JP2856974B2 - Image coding device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image coding apparatus capable of high-speed processing.

[0002]

2. Description of the Related Art In recent years, digital image processing has been actively performed. In particular, various schemes for standardization have been proposed for high-efficiency encoding for compressing image data. The high-efficiency encoding technique encodes image data at a smaller bit rate in order to improve the efficiency of digital transmission and recording. As such a high-efficiency coding method, there are a predictive coding technique and an orthogonal coding technique ("Multidimensional processing of TV images", written by Takahiko Fukinuki, detailed by Nikkan Kogyo Shimbun). Further, by performing variable-length coding on the codes compressed by these coding, further image compression is possible. Variable-length coding changes the bit width of coding in accordance with the frequency of occurrence of codes, and can reduce the bit rate compared to fixed-length codes.

However, when variable length coding is employed, the amount of compression code differs depending on the picture. Therefore, when recording compressed data on a storage medium, it is necessary to know the capacity of the storage medium required for recording before recording. And storage media cannot be used efficiently. For this reason, in order to make the amount of compression code constant, a method of fixing a variable length code to a fixed length may be adopted.

FIG. 3 is a block diagram showing a conventional image coding apparatus for converting such a variable length code into a fixed length code.

The frame memory 1 stores digital image data. The block dividing circuit 2 reads image data from the frame memory 1 and divides the image data into sub-blocks of 8 horizontal pixels × 8 vertical pixels.
Output to the CT calculator 4. The sampling circuit 3 performs a thinning process on the sub-block data, and the DCT operation circuit 4 performs a DCT (discrete cosine transform) process on the input sub-block data and outputs a transform coefficient to the quantization circuit 5. The quantization circuit 5 quantizes the transform coefficient based on the quantization coefficient from the selection circuit 6. The components below a predetermined level are deleted by quantization, and the data is compressed and output. Furthermore,
The quantized output is subjected to variable-length coding (for example, Huffman coding) by the coding circuit 7, so that the code amount is further reduced and output.

As described above, the code length of one screen changes due to Huffman coding and also changes due to quantization. In a general image, when the quantization coefficient is increased, the probability that 0 appears as a quantization output increases, and the code length (total number of bits) decreases. As described above, by controlling the quantization coefficient, the encoded output can be fixed length. However, the total number of bits when the quantization coefficient is changed differs for each image, and the total number of bits is not determined until after the coding is completed. Therefore, in the apparatus of FIG. 3, the quantization coefficient is appropriately changed by the quantization coefficient generation circuit 8 in the first scan before the original encoding, and the total number of bits after the encoding is determined is determined. The conversion rate is calculated to achieve a constant rate.

FIG. 4 is a block diagram showing a specific configuration of the quantization coefficient generation circuit 8.

The basic quantization table Qk stores basic quantization coefficients. The multipliers 10, 11, 12, 13 multiply the basic quantization coefficients by normalization factors α0 to α2, αt, respectively, and store the multiplication results in quantization tables Q0 to Q2, Qt.
The selection circuit 6 selects a quantization coefficient stored in one of the quantization tables Q0 to Q2, Qt of the quantization coefficient generation circuit 8 and outputs it to the quantization circuit 5.

That is, in the first scan, the selection circuit 6 first selects the quantization table Q0. In this case, the DCT operation circuit 4 performs DCT processing only on the sub-sampled block data from the sampling circuit 3 and supplies a transform coefficient to the quantization circuit 5. Quantization circuit 5
Quantizes the transform coefficient using the quantization coefficient from the quantization table Q0. The code length calculation circuit 15 calculates the code length when the quantized output is subjected to variable length coding in the coding circuit 7, and supplies the code length to the integration circuit 17 via the selection circuit 16. Integrator circuit
17 accumulates the code length of each block obtained by the code length calculation circuit 15, obtains the total number of bits F0 for one screen, and outputs it to the quantization characteristic determination circuit 20.

Next, the selection circuit 6 selects the quantization table Q1. The quantization circuit 5 converts the transform coefficients into a quantization table Q
Quantize using the quantization coefficient from 1. The code length calculation circuit 15 calculates the code length in this case and supplies the code length to the integration circuit 18 via the selection circuit 16. The integration circuit 18 accumulates the code length of each block and outputs the total number of bits F1 for one screen to the quantization characteristic determination circuit 20.

Next, the selection circuit 6 sets the quantization table Q2
And gives the quantization coefficient stored in the quantization table Q2 to the quantization circuit 5. The quantization circuit 5 quantizes the transform coefficient, and the code length calculation circuit 15 obtains a code length when the quantized output is subjected to variable length coding. This code length is selected by the circuit
16 to the integrating circuit 19 to provide the total bit length F for one screen
2 is given to the quantization characteristic determination circuit 20. The quantization characteristic determination circuit 20 determines the quantization characteristic based on the total bit lengths F0 to F2.

FIG. 5 is a graph for explaining the operation of the quantization characteristic determination circuit 20 by taking the normalization factor on the horizontal axis and the code amount of the sampled image on the vertical axis.

When the normalization factor of the quantization coefficient generating circuit 8 is increased, the compression ratio increases, and when the normalization factor is reduced, the compression ratio decreases. As a result, the code amount of the image sampled by changing the normalization factor changes as shown by the curve K in FIG. The normalization factor of the quantization coefficient generation circuit 8 is α0 <α1 <α2, and as described above, the code amount when the normalization factor is α0 to α2 is F0 to F2, respectively.

Now, let Ft be a value obtained by correcting the target code amount for one screen in accordance with the number of sampling blocks. Let αt be a normalization factor for obtaining the corrected target code amount Ft, and F0
If <Ft <F1, α0 <αt as shown in FIG.
<Α1 holds. In this case, the quantization characteristic determination circuit 20
For example, assuming that the change of the normalization factor with respect to the correction target code amount Ft is linear, an equation of a straight line L is derived from the normalization factors α0, α1 and the code amounts F0, F1, and is shown by a broken line in FIG. Thus, the normalization factor αt for the corrected target code amount Ft is approximately obtained. Note that the quantization characteristic determination circuit 20
In some cases, the normalization factor αt is obtained by obtaining the equation of the curve K in FIG.

That is, the quantization characteristic determination circuit 20 determines that Fi <F among the code amounts F0 to F2 from the integration circuits 17 to 19.
The code amounts Fi and Fi + 1 satisfying t <Fi + 1 (i = 0, 1) are selected, and from these values, the normalization factor .alpha.t used for the original encoding is approximately obtained by calculation. By increasing the number of quantization tables and the number of integration circuits of the quantization coefficient generation circuit 8 to reduce the interval between the normalization factors, the accuracy of calculating the normalization factor αt can be improved. . The normalization factor αt is given to the multiplier 13 of the quantization coefficient generation circuit 8. The multiplier 13 has a basic quantization table Qk
Is multiplied by the normalization factor αt and stored in the quantization table Qt.

Next, the original encoding is performed on all the image data. The image data read from the frame memory 1 is divided into
It is given to the CT operation circuit 4. The DCT operation circuit 4 performs a DCT process on the image data and supplies the image data to the quantization circuit 5. In this case, the selection circuit 6 has selected the quantization table Qt,
The quantization circuit 5 quantizes the transform coefficients using the quantization coefficients from the quantization table Qt. The coding circuit 7 performs variable length coding on the quantized output and outputs a coded output. Thus, the code amount of the encoded output is made to match the target code amount.

However, the quantization characteristic determination circuit 20 obtains the normalization factor αt by an approximate calculation, and the calculated normalization factor αt has an error. In addition, in order to determine the quantization characteristics, quantization coefficients from a plurality of quantization tables are sequentially selected, and quantization is sequentially performed using a plurality of quantization coefficients. The number is also limited, and the error of the calculated normalization factor αt becomes relatively large. Therefore, if the original encoding is performed using the calculated normalization factor αt, the target code amount may be exceeded.

Such an excess of the target code amount can be suppressed by optimizing the variable length code according to the processed image and improving the coding efficiency. In this method, a frequency distribution is obtained for each processed image. As described above, the encoding circuit 7 assigns data having a short data length to data having a high frequency of occurrence, and assigns data having a long data length to data having a low frequency of occurrence. The quantization output is converted into a variable length code by referring to the table. However, the frequency of occurrence of each data of the quantized output differs for each processing image, and when variable-length coding is performed using an initial variable-length code obtained in advance by a training method using a sample image, coding efficiency is reduced. The probability that the encoded output exceeds the target code amount increases.

Therefore, the frequency distribution of each data of the quantized output quantized by using the quantized coefficient used for the original encoding is obtained for each processing image, and the initial variable length code is corrected to obtain the optimum value. That is, a method of performing variable length coding is adopted. However, in order to perform optimal variable-length encoding, it is necessary to determine the frequency distribution after determining the optimal quantization characteristic used for the original encoding.
There is a problem that the processing time for scanning increases.

[0020]

As described above, in the above-mentioned conventional image coding apparatus, when the optimum variable length coding is performed in consideration of exceeding the target code amount due to the calculation error of the normalization factor. However, there is a problem that the processing time increases by one scan.

An object of the present invention is to provide an image coding apparatus capable of optimizing variable length coding without increasing processing time.

[0022]

According to a first aspect of the present invention, there is provided an image coding apparatus comprising: an orthogonal transform means for orthogonally transforming a digital image signal in predetermined sub-block units to output transform coefficients; A first quantization circuit that quantizes a plurality of quantization outputs with a plurality of predetermined quantization characteristics and outputs a quantized output, and a case where each of the quantized outputs based on the plurality of quantization characteristics is variable-length encoded by a predetermined variable-length code Code length calculating means for calculating the code length of each of the code lengths, and a quantization characteristic determination for obtaining a determined quantization characteristic by calculating a quantization characteristic by operation from each code length obtained by the code length calculation means and a target code length. Means,
The occurrence frequency distribution of the data of the quantized output from the first quantizing circuit is calculated for each of the plurality of quantization characteristics using the quantized output from the first quantizing circuit. When the determined quantization characteristic is detected by the quantization characteristic determining means, the occurrence frequency distribution obtained for the quantized output when quantized by the first and second quantization characteristics adjacent to the determined quantization characteristic is calculated. A frequency for predicting the occurrence frequency distribution of the quantized output at the time of the original encoding by performing an operation based on the difference between the determined quantization characteristic and the first and second quantization characteristics on the occurrence frequency distribution. Distribution generating means;
A code generation circuit for generating a variable length code based on the occurrence frequency distribution predicted by the frequency distribution generating means, and a second circuit for quantizing the transform coefficient based on the determined quantization characteristic determined by the quantization characteristic determining means. And the second quantization circuit
A coding circuit that performs variable-length coding on the quantized output from the quantizing circuit using a variable-length code generated based on the prediction value of the frequency distribution generating means, and outputs a coded output. An image encoding apparatus according to claim 2 of the present invention, comprising: a first encoding unit that orthogonally transforms a digital image signal in predetermined sub-block units and outputs transform coefficients;
Orthogonal transformation means, a second orthogonal transformation means for sub-sampling the sub-block and orthogonally transforming only a predetermined sub-block to output a transformation coefficient, and a transformation coefficient from the second orthogonal transformation means A first quantization circuit that quantizes with a plurality of quantization characteristics and outputs a quantized output, and a code in the case where each of the quantized outputs based on the plurality of quantization characteristics is variable-length coded by a predetermined variable-length code. Code length calculating means for respectively determining the length, and quantization characteristic determining means for determining a quantization characteristic by operation from each code length obtained by the code length calculating means and a target code length to obtain a determined quantization characteristic; Calculating the occurrence frequency distribution of the data of the quantization output from the first quantization circuit for each of the plurality of quantization characteristics using the quantization output from the first quantization circuit; Determining quantizer characteristics Wherein the determining the quantization characteristics is detected by the first and second adjacent to this decision quantization characteristic
Using the occurrence frequency distribution obtained for the quantized output when quantized with the quantization characteristic of the above, the calculation based on the difference between the determined quantization characteristic and the first and second quantization characteristics is performed by the occurrence frequency distribution. A frequency distribution generating means for predicting the occurrence frequency distribution of the quantized output at the time of the original encoding, a code generation circuit for generating a variable length code based on the occurrence frequency distribution predicted by the frequency distribution generating means, A second quantization circuit that quantizes the transform coefficient from the first orthogonal transform means based on the determined quantization characteristic determined by the quantization characteristic determination means, and a quantum from the second quantization circuit. And a coding circuit for performing variable length coding on the coded output using the variable length code generated by the code generation circuit based on the predicted value of the frequency distribution generating means, and outputting a coded output.

[0023]

In the present invention, the first quantization circuit quantizes the transform coefficient with a plurality of quantization characteristics. The code length calculation means calculates the code length of each quantized output, and the quantization characteristic determination means determines an optimum quantization characteristic from each code length and a target code length. The frequency distribution generation means obtains a frequency distribution corresponding to each quantization characteristic simultaneously with the calculation of the code length by the code length calculation means. When a code length approximating the target code length is obtained by the code length calculation means, the code generation circuit generates a variable length code based on a frequency distribution corresponding to the code length. The second quantization circuit quantizes the transform coefficient with the calculated optimum quantization characteristic, and the encoding circuit encodes the quantized output using the variable length code from the code generation circuit. The code generation circuit generates a variable-length code based on the quantized output based on the quantization characteristics that approximate the optimal quantization characteristics.There is no need to perform quantization using the optimal quantization characteristics. Going in a short time.

[0024]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing one embodiment of an image encoding device according to the present invention. 1, the same components as those in FIG. 3 are denoted by the same reference numerals.

The frame memory 1 stores digital image data. The block division circuit 2 reads data from the frame memory 1 in units of sub-blocks of, for example, 8 × 8 pixels and outputs sub-block data. The sub-block data from the block divider 2 is supplied to a sampling circuit 3 and a DCT operation 4. The sampling circuit 3 sub-samples the sub-block data and outputs it to the DCT operation circuit 4. The DCT operation circuit 4 performs DCT processing on the data from the sampling circuit 3 when determining the quantization characteristics and outputs the data to the quantization circuit 5, and performs DCT processing on the output of the block dividing circuit 2 and performs quantization processing on the quantization circuit 5 during the original encoding. Output.

The quantization circuit 5 receives the quantization coefficient from the selection circuit 6, quantizes the DCT transform coefficient, and
23, output to the code length calculation circuit 15 and the frequency distribution generation circuit 21. The quantization coefficient is created by the quantization coefficient generation circuit 8. The configuration of the quantization coefficient generation circuit 8 is the same as that shown in FIG. 4, and is obtained by multiplying the basic quantization coefficients stored in the basic quantization table Qk by normalization factors α0 to α2 and normalization factor αt. The quantization coefficients are stored in quantization tables Q0 to Q2 and Qt, respectively. The selection circuit 6 sequentially selects the quantization tables Q0 to Q2 of the quantization coefficient generation circuit 8 when the quantization characteristic is determined, and selects the quantization table Qt at the time of the original encoding to send the quantization coefficients to the quantization circuit 5. To give.

The code length calculation circuit 15 calculates the code length when the quantized output is variable-length coded by the coding circuit 23 when the quantization characteristics are determined, and the integration circuits 17 to 19 via the selection circuit 16. Output to The selection circuit 16 operates in conjunction with the selection circuit 6, and the selection circuit 6 generates the quantization tables Q0 to Q0.
When 2 is selected, the integration circuits 17 to 19 are selected, respectively. The integration circuits 17 to 19 accumulate the code length from the code length calculation circuit 15 for one screen, obtain the code amounts F0 to F2 of the sampled image data, and output them to the quantization characteristic determination circuit 24.

Assuming that the corrected target code amount obtained by correcting the target code amount based on the sampling number is Ft, the quantization characteristic determination circuit 24 determines the code amounts F0 to F2 from the integration circuits 17 to 19.
Of these, code amounts Fi and Fi + 1 satisfying Fi <Ft <Fi + 1 (i = 0, 1) are selected, and from these values, a normalization factor .alpha.t used for the original coding is approximately calculated. And outputs it to the quantization coefficient generation circuit 8. Further, in the present embodiment, the quantization characteristic determination circuit 24 outputs information on the normalization factors αt and i to the frequency distribution generation circuit 21.

The normalizing factors αi and αi + 1 corresponding to the code amounts Fi and Fi + 1 respectively have a relationship of αi <αt <αi + 1, and among the normalizing factors α1 and α2, The two closest values are obtained. Therefore, the normalization factors αi,
The frequency distribution of the quantized output when the quantization coefficient corresponding to αi + 1 is used is similar to the frequency distribution of the quantized output when the normalization factor αt is used. For this reason, the frequency distribution generating circuit 21 outputs the frequency distribution from the quantized output when the normalization factors αi and αi + 1 are used.

FIG. 2 is an explanatory diagram for explaining the frequency distribution generating circuit 21. FIGS. 2A to 2C show normalization factors α.
The magnitude relationship between i, αt, and αi + 1 is shown.

The magnitude relationship shown in FIG. 2A, that is, the normalization factor is compared with the difference between the normalization factors αt and αi + 1 according to the information of i from the quantization characteristic determination circuit 24 and the normalization factor αt. When it is shown that the difference between αi and αt is extremely small, the frequency distribution generation circuit 21 outputs the occurrence frequency distribution of each data obtained for the quantized output when the normalization factor αi is used. Conversely, as shown in FIG. 2B, when the difference between the normalization factors αt and αi + 1 is extremely small compared to the difference between the normalization factors αi and αt, the frequency distribution generation circuit 21 Outputs the occurrence frequency distribution of each data obtained for the quantized output when the normalization factor αi + 1 is used. Also, as shown in FIG. 2C, the normalization factors αi,
If the difference between αt and the difference between the normalization factors αt and αi + 1 are substantially equal, the frequency distribution generation circuit 21 uses the normalization factor αi
, Αi + 1, and outputs the frequency distribution obtained for the quantized output. The frequency distribution output from the frequency distribution generation circuit 21 is provided to the code generation circuit 22. Sign generation circuit
22 generates a variable-length code based on the input frequency distribution when determining the quantization characteristic, and outputs the variable-length code to the encoding circuit 23.

The quantization coefficient generation circuit 8 stores the quantization coefficient obtained by multiplying the basic quantization coefficient by the normalization factor αt in the quantization table Qt, and via the selection circuit 6 during the original encoding. Output. The encoding circuit 23 converts the quantized output at the time of the original encoding into a variable length code using the variable length code from the code generation circuit 22, and outputs the encoded output.

Next, the operation of the embodiment configured as described above will be described.

First, in order to determine the quantization characteristic, the first
In scanning, image data read from the frame memory 1 is divided by the block dividing circuit 2 and supplied to the sampling circuit 3 in units of 8 × 8 pixel sub-blocks. The sampling circuit 3 thins out the block data and outputs only the sampled block data to the DCT operation circuit 4. The DCT operation circuit 4 performs DCT processing on the output of the sampling circuit 3 and outputs the result to the quantization circuit 5.

In the first scan, the selection circuit 6 sequentially selects the quantization tables Q0 to Q2. First, the quantization circuit 5 quantizes the DCT transform coefficient using the quantization coefficient from the quantization table Q0, and outputs a quantized output to the code length calculation circuit 15 and the frequency distribution generation circuit 21. The code length calculation circuit 15 obtains a code length when the input quantized output is encoded by the encoding circuit 23, and outputs the code length to the selection circuit 16. The selection circuit 16 operates in conjunction with the selection circuit 6, and in this case, gives the code length to the integration circuit 17. The integrating circuit 17 accumulates the code length for the sampling block (for one screen) to obtain the code amount F0 and outputs it to the quantization characteristic determination circuit 24.

Similarly, the quantization circuit 5 outputs the quantization table Q
Quantization using the quantization coefficients from 1 and Q2 is also performed. The code length of the quantization output when using the quantization coefficient from the quantization table Q1 is supplied to the integration circuit 18 via the selection circuit 16, and the quantization when using the quantization coefficient from the quantization table Q2 is performed. The code length of the output is determined by the integration circuit 19 via the selection circuit 16.
To supply. The code amounts F1 and F2 for one screen are supplied to the quantization characteristic determination circuit 24 from the integration circuits 18 and 19, respectively.

The quantization characteristic determination circuit 24 calculates the corrected target code amount F
Select Fi and Fi + 1 such that t is Fi <Ft <Fi + 1,
Using the code amount Ft and the selected code amounts Fi and Fi + 1, a normalization factor .alpha.t used for the original coding is obtained. The quantization characteristic determination circuit 24 converts the normalization factor αt into a quantization coefficient generation circuit 8.
And the information of i and the normalization factor αt are given to the frequency distribution generating circuit 21. The frequency distribution generating circuit 21 uses the outputs of the quantization tables Q0 to Q2 based on the magnitude relation between the normalizing factors αi, αi + 1 and αt corresponding to the code amounts Fi, Fi + 1, respectively. Of the quantized outputs
A frequency distribution obtained from one or two quantized outputs is output. Since at least one of the normalization factors αi and αi + 1 is a value approximate to the normalization factor αt, the frequency distribution generation circuit 21
Can obtain the same frequency distribution as in the original encoding. The code generation circuit 22 generates a variable length code based on the frequency distribution from the frequency distribution generation circuit 21.

Next, in the second scan, the original encoding process is performed. That is, the image data read from the frame memory 1 is divided by the block dividing circuit 2 and applied to the DCT operation circuit 4 in sub-block units. DCT
The arithmetic circuit 4 performs a DCT process on the sub-block data and supplies a transform coefficient to the quantization circuit 5. In this case, the quantization coefficient based on the normalization factor αt calculated in the first scan is stored in the quantization table Qt.
Calculates the quantization coefficient from the quantization table Qt into the quantization circuit 5
Give to.

The quantization circuit 5 quantizes the transform coefficient and outputs the result to the encoding circuit 23. The encoding circuit 23 converts the quantized output into a variable length code using the variable length code corrected by the code generation circuit 22 in the first scan, and outputs the result. The variable length code is corrected by obtaining the frequency distribution for each processed image, and optimal variable length coding is possible.

As described above, in the present embodiment, it is detected that the quantization process has been performed using the quantization coefficient based on the normalization factor approximated to the normalization factor αt used for the original encoding. A variable length code corresponding to each processed image is created by obtaining the frequency distribution of the quantized output by this quantization process in one scan. This makes it possible to perform optimal variable-length coding on the quantized output in the second scan, improve coding efficiency, and suppress the target code amount from being exceeded due to an error in determining the quantization characteristic. Can be. The frequency distribution is obtained in the first scan, not after the determination of the quantization characteristics, and the optimum variable length coding can be performed without increasing the time required for coding.

[0041]

As described above, according to the present invention, there is an effect that the variable length coding can be optimized without increasing the processing time.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of an image encoding device according to the present invention.

FIG. 2 is an explanatory diagram illustrating an embodiment.

FIG. 3 is a block diagram showing a conventional image encoding device.

FIG. 4 is a block diagram showing a specific configuration of a quantization coefficient generation circuit.

FIG. 5 is a graph for explaining the operation of the quantization characteristic determination circuit.

[Explanation of symbols]

2 ... Block dividing circuit, 3 ... Sampling circuit, 4 ... D
CT circuit, 5 quantization circuit, 6, 16 selection circuit, 8 quantization coefficient generation circuit, 15 code length calculation circuit, 17 to 19 integration circuit, 21 frequency distribution generation circuit, 22 code generation circuit ,twenty three
... Encoding circuit, 24 ... Quantization characteristic determination circuit, Q0 to Q2
, Qt ... quantization table,

Claims (2)

(57) [Claims] An orthogonal transform unit for orthogonally transforming a digital image signal in predetermined sub-block units to output a transform coefficient; and quantizing the transform coefficient by a plurality of predetermined quantization characteristics to output a quantized output. A first quantization circuit; code length calculation means for respectively obtaining a code length when each of the quantized outputs based on the plurality of quantization characteristics is variable-length coded by a predetermined variable-length code; Quantization is determined by determining the quantization characteristics by operation from the obtained code lengths and the target code length .
Quantization characteristic determining means for obtaining characteristics; and generation of quantized output data from the first quantization circuit.
The frequency distribution is calculated using the quantized output from the first quantizing circuit.
In addition to calculating for each of the plurality of quantization characteristics,
The determined quantization characteristic is detected by the quantization characteristic determination means.
When issued, the first and second adjacent
About the quantized output when quantizing with the quantization characteristic of 2
Using the determined occurrence frequency distribution, the decision quantization characteristics and
Before the calculation based on the difference between the first and second quantization characteristics,
Applied to the occurrence frequency distribution to calculate the quantum
Frequency distribution generating means for predicting the occurrence frequency distribution of the coded output, a code generation circuit for generating a variable length code based on the occurrence frequency distribution predicted by the frequency distribution generating means, and a decision determined by the quantization characteristic determining means A second quantization circuit that quantizes the transform coefficient based on a quantization characteristic; and a code output circuit that outputs a quantized output from the second quantization circuit based on a predicted value of the frequency distribution generation unit. An encoding circuit for performing variable length encoding using the generated variable length code and outputting an encoded output. 2. A first orthogonal transformation means for orthogonally transforming a digital image signal in a predetermined sub-block unit and outputting a transform coefficient, and sub-sampling the sub-block and orthogonally transforming only the predetermined sub-block. A second orthogonal transform unit that outputs a transform coefficient, a first quantizer that quantizes the transform coefficient from the second orthogonal transform unit with a plurality of predetermined quantization characteristics and outputs a quantized output, Code length calculating means for respectively obtaining a code length when each of the quantized outputs based on the plurality of quantization characteristics is variable-length coded by a predetermined variable-length code; and each code length and the target obtained by the code length calculating means. Determines the quantization characteristics by calculation from the code length to be calculated and determines the quantization
Quantization characteristic determining means for obtaining characteristics; and generation of quantized output data from the first quantization circuit.
The frequency distribution is calculated using the quantized output from the first quantizing circuit.
In addition to calculating for each of the plurality of quantization characteristics,
The determined quantization characteristic is detected by the quantization characteristic determination means.
When issued, the first and second adjacent
About the quantized output when quantizing with the quantization characteristic of 2
Using the determined occurrence frequency distribution, the decision quantization characteristics and
Before the calculation based on the difference between the first and second quantization characteristics,
Applied to the occurrence frequency distribution to calculate the quantum
Frequency distribution generating means for predicting the occurrence frequency distribution of the coded output, a code generation circuit for generating a variable length code based on the occurrence frequency distribution predicted by the frequency distribution generating means, and a decision determined by the quantization characteristic determining means A second quantization circuit for quantizing a transform coefficient from the first orthogonal transform means on the basis of a quantization characteristic; and a code output circuit which outputs a quantized output from the second quantization circuit to the frequency distribution. An encoding circuit for performing variable-length encoding using a variable-length code generated based on a prediction value of a generating unit and outputting an encoded output.