JPH07321665A - Coding method - Google Patents

Abstract

PURPOSE:To control code quantity through only offset control of a quantization section by allowing an offset processing section provided to the post-stage of an arithmetic section to add/subtract an offset to/from prescribed quantization data generated by the arithmetic section based on the setting of rough or accurate degree of quantization. CONSTITUTION:A digital image signal 1 is given to a DCT processing section 2 in which the signal is converted into DCT coefficient data 3 representing power distribution for each of divided areas. The data 3 are quantized by a quantization section 4, quantized data 5 are given to a variable length coding section 6, from which variable length code data 7 are outputted. The data 7 are fed to a transmission line via a buffer 8. Furthermore, a control section 11 detects the storage code quantity in the buffer 8 and generates a quantization parameter 12, which is provided as an output to the quantization section 4. On the other hand, an offset matrix section 21 of the quantization section 4 selects data corresponding to matrix data based on the value of a quantization parameter 12 and provides the output to an adder 16 as offset data 23, which are added to the quantization data.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving image compression coding system, and more particularly to generation code amount control.

[0002]

2. Description of the Related Art Generally, in a system in which data encoded using a variable length encoding method is transmitted using a transmission line having a predetermined transmission rate, the code amount control as described below is performed. is necessary. That is, the accumulated code amount of the difference between the transmission code amount per unit time obtained from the transmission rate of the transmission line and the generated code amount that constantly varies in the variable encoding unit is the buffer provided in the transmission line or both ends thereof. It is necessary to control so that the amount of code that can be stored by the means is not exceeded or that the amount of stored code is not less than the required minimum amount of stored code.

By the way, the generated code amount in the variable length coding unit for generating the code data transmitted on the transmission path is, for example, DCT (obtained by the discrete cosine transform of the video data in the preceding stage of the variable length coding unit. Discrete Cosine Transf
orm) When quantizing coefficient data, it depends on the setting value of the quantization parameter, which is a parameter for setting the coarse density of quantization, and in the example of image transmission, it depends on the fineness of the image to be transmitted. The generated code amount fluctuates depending on the amount of information contained in the image data before quantization, which is obtained by

From the above, as the conventional code amount control, the difference between the target code amount (the code amount corresponding to the transmission amount per unit time obtained from the transmission rate) and the generated code amount per unit time (or the accumulation of the differences) ), The quantization parameter is controlled so that the accumulated code amount of the difference is within a certain range.

However, in the case of transmitting a special input image, etc., the accumulated code amount of the difference does not necessarily converge within a given target accumulated code amount only by controlling the above-mentioned quantization parameter. In the accumulating means, an overflow operation or an underflow operation may occur.

[0006] For example, in the case of image data having a relatively large degree of redundancy, a simple data content before quantization, and a small amount of information contained in the data, such as image data of a completely black image, By the above code amount control, the quantization parameter is made small, the coarse density of quantization is made fine, and the value of the data after quantization is made large. Even if the number is increased, the generated code amount is still too small, and the accumulated code amount of the difference may not converge within a given target accumulated code amount range.

Conventionally, this problem is generated by adding a stuffing byte, which will be described later, to the code data output from the variable length coding unit in the latter stage of the variable length coding unit to compensate for the generated code amount. It solves the problem that the code amount becomes insufficient with respect to the target code amount, the accumulated code amount of the difference becomes less than a given target accumulated code amount, and does not converge to the range of the target accumulated code amount. . This stuffing byte is added to code data only for the purpose of increasing the generated code amount.
The world standard for image transmission, MPEG1 (ISO /
In the IEC JTC1 / SC2 / WG11) system, a stuffing byte system called macroblock stuffing is used.

FIG. 2 shows an example of a block configuration of a transmission unit of a transmission apparatus which realizes a conventional variable length coding system having means for adding stuffing bytes. In FIG. 2, first, in the DCT processing unit 32, the image signal 1 is input, and for example, 8 × in the horizontal and vertical directions of the image.
Discrete cosine transform is performed with a matrix of 8 pixels, and the horizontal and vertical frequency axes are divided into eight,
It is converted into DCT coefficient data 33 that represents the power (energy) distribution for each of the 64 divided areas. The DCT coefficient data 33 is quantized by the quantizer 34 to be quantized DCT coefficient data (hereinafter referred to as quantized data) 35. Further, the quantized data 35 is stored in the variable length coding unit 3
6 and is output as variable length code data 37. In addition, the stuffing byte generator 40 outputs the stuffing byte as output data 41.

The above-described variable length code data 37 is supplied to the selector 42 and the buffer 38, and the buffer 38 stores a part of the supplied variable length code data 37 under the control of the control unit 48. The selector 42 is supplied with the output data 39 of the buffer 38 and the output data 41 of the stuffing byte generator 40, in addition to the variable length code data 37.

Further, the selector 42, the buffer 38, and the stuffing byte generating section 40 are controlled by the control section 48 in the stuffing byte generating section 40 so that the timing at which the stuffing byte is generated is controlled. Selects the data to be accumulated and held among the data supplied to the buffer 38 and the output timing of the data. The control unit 48 controls which of the above three input data is selectively output to the selector 42 by taking control of the stuffing byte addition operation described later to control the output data amount of the selector 42. It can be the amount of data, including more than the amount of stuffing bytes.

The output 43 of the selector 42 is the buffer 44.
Are stored in the buffer 44 and are sent from the buffer 44 to a transmission line (not shown) at a subsequent stage at a speed matching the transmission rate.

Next, the quantization parameter control method of the control unit 48 will be described. The control unit 48 determines the amount of data accumulated in the buffer 44 per unit time as the buffer 4
4 is calculated from the number of clocks of the enable clock signal. Furthermore, a target code amount value 46, which is a code amount value related to the transmission rate and transmitted per unit time, is input from a target code amount setting means (not shown). The control unit 48
The calculated data storage amount per unit time is compared with the target code amount value 46, and as a result of this comparison, the output value of the quantization parameter 49 is controlled as in the following (1) to (3). Due to the change of the quantization parameter 49, if the data storage amount calculated after the change is in a range that does not require the addition of the above-mentioned stuffing byte, it approaches the target code amount and eventually becomes equal to the target code amount. Alternatively, the generated code amount is controlled so as to be substantially equal.

(1) (data storage amount> target code amount)
If so, the value of the quantization parameter 49 is increased and the quantization in the quantization unit 34 is made coarse. (2) If (data storage amount <target code amount), the value of the quantization parameter 49 is reduced, and the quantization in the quantization unit 34 is finer.
(3) If (data storage amount = target code amount), the value of the quantization parameter 49 is not changed.

Regarding the above-mentioned quantization parameter control means, the case where the generated code amount can be controlled within the range where the addition of the stuffing byte is not necessary has been explained. Furthermore, the quantization parameter 49 has the redundancy as described above. Even if the quantization parameter 49 reaches the minimum value, the accumulated data amount does not reach the target code amount and the accumulated code amount becomes a predetermined value. The stuffing byte addition operation newly performed in addition to the above operation when there is a possibility that the storage amount will be less than the range will be described below.

Even if the control unit 48 continues to output the quantization parameter 49 with the minimum value, the data storage amount per unit time accumulated in the buffer 44 calculated as described above is equal to the target code amount 46. When a relatively small amount of time has elapsed for a predetermined time, the stuffing byte generation unit 40 is instructed by the control signal 50 to generate a stuffing byte of an amount equal to or more than or close to the deficient data amount. At the same time, the control unit 48 controls the control signal 52
, The selector 42 selects the output data 41 from the stuffing byte generator 40, and instructs the selector 42 to output it. Note that this output data 41
Even while the selector 42 is being selected, the variable-length coding unit 36 may intermittently output the variable-length code data 37. In order to temporarily hold the data,
The buffer 38 holds the output of the variable length coding unit 36 while the selector 42 selects and outputs the stuffing byte as described above in response to the control signal 51 from the control unit 48.

Next, when the writing of the predetermined amount of stuffing bytes is completed, the selector 42 causes the buffer 38 to
The data 39 which is the output of is selected, and the accumulated data is output to the buffer 44 within a predetermined time. At the end of that, the selector 42 determines that the variable length code data 37
Is selected again and the data is output to the buffer 44.

By this operation, the stuffing byte is added to the data stored in the buffer 44 by the quantization parameter 49 so that the data storage amount per unit time approaches the target code amount.

[0018]

As described above, the prior art has a drawback in that a complicated control circuit for adding the stuffing byte is required.

In order to solve this drawback, the present invention does not use a stuffing byte generating means, a selector after variable length coding, a buffer before the selector for temporarily holding code data, etc. An object of the present invention is to enable code amount control by using the quantized data output from the quantizing unit in the preceding stage of the long coding unit.

[0020]

In order to achieve the above-mentioned object, the present invention provides a quantization unit which is generated based on a quantization parameter input from a control unit in a quantization unit before variable length coding. An offset value for increasing the amount of data generated in the variable-length coding unit in the subsequent stage is added to or subtracted from the subsequent data, and the quantized data obtained by adding or subtracting the offset value is subjected to the variable-length coding in the subsequent stage. It is designed to be output to the department.

[0021]

As a result, by providing the quantizing section with an offset function that can be configured by a simple circuit, when the value of the quantizing parameter becomes smaller than a predetermined value, the quantized data is quantized by the quantizing section. Since a predetermined offset value is added to and output to the variable length coding unit in the subsequent stage, when the amount of code generated in the variable length coding unit is smaller than when the offset value is not added, It is possible to increase the code amount generated in the long coding unit. Further, a means for adding a stuffing byte is not required in the subsequent stage of the variable length coding unit, and a complicated circuit therefor can be omitted.

[0022]

An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram showing an example of a block configuration of a transmission unit of a transmission device using a quantization unit according to the present invention. In FIG. 1, the digital image signal 1 is input to the DCT processing unit 2, and is converted into the DCT coefficient data 3 representing the power distribution for each divided area, as described above. The DCT coefficient data 3 is quantized by the quantizer 4 and becomes quantized data 5. Further, the quantized data 5 is input to the variable length coding unit 6 and output as the variable length code data 7. The variable-length code data 7 is accumulated in the buffer 8 and is sent from the buffer 8 to a transmission line (not shown) in the subsequent stage at a speed matching the transmission rate.

Further, the quantization parameter control unit 11 is
As described in the description of the conventional technique, the accumulated code amount of the buffer 8 is detected, and the target code amount setting means (not shown)
, The target code amount value 13 is input, the quantization parameter 12 is generated based on these, and the quantization parameter 12 is generated.
To the quantizer 4.

Next, FIG. 3 shows an example of a block configuration of the quantizer 4 having an offset function, which is a feature of the present invention.
In FIG. 3, the DCT coefficient data 3 output from the DCT processing unit 2 is input to the calculator 16 of the quantization unit 4. The quantization matrix section 27 outputs the quantization matrix coefficient data 17 held therein for determining the degree of quantization which is different for each area of the matrix, and the scaling matrix section 20. Outputs scaling data 22 for determining the quantization scale of the entire matrix, and the quantization matrix coefficient data 17 and scaling data 22 are input to each multiplier 18. The multiplier 18 multiplies these data and outputs the coefficient data 25 which is the multiplication result to the arithmetic unit 16 described above. It should be noted that the scaling matrix unit 20 inputs the quantization parameter 12, selects a matrix portion suitable for scaling based on the value of the input quantization parameter 12 from the data held in the scaling matrix unit 20, and stores it there. The scaling data 22 that is held is output.

Further, with respect to the quantization matrix unit 27, the scaling matrix unit 20, and the offset matrix unit 21 which will be described later, of the DCT coefficient data 3 input to the quantization unit 4 in a predetermined order, the data that is being input is exactly the data. Matrix position data 29, which is data for indicating which position on the matrix corresponds to, is input.
The data regarding the matrix positions in a predetermined order is output at the timing determined by the matrix position data 29.

Further, the matrix position data 29 is generated by the matrix position designating unit 28, and the matrix position designating unit 28 is inputted from the preceding stage (not shown) in synchronization with the DCT coefficient data 3 inputted to the quantizing unit 4. Matrix position data 29 is generated from the synchronization timing signal.

Further, the arithmetic unit 16 has the above-mentioned D
The CT coefficient data 3 and the coefficient data 25 are input, a calculation for quantization is performed, and the result is output as the quantized data 24 before offset.

Of the description of the block structure of the quantizer 4,
The above is a description of a portion (outside of the broken line portion in FIG. 3) that performs the same operation as that of the conventional technique.

The operation of the block shown in the broken line portion of FIG. 3 for realizing the offset function of the quantizer will be described below. The offset matrix unit 21 selects a corresponding value of the matrix data held in the offset matrix unit 21 based on the value of the quantization parameter 12, and outputs it as offset data 23 which is matrix data corresponding to the quantization matrix. The offset data 23 is input to the adder 26.

Here, for example, it is assumed that all the values of the matrix of the output offset value 23 are 0, and after that time, images having a relatively large degree of redundancy are continuously transmitted. As a result of quantizing images that are continuously input, for example, quantized data 24 before offset
When all the values of 0 are 0, the offset values 23 are all 0, so that the quantized data 5 output from the quantizer 4 is also 0, and is input to the variable length encoder 6 (FIG. 1) in the subsequent stage. To do. The variable-length coding unit 6 increases the generated code amount when the value of the quantized data 5 is larger relative to the value of the quantized data 5, and in this case, the quantized data 5 are all 0. Therefore, the generated code amount becomes the minimum value.

As a result, the value of the quantization parameter 12 becomes a minimum value or a predetermined value or less under the control of the quantization parameter control unit 11 (FIG. 1), and the quantization parameter 12 that is less than or equal to one of these values is offset. Matrix part 2
When inputting to 1, the offset matrix unit 21
For example, like the shaded area in the matrix of FIG. 5, the offset data 23 in which some offset values are values other than 0 (± 41 in the example of FIG. 5) is output.

Therefore, the offset data 23 having a non-zero value is added, and the quantized data 5 output from the quantizing unit is variable-length coded, so that only the offset data 23 having a non-zero value is added. Since it is possible to increase the generated code amount extra, the offset data 23
By controlling the above, it is possible to bring the amount of data accumulated in the buffer 44 per unit time close to the target code amount.

On the contrary to the above, when the image with relatively small redundancy is transmitted continuously and the quantization parameter 12 is larger than a predetermined value, all offset values are 0.
The value of the quantized data 24 before the offset remains as it is, and the data having the same value as the quantized data 24 before the offset remains as the quantized data 5 from the adder 26 in the variable length coding unit 6 (FIG. ) Is output.

As described above, the value of the quantized data 24 before the offset decreases for a certain period, for example, all become 0, the offset value is not added to the value, and the value remains as it is, and the variable length code in the subsequent stage. Even if the amount of code generated in the variable-length coding unit 6 decreases as a result of being sent to the quantization unit 6, and as a result the quantization parameter 12 becomes smaller, the value of the quantized data 24 before offset still decreases. Even if the quantization parameter 12 continues to decrease and there is a risk of underflow operation, before the underflow operation, the quantization parameter 12 becomes the minimum value or a predetermined value or less, and before the offset. Quantized data 24
By adding the offset value to, the code amount generated in the variable-length coding unit 6 can be set to an amount that does not cause the underflow operation.

In other words, as described above, the offset data 23 can be added, so that when the quantization parameter 12 becomes the minimum value or a predetermined value or less, It is possible to define the minimum value of the output data amount of the variable length coding unit 6. As a result, it is possible to eliminate the need for the additional operation of the stuffing byte, which requires a complicated control circuit as in the conventional case.

Next, the circuit configuration for realizing the quantizing unit 4 of the present invention will be described with reference to FIG. FIG. 4 is a diagram showing an embodiment of the circuit configuration of the quantizing unit 4 of the present invention. In FIG. 4, the DCT coefficient data 3 input to the quantization unit 4
Is input to a ROM (Read Only Memory) 60. Further, the matrix position data 62 is input to the ROM 60. The matrix position data 62 is data for indicating to which position in the matrix the DCT coefficient data 3 currently being input among the DCT coefficient data 3 input to the ROM 60 in a predetermined order corresponds to the DCT coefficient data. In synchronization with the data 3, it is generated by the matrix position designating circuit 61 to which the synchronization timing signal with the DCT coefficient data 3 is input. Further, the quantization parameter 12 from a control unit (not shown) is input to the ROM 60. RO
In M60, quantized data before offset having a predetermined value are stored at a plurality of predetermined address addresses. Explaining the operation of the ROM 60, the three signals of the DCT coefficient data 3, the matrix position data 62 and the quantization parameter 12 described above are address inputs and become the quantized data corresponding to the values of the input signals. The value is stored in the address designated by the address input, and the value is output as the quantized data 64 before the offset.

After this, the quantized data 64 before offset
Is input to the adder 65. Further, the sign bit 68 indicating either the positive or negative sign of the quantized data 64 before offset is input to the ROM 67 as an address input. Further, as in the ROM 60, the matrix position data 62 and the quantization parameter 12 are input to the ROM 67 as address inputs.

In the ROM 67, a value serving as an offset value corresponding to the values of the above three address input signals is stored at the address indicated by the above address input, and the values are offset data. It is output as 66. After that, the offset data 66 output from the ROM 67 is input to the adder 65, and is added to the quantized data 64 before the offset in the adder 65.

Here, the DCT coefficient data 3 is, for example,
It is assumed that the block is composed of a matrix of 8 rows × 8 columns. Further, the horizontal direction of the matrix is a discrete horizontal frequency axis, and the vertical direction is a discrete vertical frequency axis, and the DCT coefficient data 3 is for each frequency domain represented by these frequency axes. The power distribution of

Based on the above, further, an example of the offset data value set in the ROM 67 when the number of bits of the quantization parameter 12 is, for example, 4 bits (takes a value of 0 to 15) is shown in FIG. Shown in.

In FIG. 5, the set values in the ROM 67 are represented by a matrix of 8 rows × 8 columns for each of the horizontal frequency axis and the vertical frequency axis so as to correspond to the matrix configuration of the DCT coefficient data 3. The set values in the ROM 67 are set corresponding to the values 0 to 15 of the quantization parameter 12, respectively. In FIG. 5, a matrix structure is shown corresponding to the values 0, 1, 2, 3 and 15 of the quantization parameter 12. The value of the quantization parameter 12 is 4 to
The setting value corresponding to the case of 14 is the quantization parameter 12
Is the same as the set value corresponding to 3 and 15, and is omitted in FIG.

In FIG. 5, of the squared squares, the shaded squares have offset data values.
In this example, positive or negative offset data values of ± 41 are set, and all other squares not shaded are set to 0. It should be noted that the difference between the offset data value being positive and negative depends on whether the above-mentioned sign bit 68 represents positive or negative, and the offset data value having the same sign as the sign bit 68 is set. The setting including the value of ± 41 is made in the case of offset data corresponding to a small value of the quantization parameter 12, and in the example of FIG. 5, the quantization parameter 12 is 0 to 2 For the value of.
In addition, the smaller the value of the quantization parameter 12, the larger the number of squares for which a value of ± 41 other than 0 is set.

In this example, the offset data value whose value is not 0 is set to ± 41, which is a worldwide standard variable length coding system in the variable length coding unit in the subsequent stage, MPEG2.
(ISO / IEC JTC1 / SC29 / WG11 C
This is because the case where D) is used as an example is adapted to the code allocation method. FIG. 6 shows that MPEG2.
3 illustrates an example of a Huffman table for realizing the variable length coding method of FIG. The Huffman table shown in FIG. 6 is a table in which the RUN value is 0 to 63 in the horizontal direction and the absolute value of the LEVEL value is 0 to 127 in the vertical direction. Here, the RUN value is a value representing the so-called RUN length, which is the number of zero coefficient codes preceding immediately before the code, and the LEVEL value is, for example, the above-mentioned ROM.
It is the value of the quantized data 64 before offset of the non-zero code output by 60. In this table, the blank part is converted into code data including a header code called ESC code, and the length of this code data is a fixed length of 24 bits in this example. There is. In other words, in the case of the variable length coding using the Huffman table of FIG. 6, if the absolute value of the LEVEL value is 41 or more, the coded data will include the ESC code regardless of the RUN length. Code length must be 24
Become a bit. In addition, the part of the star in the table is a variable length code code other than the code data including the ESC code, and the code length is 24 bits or less.

From the above, the variable length coding unit 6 in the subsequent stage
If the absolute value of the LEVEL value of the data to be input to is set to 41 or more, the amount of variable length code generated by the variable length coding unit 6 per coded data is fixed to 24 bits in this example. it can. Therefore, the offset data value ± 41 is added to a predetermined number of data (corresponding to the number of shaded squares in FIG. 5) of the 8 × 8 matrix to obtain the data after the addition. L
The absolute value of the EVEL value is always set to 41 or more and is output to the variable length coding unit 6 in the subsequent stage, so that the code generation amount corresponding to a certain block is always set to the predetermined number × 24 bits or more. You can

In the 8 × 8 matrix, the arrangement of the data whose offset data value is ± 41 (corresponding to the shaded squares in FIG. 5) on the matrix is set from the high frequency region side. I have to. In general, if offset data is added to the low frequency region side, the original dynamic range may easily be exceeded. However, since the value corresponding to the high frequency component of the quantized data 64 before offset is almost 0 in a normal image, even if the offset data is added, there is a harmful effect (overflow or underflow with respect to the dynamic range). It can prevent the occurrence.

Here, the operation described above will be considered by considering an extreme example in which the degree of redundancy is large with a specific number of bits. In this case, it is assumed that the quantized data 64 before offset is all zero. If the offset values shown in FIG. 5 are added, the value of the quantization parameter 12 is set to 0, and the quantized data 5 has three +41 in one block.
Value exists.

The fact that there are three values of +41 for each block means that a code of 72 bits (24 bits × 3) is always generated. Therefore, when converted into frame units, the number is 583200 bits. Furthermore, the minimum code generation rate is 17.496 Mbps.

The calculation process is shown below.

Number of pixels per frame: 518400 bytes {720 pixels (luminance component / horizontal) × 480 pixels (luminance component / vertical) +360 pixels (color difference component / horizontal) × 240 pixels (color difference component / vertical) × 2} per frame Number of blocks: 8100 {Number of pixels per frame / 64 bytes (number of pixels per block)} Minimum code amount per frame: 583200 bits {8100 × 72 bits (Minimum code amount per block)} Minimum code generation rate : 17.496 Mbps {minimum code amount per frame × 30 frames (number of frames per second)} Due to the above, in the example of FIG. 5, an image with extremely high redundancy is continuously input, and the code in the variable length coding unit 6 is input. When the amount of generation decreases and the value of the quantization parameter 12 becomes 0, the code generation rate in the variable length coding unit 6 is set to Can be 17.496Mbps, it is possible to increase the amount of generated code in the variable length coding unit. Therefore, only by controlling the quantization parameter 12,
It is possible to control the code amount so that underflow does not occur, and there is no need to use the stuffing byte method, which requires a complicated and special control circuit as in the prior art.

FIG. 7 shows a second embodiment of the circuit structure of the quantizing unit of the present invention. FIG. 7 shows the ROM 6 of the embodiment shown in FIG.
In this embodiment, the functions of 0 and the ROM 67 are combined into one ROM 70.

In FIG. 7, the number of offset data and its value can be adjusted according to the transmission capacity and the content of variable length coding. In addition, each ROM in FIGS.
It can be realized even if it is replaced with AM (Random Access Memory) or another storage element.

In the dequantization operation at the time of decoding, there is no problem if the same value as the added offset data value is subtracted and then the same processing as the conventional one is performed.

If the added offset data has no visual effect on the quality of the decoded image,
In the dequantization operation at the time of decoding, the same value as the offset data may not be subtracted and the same process as the dequantization of the conventional technique may be performed.

In the above description, the case where the discrete cosine transform is performed before the quantizer and the DCT coefficient which is the transform result is used as the input data of the quantizer has been described as an example. Also, there is no problem even if a configuration using another conversion method such as a wavelet conversion method instead of the discrete cosine conversion is used.

[0055]

According to the present invention, without using the stuffing byte system which requires a complicated control circuit,
For image input of any degree of redundancy, it is possible to realize code amount control that does not cause the buffer to underflow, only by controlling by the quantization parameter.

[Brief description of drawings]

FIG. 1 is a diagram showing a block configuration example of a transmission unit of a transmission device using a quantization unit according to the present invention.

FIG. 2 is a diagram showing a block configuration example of a transmission unit of a conventional transmission device.

FIG. 3 is a diagram showing a block configuration example of a quantization unit of the present invention.

FIG. 4 is a diagram showing an embodiment of a circuit configuration of a quantization unit of the present invention.

FIG. 5 is a diagram showing an example of offset data values in a ROM 67.

FIG. 6 is a diagram showing an example of a Huffman table.

FIG. 7 is a diagram showing a second embodiment of the circuit configuration of the quantization unit of the present invention.

[Explanation of symbols]

 1 Image Signal 2 DCT Processing Section 3 DCT Coefficient Data 4, 34 Quantization Section 5 Quantization Data 6, 36 Variable Length Coding Section 7 Variable Length Code Data 8, 38, 44 Buffer 11 Quantization Parameter Control Section 12 Quantization Parameter 13 target code amount value 16 arithmetic unit 17 quantization matrix coefficient data 18 multiplier 20 scaling matrix unit 21 offset matrix unit 22 scaling data 23 offset matrix data 24, 64 quantization data before offset 25 coefficient data 26, 65 adder 27 Quantization matrix unit 28 Matrix position designating unit 60, 67, 70 ROM 61 Matrix position designating circuit 62 Matrix position data 66 Offset data 68 Sign bit

Claims (4)

[Claims] 1. An encoding method using a variable length encoding method, wherein an input data is quantized by using an arithmetic unit capable of changing a coarse density of quantization to generate quantized data. In the quantization method, an offset processing unit is provided at a stage subsequent to the calculation unit, and the offset processing unit adds and subtracts the offset value to the predetermined quantized data generated by the calculation unit based on the setting of the coarse density of quantization. An encoding method comprising: 2. The encoding method according to claim 1, wherein the offset value is added to or subtracted from the quantized data based on a coarse density of quantization and a matrix arrangement of input data composed of a matrix. The encoding method according to claim 1, wherein 3. In the above encoding method, input data composed of the matrix is DCT (Discret).
e Cosine Transform) coefficient data, and one or more data arranged in a higher frequency region in the horizontal frequency axis direction and the vertical frequency axis direction of the matrix of the data obtained by quantizing the DCT coefficient data are selected, The encoding method according to claim 2, wherein the offset value is added to or subtracted from the data by the offset processing unit. 4. The encoding method according to claim 2, wherein the selection range of the matrix arrangement of the quantized data to which the offset value is added and subtracted is controlled based on the setting of the coarse density of the quantization. 3. The encoding method described in 3.