JPH01237774A - Encoding method in compression of gradation picture data - Google Patents

Abstract

PURPOSE:To improve the encoding efficiency of compressing gradation picture data by separately using a prescribed two-encoding system for the unit of blocks. CONSTITUTION:The gradation picture data for the unit of blocks are orthogonally transformed by a two-dimensional discrete cosine transformation device 3 and go to be a transformation coefficient equivalent to one block. Then, the data are uniformly quantized by a first quantizing device 4 and the coefficient number and polarity number of a fixed length code are obtained. These pair numbers are stored in a block buffer memory 6 and a ranking device 5 distributes the block from the coefficient number into plural ranks. In a second quantizing device 7, quantization is executed by a quantizing width, which is determined based on a rank number and setting compression factor, and the cut rate of a high frequency component. Then, in an encoding device 8, the two systems of encoding are executed. Then, one system of the encoding uses both a Huffman code and run-length code and the other system of the encoding uses the Huffman code. In the block of the small quantizing width, or, in the block of a large high frequency cut degree, only the Huffman encoding is used. Then, a code length is shortened and the encoding efficiency is improved.

Description

Detailed Description of the Invention (Field of Industrial Application) The present invention is directed to an encoding method for compressing data of gradation images such as medical images such as X-ray photographs and CT images. Background) With the advancement of digital technology, gradation images are frequently being digitized for storage and transmission, as well as being subjected to various digital image processing. However, a gradation image has a larger amount of information than a binary image, and therefore, when digitizing a gradation image, the large amount of data becomes a problem. In medical images in particular, the number of pixels and bits required for each pixel when digitized is enormous, for example, 4 million pixels and 8 to 10 bits for a chest X-ray photograph, and it is difficult to efficiently store and transfer data. It's bad.

Therefore, today, data compression technology that compresses a huge amount of data of gradation images including medical images and converts them into data is in the spotlight.

Data compression techniques are broadly classified into reversible compression and irreversible compression, but reversible compression can only achieve a low compression rate of about 1/2 to 1/3, whereas Reversible compression methods, especially transform coding methods, are attracting attention.

Transform encoding is an irreversible compression method that divides the entire image into small blocks, performs orthogonal transformation on a block-by-block basis, and quantizes and encodes the resulting transform coefficients. This is the most suitable compression method for compressing .

In transform coding, it is known that the distribution of AC components of transform coefficients is approximated to a Gaussian distribution with a peak at zero, and the quantization of transform coefficients with such a distribution is Mid-riser type quantization including M i d -t including zero in the quantization output level
When classified into r a c e type quantization, M i d
It has been reported that -tr ace type quantization is more preferable because it causes less random noise within a block.

In order to obtain a high compression ratio in transform encoding, there are two methods: cutting high frequencies and increasing the quantization width.

When cutting high frequencies, there is a property that the larger the cutting rate, the more likely blurring will occur in the restored image. Also,
If you increase the quantization width without cutting high frequencies, M
i d -trace5 quantization has the property that component-specific pattern images are likely to occur (
Mid-rise! Quantization increases random noise), but in order to obtain high quality images, that is, restored images that are faithful to the original image, it is sufficient to quantize all frequency components with a small quantization width. As the amount increases, the compression ratio decreases.

Therefore, the present inventors investigated the relationship between the high frequency cut degree, quantization width, and image quality, and by adjusting the quantization width and high frequency cut degree according to the properties of each block, we achieved high fidelity with a high compression ratio. It was discovered that a high degree of restoration of the original image could be obtained.

On the other hand, encoding usually uses Huffman 9. A method is used in which the number of components whose quantization approaches zero is run-length encoded. However, when changing the quantization width and high frequency cut degree for each block, run-length encoding results in shorter blocks and run-length encoding.
There are blocks whose code length is shorter if the code length is shorter. In general, for blocks with a small quantization width or blocks with a large degree of high-frequency cut, it is better to use only Huffman encoding without run-length encoding, as the code length will be shorter and the encoding efficiency will be lower. will improve.

(Objective 3 and Structure of the Invention) The present invention has been made in view of the above points.It is an object of the present invention to improve encoding efficiency when compressing gradation image data including medical images using a transform encoding method. To achieve this goal, we prepared two encoding methods: one that uses both Huffman codes and run-length codes, and one that uses only Huffman codes. It is designed to be used in block units.

(Example) The present invention will be explained below based on two drawings.

FIG. 1 is a block diagram showing an embodiment of a gradation image data compression method using the encoding method according to the present invention. - In the figure, the frame memory l stores gradation image data to be compressed (in this example, the number of bits per pixel is 8 bits), and the reading device 2 first reads the image data from the frame memory l. The data is read out in blocks (in this example, the block size is 166 pixels each in the line direction and column direction), and the read block data is subjected to two-dimensional discrete cosine transformation (2D-
DCT) is cosine transformed by device 3 and is divided into 2
Obtain 56 transform coefficients.

Next, among the 256 transform coefficients obtained in this way, 255 AC components excluding one DC component are uniformly quantized by the first quantizer 4 with a basic quantization width W0, and each transform coefficient is Obtain the coefficient number and polarity number bare, which are fixed length codes, for each time. The bare coefficient numbers and polarity numbers are temporarily stored in the block buffer memory 6, and the coefficient numbers corresponding to ranks are sent to the device 5.

For rank, device 5 uses the first quantization device! The second Ek child conversion device 7 performs a process of sorting the blocks into a plurality of ranks based on the coefficient numbers for each block sent from the second Ek child conversion device 7.
Outputs the herank number.

The ff52ffi childization device 7 reads out the bare coefficient numbers and polarity numbers from the niblock buffer memory 6, and performs quantization with the quantization width and high frequency component cut rate determined based on the rank number and the degree of compression requested by the user. , the encoding device 8 is a second quantizer! The fixed-length code of the alternating current component obtained in step 7 is converted into a variable-length code and outputted to the terminal 9. Code data is transmitted from terminal 9 or stored in memory.

FIG. 2 is an example of the first quantization device 4. In FIG.

First, 255 alternating current components excluding the direct current component among the transform coefficients for terminals 41 or 61 blocks are input. The absolute values of the 255 AC components are taken by the absolute value circuit 42, and at the same time, the polarity determining circuit 43 determines whether the conversion coefficients are positive or negative. The polarity determination circuit 43 outputs the determination result to the terminal 47 as a polarity number. Here, the polarity number is represented by a 1-bit code j, and when the conversion coefficient is negative, it is 1'', and when it is positive, it is ``tsu''.

On the other hand, the absolute value of the conversion coefficient output from the absolute value circuit 42 is divided by the basic quantization width w0 by the division circuit 44. The 0 division result is rounded down to the decimal point by the truncation circuit 45 (this result is divided by the coefficient number ),
It is output to terminal 46 and terminal 47. The terminals 46 are connected to the device 5 for each rank, and the m child 47 is connected to the block buffer memory 6. The polarity number and coefficient number obtained for one transform coefficient inputted from the terminal 41 form a bare set, and are temporarily stored in the block buffer memory 6 from the terminal 47.

On the other hand, the DC component of each block has a sufficiently small quantization width wd
It is uniformly quantized at e, or it is temporarily stored in the block buffer memory 6 like the other 255 AC components without quantization.

FIG. 3 shows the first quantization of the AC component.

The horizontal axis is the value of the conversion coefficient/number, and the vertical axis is the frequency of occurrence of the conversion coefficient. The number sequence shown in FIG. 3(a) is a sequence of coefficient numbers output from the truncation circuit 45, and the number sequence shown in FIG. 3(b) is a sequence of polarity numbers j output from the polarity determination circuit 43. It is.

Next, FIG. 4 shows an example of the device 5 for the ranks when the number of ranks is 8.

The terminal 51 is connected to the terminal 46 in FIG. 2, and the coefficient number output from the first quantizer fi4 is input from here. The coefficient number k (k=0.1, 2,
3.4-) is an 8-bit counter control code 'Ct (i = 0, 1,
2, = , ? ) is converted to Figure 5 (a) shows counter control code table 5, 2. - Give an example.

Next, the counter control circuit 53 takes in the counter control code Ci, and controls the eight counters from counter O to counter 7 in the counter circuit 54 to each bit of C4.
/ o ff Controlled by state. Each bit of the counter control code C1 is set to counter 0 in order from the lower bit.
, counter l, counter 2. ... is associated with the counter 7, and when the bit is on (=1), the counter control circuit 53 increments the corresponding counter by +1, and when the bit is off (=O), the counter control circuit 53 increments the corresponding counter by +1. CI is shown in Figure 5 (b) without counting up.
The correspondence between each bit and each counter is shown.

FIG. 6 is a diagram showing the operating range of each counter. Each counter operates only for coefficient numbers k in the range indicated by the arrow in FIG. Since the counter 7 operates for all coefficient numbers, the count value of the counter 7 will be 255 at the end of the work for one block of ranks. When the count value of the carantuff reaches 255, the comparison circuit 55 starts comparison processing.

A threshold value Sp for the count value is applied to the comparator circuit 55 via a terminal 56 at the start of the compression process. rRmS
p is a function of the compression rate number P, and the compression rate number P is a parameter indicating the size of the compression rate requested by the user from the terminal (not shown). In this example, p=o, 2 of i If the user requests a high compression rate, l is assigned, and if the user requests a low compression rate, 0 is assigned. The threshold value Sp is set to satisfy the following relational expression.

0<Sp≦Number of total AC components If P*<P*, SPt≦SPt comparison circuit 55 compares the threshold value Sp with each count value, and sets a rank number m based on the comparison result.

An example of the comparison procedure of the comparison circuit 55 is shown in the flowchart of FIG.

First, the count value of the counter 6 and the threshold value Sp are compared (F-1), and if sp is larger than the count value of the counter 6, the rank number m is set to 7 (F-2), and the rank number m1 is input from the terminal 51. However, if Sp is less than the count value of counter 6, the count value of counter 5 and Sp7 are compared (F-3).

Thereafter, in the same manner, find the counter number where the count value is equal to or less than Sp and is the maximum, and set the rank number mr according to the counter number (F-1) to (F-14). In comparing the count value and Sp,
If Sp is less than or equal to the count value of counter 0, the rank number m,, is set to O (F-15), and when the rank number m1 is set, the rank number mr is output from the terminal 57 (F-16). ), the work for l blocks of ranks is completed, and the count values of each counter from counter O to counter 7 are cleared to zero (F-17).

FIG. 8 shows an example of the second quantizer M7 and the encoder 8.

FIG. 8(A) shows an example of the second quantization device fi7 and the encoding device 8 when the encoding method is selected before encoding, and FIG. 8(B) shows an example of the encoding method selected after encoding. An example of the encoding device 8 in the case of selection is shown.

First, the second quantization device 7 and encoding bag W8 shown in FIG. 8(a) will be explained.

The second quantizer g17 includes a condition setting section 70 that sets conditions for quantization and encoding, and a quantization section 71 that performs quantization.

The encoding device 8 includes a mode 0 encoding section 80 and a mode I encoding section 81. Here, mode 1 (code mode 1) refers to encoding that assigns one code to one component (Huffman encoding in this example) and run-length encoding that assigns one code to the number of components. Mode O (code mode 0) refers to an encoding method that uses only encoding that assigns one code to one component. Refers to the method.

First, the operation of the second quantizer fi7 shown in FIG. 8(a) will be explained.

At the initial condition setting stage at the start of the compression process, the terminal 70
The compression ratio number P from 2 is read into the condition setting section 70.

When the compression ratio number P is read, one of the two conversion tables, table A and table B, in the condition conversion table 704 is selected depending on the value of P. 0 The selected table is the image l to be compressed. Applies to all blocks of the frame.

FIGS. 9A and 9B show examples of tables A and B in the condition conversion table 704. In this example, P=0
When P=1, table A is selected, and when P=1, table B is selected. Further, the switch circuit 801 of the encoding device 8 reads the encoding mode and selects the destination of the quantized AC component data as either the mode 0 encoding unit 80 or the mode 1 encoding unit 81.
Set to one of the following. For example, in the case of y=O (code mode 0), the quantization unit 71 and the mode 0 encoding unit 80
If y=1 (code mode l), the quantizer 7
1 and the mode 1 encoding section 81 are connected.

When the initial condition setting stage is completed, compression processing is started for each block, and when the processing for ranks for l blocks is completed, the rank number m is read from the terminal 701 to the condition conversion table 04, and at the initial condition setting stage. Rank number m1 is converted into class number m6, quantization end component number X, and code mode y using either table A or table B in the selected condition conversion table 704. The class number mc is output to the terminal 703 and simultaneously read into the setting table 705 and converted into a quantization offset α and a quantization divisor β. $9 Figure (c) shows an example of the condition setting table 705. In the case of zero cases, α=
The following relationship holds true: m, ÷2, β=αX2.

The condition setting unit 70 reads the rank number m, and sets the class number me, the quantization end component number X, and the quantization offset α.
, quantization divisor β, and encoding mode y, quantization processing is started according to these parameters.

Next, quantization will be explained.

First, the data reading circuit 712 starts reading the AC component (the bare coefficient number and polarity number j) from the block buffer memory 6 via the terminal 711. At this time, the number of AC components read out is given by the quantization end component number X, and X AC components are read out in the order shown in FIG. 12 (a) or (b). The coefficient number k of the read AC component is taken into the addition circuit 713, and the quantization offset α set by the condition setting table 705 in the quantization condition setting section 70 is added to the second coefficient number k.
The second coefficient number '(k'' = +α) is taken into the division circuit 714 and divided by the quantization divisor β set by the 1st condition setting table 705 to obtain the code. Look at the number (i=int(k'/β)
) to obtain a fixed length code. The code numbers are sequentially sent to the encoding device 8 via the terminal 715. In addition, the polarity number j that is bare with the above code number is the data readout circuit 7.
12 and are sequentially sent to the encoding device 8 via a terminal 716.

Next, the operation of the encoding device 8 shown in FIG. 8(a) will be explained.

First, to explain the case where the code mode is O (y=o), the code number sentence output from the quantizer 71 of the quantizer 7 passes through the switch circuit 801 to the mote 0 encoder 80.
At the same time, the polarity number j which is bare with the code number is sent from the data reading circuit 712 of the quantization unit 71 to the mode 0 encoding unit 80. The code number sentence and polarity number j are converted into a variable length code hn3 and output from the terminal 815.

FIG. 11(c) shows a mode 0 Huffman code table 802.
An example is shown below.

As mentioned above, in the case of code mode O, the encoding unit device 8 generates X variable length codes h'zJ for χ code number sentences.
Output.

Next, the operation in the case of code mode 1 (y=1) will be explained using the flowchart of FIG. 1O.

When the code number output from the quantization unit 71 of 7 is read into the zero judgment circuit 811 in the mode I encoding unit 81 via the switch circuit 801 (P-1), the zero judgment circuit 811 Please make sure to judge the code number statement (P-
2) If the sentence is zero in the judgment of step (P-2), count up the zero count value Z of the zero counter 813 by +1 (P-3), if the sentence is not zero,
Determine whether zero count value 2 is zero (P-4)
, in the determination of step (P-4), if the zero count value Z is zero, the code number d and the polarity number j that is bare thereto are converted into a variable length code by the mote 1 Huffman code table 812 ( p-s) Output hzzj from the terminal 815 (P-6). In the judgment of step (P-4), if the zero count value 2 is not zero, the zero count value Z is converted into a variable length code r8 by the Bl code table 814. Convert to P-7), then clear the zero count value 2 to zero P-8), then output "8" from terminal 815 (P-9), step (P-5), step (P-8).
-6), the variable length code hfj for the code number sentence and polarity number j is output to the terminal 815. .

The above processing steps are classified into the following three steps.

(P-1) ↓ (P-2) (P-3) (P-4) (P-5) (P-7) Run length ↓
↓Count (P-6)
(P-8) Process ↓ Huffman code (P-9) Process
↓ (p-s) ↓ (P-6) Run-length Huffman encoding process Any of the above three processing processes is repeated for X components (code number sentences) for one pro-nk. (P-10), when any of the above processing steps for the If not zero, run length -
The first step (P-7) of the Huffman encoding process, (
After passing through steps P-8) and (P-9), the encoding work for l blocks is completed.

As explained above, in the case of code mode l, the encoder!
8 is a variable length code (h9
Or output "t).

FIG. 11(a) shows a mode 0 Huffman code table 802.
An example of the mode 1 Huffman code table 812 is shown in FIG. 11(b).

This example uses a Huffman code that is common to each class. When performing uniform quantization as in this example, there is a case where a Huffman code that is common to each class is used, and a Since there is no big difference in compression rate when Huffman codes are used, simpler device implementation is possible by using a common Huffman code for each class.

FIG. 11(c) shows an example of the B1 code table 814.

In this example, the Bl code common to each class is used as the run length code assigned to the zero count value 2. When 0-gradation image data is compressed using this method, the frequency distribution of each zero run length is the Bl code. Since the Bl code has an exponential distribution suitable for , it is possible to obtain a high compression rate by using the Bl code. In FIGS. 11(a) and 11(c), C3 and C are 1-bit codes that take values of 0 or 1 and satisfy ≠C.

Next, the encoding device 8' shown in FIG. 8(b) will be explained. Furthermore, the encoding device shown in FIG. 8 (a)! In order to distinguish it from is, it is referred to as encoding device 8'.

The encoding device 8' includes a mode 0 encoding section 80', a mode 1 encoding section 81', and a mode selection section 82'. Here, the mode O encoding section 80' and the mote I encoding section 81' have the same functions as the mote 0 encoding section 80 and the mode 1 encoding section 81 described in FIG. 8(A).

The terminal 801' is connected to a second quantizer 7' (not shown) from which the code number sentence is read. Kaso-g
17, but the second quantizer 7' does not output the mode number y to the encoder 8'.

The operation of the encoding device 8' shown in FIG. 8(b) will be explained below.

First, the code number sentence read from the terminal 801' is sent to the mode 0 encoding section 80' and the mode I encoding section 81'.
Encoding in code mode 1 and encoding in code mode 0 are performed in parallel. The variable length code h2 output from the mode 1 encoder 81' and the variable length code h2 output from the mode 0 encoder 80' are each output from the mode selector 82.
' is sequentially stored in the mode l code buffer 824' and the mode 0 code buffer 823'. At this time, mode 1 bit counter 822' calculates the cumulative number of bits of all variable length codes output from mode 1 encoding section 81', and mode 0 bit counter 821' calculates the cumulative number of bits of all variable length codes output from mode 0 encoding section 81'.
Find the cumulative number of bits of all variable-length codes output from '. When the encoding for l blocks is completed, the mode selection circuit 825' compares the count values (the cumulative number of bits of the code for one block) of the mote l bit counter 822' and the mode O bit counter 821'.

If the count value of the mote 0 bit counter 821' is larger than the count value of the mode 1 bit counter 822', the code mote 1 (y=1) is output to the terminal 827', and the code mote 1 (y=1) is output from the mode I code buffer 824'. One block worth of code string is output to terminal 826'.

On the other hand, if the count value of mode O bit counter 821' is less than the count value of mode 1 bit counter 822', code mode O (y=o) is output to terminal 827' and stored in mode 0 code buffer 823'. The code string for one block is output to the terminal 826'. The count value of each bit counter is cleared to zero when the comparison of the count values is completed.

FIG. 12 shows the data reading order of the data reading circuit 712 in FIG. 8.

Gradation painting t! 4) Since the conversion coefficient has a strong tendency for the amplitude to become smaller for high frequency components, the probability of occurrence of a component that is quantized to zero also increases for high frequency components.

Therefore, as shown in Figure 12 (a), if the missing part is (b),
If quantization and encoding are performed from the low frequency side toward the high frequency side, the probability that a long zero run will occur increases, and effective run length encoding can be performed. Note that when the quantization end component number is X, only X components from the low frequency side are read, and AC components from x+1th to 255th are not read.

FIG. 13 shows an example of the code structure.

FIG. 13(a) is an example of a code resulting from compressing the entire image of 1 frame, which consists of a header part, codes for each block following it, and a data end code.

FIG. 13(b) is an example of the header structure, and stores the block size (16 in this example), the number of blocks in the image line direction, the number of blocks in the image column direction, and the compression ratio number P.

FIG. 13(c) is an example of the code structure of each block,
Class number me (3-bit fixed length code in this example) Code mode y (fixed length code), DC component code (fixed length code)
, followed by an AC component code string (variable length code string). The AC component code string is q(I(1
≦9≦255) coefficient code string (variable length code string hIL, 7I).

hlzzz, ``1J1r'') and the run length code ゛ (variable length code rz) as shown in the same figure (e) are arranged alternately. Below are examples of various parameters used in this example. show,
Note that Ao-A is a constant indicating the operating range of each counter shown in FIG.

Basic quantization width WO=o, zs A o =0-50. A I = 0.75 A t
= 1.00. A3=1.25A4=1.50
．． AI! =1.75A6 =2. DD, At =2
．． 25S, =S, =200 In this example, when P;O, the compression ratio is 115, and when P=1, the compression ratio is around 1710.

Note that the above parameters are such that the amount of information per pixel of the original image is n bits (n = 8 in this example) and the block size is N x N.
− (N=16 in this example), the plotted image is f (
x, y), the transformation coefficient matrix F (u.

■) Two-dimensional discrete cosine transformation formula (
(shown in the following formula) is used.

Further, in the present invention, other orthogonal transform devices such as a quadratic complete Hadamard transform device, a slant transform device, etc. may be used instead of the two-dimensional discrete cosine transform device 3. Also, in the present invention, the block size is set to 16x16. The size is not limited to 8 x 8.32 x 32.
It is also applicable to sizes such as 64x64, and the number of bits per pixel of the original image is not limited to 8 bits.Also, the rank is determined by the device 5, when counting the coefficient number k within a specific section. Without making many steps in a specific section,
It is also possible to perform rankings based on the number of coefficient numbers existing within one specific section. Also,
In the present invention, it is also possible to perform classification by combining parameters for ranks with parameters for other classes.

(Effects of the Invention) As explained above, in the present invention, two encoding methods are prepared, one using both a Huffman code and a run-length code, and the other using only a Huffman code. When encoding transform coefficients (alternating current components) for 1 block, the encoding method that shortens the code length for 1 block is determined in advance as a function of the compression degree and quantization width, or two encoding methods are used. Either compare the code lengths after code death and select the encoding method with the shortest code length.
By selecting an encoding method with good encoding efficiency for each block, efficient compression can be performed.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a gradation image data compression method using the encoding method according to the present invention, FIG. 2 is a block diagram of an example of the first quantization device shown in FIG. 1, and FIG. Figure 4 is a diagram showing the coefficient number and polarity number obtained by quantizing the transform coefficient AC component in relation to the frequency of occurrence of the transform coefficient. Figure 5 (
(a) is a counter control code table, (b) is a diagram showing the correspondence between the counter control code and each counter, Figure 6 is a diagram showing the operating range of each counter in the counter control circuit, and Figure 7 is a diagram of the comparison circuit. Flowchart showing the comparison procedure, Figure 8 (
B) is a block diagram of an example of the second quantization device and the encoding device shown in FIG. 1 when the encoding method is selected before encoding, and FIG. The block diagrams of other examples of the encoding device in the case of selection, FIGS. 9(a) and (b), are the condition conversion table shown in FIG.
C) shows the quantization off, set α, and quantization divisor β when the addition circuit and division circuit shown in FIG. 8 are tabulated; FIG. Flowchart showing component encoding procedure, j1th
Figures (A) and (B) are examples of the Huffman code table in Figure 8, Figure (C) is an example of the B1 code table in Figure 8, and Figures 12 (A) and (B) are examples of the Huffman code table in Figure 8. A diagram showing the data read order in the data read circuit of
FIGS. 13(a), (b), (c), (d), and (e) are diagrams showing an example of the contents of the code structure. l...Frame memory, 2...Reading device, 3.
...Two-dimensional discrete cosine transform device, 4...
First quantizer, 5... Device for ranks, 6... Block buffer memory, 7-- Second quantizer, 8...
...The encoding part is

Claims (5)

[Claims] (1) In an encoding method for compressing gradation image data, in which digitized gradation image data is divided into a plurality of blocks, and transform coefficients obtained by performing orthogonal transformation on each block are quantized and encoded. A first encoding method uses two code sequences: a code sequence that assigns one variable-length code to each component, and a code sequence that assigns one variable-length code to each number of components; te1
The second method uses only code sequences that allocate two variable length codes.
The first encoding method is prepared for each block.
, a second encoding method, and encoding is performed using the selected encoding method. (2) The first,
The encoding method according to claim 1, further comprising a table indicating which of the second encoding methods is used, and determining the encoding method by referring to the table before encoding. . (3) The encoding method according to claim 1, wherein encoding is performed using two encoding methods, the code lengths are compared after encoding, and the encoding method with the shorter code length is selected. (4) The encoding method according to claim 1, wherein the code sequence that assigns one variable length code to one component is a Huffman code sequence. (5) The encoding method according to claim 1, wherein the code sequence in which one variable length code is assigned to each number of components is a B1 code sequence.