination data of variable-length codes, corresponding indexes, and runs are stored in advance in a table format.

Encoded data supplied to the decoding device via a transmission path etc. is first input to the decoding section 211. This decoding unit 211
The above-mentioned decoding table 212 is searched based on the input encoded data, and the corresponding index and run combination data is read out from the decoding table 212 and supplied to the dequantization unit 220 as decoded data. be done.

In the inverse quantization unit 220, the DCT coefficients are restored by performing inverse quantization processing on the quantization coefficient K)011 based on this decoded data, and the inverse DCT transformation unit 241
Then, the DCT coefficients are subjected to inverse DCT transformation and then orthogonally transformed again to restore the image data.

The configuration and operation of the inverse quantization section 220 will be described below.

The dequantization unit 220 includes a demultiplexer 221 that separates decoded data into index data and run data;
An address calculation section 222 that calculates an address (described later) based on run data supplied from the demultiplexer 221, an address holding section 223 that stores the addresses calculated by the address calculation section 222, and an address storage section 223 that counts the number of effective coefficients. counter 225 and demultiplexer 22
1 and the quantization matrix VTM described above.
A quantized dark value holding section 227 that stores the quantized dark value holding section 227, a multiplier 228 that multiplies the output of the quantized coefficient holding section 226 and the output of the quantized dark value holding section 227, and controls the timing of the operation of each section. The timing control section 224 is configured to include a timing control section 224.

The decoded data supplied from the decoding unit 211 is input to the demultiplexer 221, and the timing control unit 224 generates a selection signal that becomes logic “1” when index data is supplied as the decoded data. CsL is the control terminal S of this demultiplexer 221
is supplied to.

The demultiplexer 221 outputs a selection signal C3 to a control terminal S.
The decoded data (index data) input when a logic "1" is supplied as L is stored in the quantization coefficient holding unit 2.
26, and when a logic "0" is supplied, manually input decoded data (run data) is supplied to the address calculation section 222.

The counter 225 is an up/down counter, and can be set to an addition mode or a subtraction mode according to a counting control signal inputted to the control terminal 5LID.
Further, depending on the set mode, the counting operation is performed in synchronization with a synchronization signal supplied to the clock terminal.

Quantization coefficient holding unit 226 and quantization dark value holding unit 227
is formed from storage areas corresponding to each component of a matrix of 8 rows and 8 columns, and the address of each area is specified by the zigzag scan order shown in FIG. Further, the quantization coefficient holding unit 226 has a control terminal S
The write mode and the continuous write mode can be set according to the continuous write control signal manually inputted to Il, 1.
For example, when the logic 111'' is input to the control terminal S, the writing mode is set, and when the logic is ``0'', the writing mode is set.

The quantized dark value holding unit 227 reads and outputs the index data stored at the address specified by the address input when a logic "O" is supplied to the control terminal SR as a successive control signal. There is.

Further, as shown in FIG. 4, the address holding unit 223
For example, it is formed from 64 storage areas, and numbers 1 to 64 correspond to each storage area. The addresses supplied from the address calculation unit 222 are sequentially stored in these storage areas, and the last stored address is output and supplied to the address calculation unit 222 and the quantization coefficient holding unit 226. It has become.

In the dequantization section 220 shown in FIG. 2, first, based on the decoded data supplied from the decoding section 211, a process of restoring the quantization coefficient D0 into a matrix format is performed.

When starting the restoration process for the decoded data corresponding to each block, the contents of the address holding unit 223 described above are cleared, and fil is stored as the initial value of the address in the storage area with number l.

Further, the timing control unit 224 controls the quantization coefficient holding unit 2
26 control terminal S□ as a continuous write control signal.
" is input to set the write mode, and at the same time, the contents of the storage area corresponding to each address of this quantization coefficient holding section 226 are cleared.

Further, the timing control unit 224 controls the counter 225
is set to addition mode, and this counter 22
The count value of 5 is reset to the initial value rO.

Decoded data output from the decoder 211 (see Figure 3)
is the selection signal C! supplied from the tying control unit 224 by the demultiplexer 221 of the inverse quantization unit 220.
The data is separated into index data and run data according to L, and supplied to the quantization coefficient holding section 226 and the address calculation section 222, respectively.

When the first index data is input, the address holding unit 223 uses the pre-stored initial value “
11 is supplied to the quantization coefficient holding unit 226 as an address, and this index data is stored corresponding to address r1j.

Here, the run data indicates the number of invalid coefficients following the valid coefficient indicated by the corresponding index data. Therefore, the result of integrating the number indicated by the supplied run data plus rll indicates the total number of components of the one-dimensional quantized coefficient I)ou supplied as decoded data up to that point. and this total number corresponds to the address of the last component.

The address calculation unit 222 uses the run data supplied from the demultiplexer 221 and the address holding unit 223 described above.
By performing addition with the output of and further adding rlJ to the addition result, the above-mentioned integration processing is performed, and this integration result is supplied as an address to the address holding unit 223,
This address is stored by the address holding section 223.

However, in this case, rl4 stored in the address holding unit 223 as the initial value of the address is further added. Therefore, the address calculated by the address calculation unit 222 is the address of the effective coefficient corresponding to the index data to be supplied next as decoded data, and the next input index data corresponds to this address. , are stored in the quantization coefficient holding unit 226.

For example, the index data D corresponding to the DC component of the decoded data shown in FIG. , the number rl in Figure 10
The quantization coefficients indicated by , are stored in the storage area corresponding to the component (1, 1) of u. This index data D1
According to the run data R0 corresponding to
22, the address r24 corresponding to the next index data (I2) is calculated. By performing similar processing on the subsequently supplied decoded data Hz, R-, IU, . . . , I6°R3), the address calculation unit 2
23, address r3”.

r4i, r5i, I6'', rl OJ are calculated sequentially, and corresponding to these addresses, index data It, I! + Ia, Is, Th, I? is stored in the quantization coefficient holding unit 226. Also, in this case,
In the address holding unit 223, as the address of the effective coefficient,
'IJ, r2J, =..., I5'.

'6J, rlO' are sequentially accumulated (see FIG. 4).

In this way, by repeating the above-described process every time run data is supplied, the address calculation unit 222
The addresses of the effective coefficients are sequentially calculated by , and the addresses of the effective coefficients are sequentially stored by the address holding unit 223 . In addition, corresponding to this address, the index data supplied subsequent to the run data described above is stored in the quantization coefficient holding unit 226, whereby the quantization coefficient I)ou in a matrix format of 8 rows and 8 columns is stored. will be restored.

Further, the timing control section 224 supplies a synchronization signal synchronized with the index data storage operation by the quantization coefficient holding section 226 described above to the counter 225, and the addition operation by the counter 225 causes the quantization coefficients to be calculated for each block. The total number of valid coefficients included in U is counted.

The address calculation unit 222 also uses Ro as run data.
. is supplied, it is determined that the supply of decoded data corresponding to one block has ended, and the block end signal (
BEN) and supplies it to the timing control section 224.

In response to this block end signal, the timing control section 22
4 is the quantization coefficient DTILI for one block
It is determined that the restoration process has ended, and thereafter, the inverse quantization process of the quantization coefficient N)@U corresponding to this block is started.

Hereinafter, the quantization coefficients restored as described above]D
The inverse quantization process of oa will be explained.

First, the timing control unit 224 controls the quantization coefficient holding unit 2
Logic "0" is supplied to the control terminal S□ of 26 as a continuous write control signal - (R/W), and logic "'0°" is supplied to the quantization threshold value holding section 227 as a read control signal (RED). As a result, from now on, the quantization coefficients 1)o are obtained from the quantization coefficient holding section 226 and the quantization dark value holding section 227 in accordance with the address supplied from the address holding section 223.
t+ and the corresponding components of the quantization matrix VTH are read and provided to multiplier 228.

Further, the timing control unit 224 has a control terminal S of the counter 225. Logic "0" is supplied as a counting control signal to set the counter 225 to the subtraction mode, and the storage area of the address holding section 223 is sequentially designated, and a synchronization signal synchronized with this designated operation is supplied to the counter 225. do.

As a result, the addresses stored in the address holding section 223 are sequentially output and supplied to the quantization coefficient holding section 226 and the quantization dark value holding section 227. 226 to quantization coefficient I)o
The corresponding effective coefficient of u is output, and the quantized dark value holding unit 2
27 to the quantization matrix V corresponding to this effective coefficient,
The components of the process are output.

For example, when the quantization coefficient ID, Ll is restored with respect to the decoded data shown in FIG. 3, the address r11. r21゜..., r6'',
rloJ is sequentially output (see FIG. 4). According to these addresses, the components of the quantized coefficient I)ou stored in the quantized coefficient holding unit 226 are ``64'', '-16J,
・, r-8J, r-8J are read out sequentially (
(see FIG. 8), the quantization dark value holding unit 227 stores the quantization matrix V corresponding to the components of these quantization coefficients DQLI as r16'', rll'', -, rlo
", r14" are read out (see FIG. 7), and multiplier 2
28.

In parallel with this inverse quantization process of the effective coefficients, the counter 22
5, the count value is subtracted in synchronization with the synchronization signal described above, and when the count value reaches r□, a stop signal is output from the output terminal O of this counter 225, and the timing control unit 224 It is now being supplied.

In response to this stop signal, the timing control unit 224 determines that the inverse quantization process of the quantization coefficient I)ou corresponding to one block has been completed, and ends the operation of specifying the storage area for the address holding unit 223. It looks like this.

In this way, by repeating the reading operation described above until the count value of the counter 225 reaches 101, the effective coefficients included in the quantization coefficient Dou and the quantization matrix V
By sequentially supplying the corresponding components of B to the multiplier 228 and performing multiplication processing, it becomes possible to selectively dequantize only the effective coefficients included in the quantization coefficients Ll.

Further, the output of the multiplier 228 and the corresponding address are supplied to an input buffer (not shown) provided inside the inverse DCT conversion unit 241, and the output of the multiplier 228 is By storing the output 8 rows 8
The DCT coefficients in column matrix form are restored. This DCT
One block of image data can be obtained by performing inverse DCT transformation on the coefficients as in the conventional method.

Here, in the dequantization process, the multiplication process by the multiplier 228 is the most time-consuming process, so by omitting the multiplication process regarding invalid coefficients as described above, the dequantization process can be significantly reduced. You can expect faster speeds.

In addition, the quantization coefficient D0 normally includes many invalid coefficients. For example, all 57 components except for the 7 valid coefficients included in the quantization coefficient pau shown in FIG. 8 are invalid coefficients. be. In this case, the multiplication process for the 57 invalid coefficients is omitted, and the 64 multiplication operations performed in the conventional method are replaced with 7 multiplication operations.
Compared to conventional methods, dequantization processing can be performed in about 1/10 of the time.

In this way, by selectively dequantizing only the effective coefficients included in the quantization coefficient Dllll, the time required for the dequantization process can be significantly shortened and the restoration process can be accelerated.

Note that the quantization coefficient holding unit 226 of the dequantization unit 220 shown in FIG. Good too. In this case, the effective coefficients are sequentially read out from the quantization coefficient holding section 226, and the quantization value corresponding to the effective coefficients is read out from the quantization dark value holding section 227 in accordance with the addresses sequentially output from the address holding section 223. All you have to do is read out the components of the matrix ■Ao.

〔Effect of the invention〕

As described above, according to the present invention, it is possible to selectively dequantize only the effective components included in the quantized coefficients, and the inverse quantization of the invalid components included in the quantized coefficients becomes possible. This method is extremely useful in practice because it can omit processing, reduce the number of multiplication operations, and perform inverse quantization processing at high speed.

[Brief explanation of drawings]

Fig. 1 is a principle block diagram of the image data restoration method of the present invention, Fig. 2 is a block diagram of a restoration device to which the image data restoration method according to an embodiment of the invention is applied, and Fig. 3 shows an example of decoded data. Figure 4 is an explanatory diagram of the address holding unit, Figure 5 is a configuration diagram of a conventional encoding device, Figure 6 is an explanatory diagram of blocks into which an image is divided, and Figure 7 is an explanatory diagram of the D
Figure 8 shows the quantization matrix VTH, Figure 9 shows the quantization coefficient D*u, Figure 10 shows the zigzag scan, and Figure 11 shows the conventional restoration. It is a block diagram of a device. In the figure, 111 is a decoding means, 120 is an inverse quantization means, 121 is a selection means, 122 is a storage means, 123 is a multiplication means, 141 is an orthogonal transformation means, 211.611 is a decoding unit, 212.612 is a decoding table, 220.620 is a dequantization unit, 221.621 is a demultiplexer, 222 is an address calculation unit, 223 is an address holding unit, 224.622 is a timing control unit, 225 is a counter, 226.628 is a quantization coefficient holding unit, 227.626 is a quantized dark value holding unit, 228.629 is a multiplier, 4 1 2 3 3 2 2 2 2 1.631 is an inverse DCT transformation unit, 1 is a DCT transformation unit, 0 is a linear quantization unit, 1 is an encoding unit, 2 is a code table, 3 is a multiplexer, 4 is a run length determination unit, 5 is a zero generator, and 7 is an address counter. FIG. 5 is an explanatory diagram of a conventional encoding device. FIG. 6 is a diagram showing D CT coefficients. FIG. 7 is a diagram showing quantization matrix ■TN. FIG. 8 is a diagram showing quantization coefficient ID. Figure 9 showing U Figure 9 Explanatory diagram of zigzag scan Figure

Claims (1)

[Claims] (1) Divide the image into multiple blocks each consisting of multiple pixels, quantize each component of the transform coefficient obtained by orthogonally transforming the pixel data of these blocks, and quantize the quantization generated by this quantization process. Encoded data generated by encoding coefficients is introduced, and the decoding means (111) decodes the encoded data to dequantize the obtained quantized coefficients, and the dequantization means (120) dequantizes the dequantized coefficients. In an image data restoration method in which image data is generated by orthogonally transforming transform coefficients generated by a dequantizing means (120) by an orthogonal transform means (141), the dequantizing means (120) comprises: the decoding means (111) Selection means (121
), storage means (12) storing quantization thresholds corresponding to each component of the transform coefficients and outputting the quantization thresholds corresponding to the valid components selected by the selection means (121);
2), multiplying the effective component selected by the selection means (121) by the component of the quantization threshold corresponding to the effective component output by the storage means (122), and applying this multiplication result to the An image data restoration method comprising: a multiplier (123) for outputting a transform coefficient as a component corresponding to an effective component;