JPH08214170A - Image compressor - Google Patents

Abstract

PURPOSE: To attain image compression in response to image quality of each image by deciding a quantization coefficient corresponding to a spatial frequency. CONSTITUTION: DCT data statistics obtained corresponding to the quantized DCT coefficients using a quantization table whose quantization coefficients are all 1 as a default and a set sum code quantity are input to a spatial frequency data quantity setting section 24. The spatial frequency data quantity setting section 24 sets set data quantity for each spatial frequency. A quantized table generating section 25 predicts a data quantity of coding data in the case of coding through the use of prescribed quantization coefficients. When the data quantity is less than the set data quantity set by the spatial frequency data quantity setting section 24, the quantization coefficients thus determined are adopted as the quantization table.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention
The present invention relates to an image compression device that compresses information based on the EG algorithm.

[0002]

2. Description of the Related Art A standardized algorithm for encoding high-resolution images and exchanging information via a communication transmission line is JPE.
Recommended by G (Joint Photographic Expert Group). Algorithm recommended by this JPEG,
That is, in the baseline process of the JPEG algorithm, since a large amount of information is compressed, first, the two-dimensional DC is
The original image data is decomposed into components on the spatial frequency axis by T conversion, each data represented on the spatial frequency axis is quantized using a quantization table, and each quantized data is encoded. . JPEG recommends a predetermined Huffman table for this encoding.

In a conventional image compression apparatus, a default quantization table is usually used, and if necessary, a quantization table modified by multiplying each quantization coefficient of the quantization table by a single coefficient is used. Has been created.

[0004]

As described above, since the modified quantization table is obtained by uniformly multiplying each spatial frequency by a coefficient, the characteristics of each image (for example, a low frequency component Therefore, it is difficult to say that the image data is compressed without degrading the image quality because it is not possible to perform the quantization in accordance with the characteristics such as a lot of high frequency components.

In view of the above problems, it is an object of the present invention to provide an image compression apparatus capable of achieving image compression according to the characteristics of individual images.

[0006]

An image compression apparatus according to the present invention comprises an orthogonal transform means for performing an orthogonal transform on original image data to obtain an orthogonal transform coefficient for each spatial frequency, and an orthogonal transform coefficient for a predetermined quantization coefficient. Quantizing means for quantizing a quantized orthogonal transform coefficient by a quantization table consisting of, and quantized orthogonal transform coefficients are rearranged in a predetermined one-dimensional array data with respect to spatial frequency, and then based on the quantized orthogonal transform coefficient. A coding unit that performs coding to obtain coded data, a unit that sets a target value of the data amount of the coded data for each spatial frequency, and a first unit based on the statistical amount of the coded data of the first spatial frequency A means for predicting the data amount of encoded data of the spatial frequency of 2 and a quantization coefficient corresponding to the spatial frequency so that the predicted data amount of the second spatial frequency is equal to or less than the target value. It is characterized in that a Mel quantization coefficient calculating means.

[0007]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below with reference to illustrated embodiments. FIG. 1 is a block diagram of an image compression apparatus according to an embodiment of the present invention.

The light coming from the subject S is collected by the condenser lens 11
The image of the subject is collected by the CCD (solid-state image sensor)
An image is formed on 12 light receiving surfaces. A large number of photoelectric conversion elements are arranged on the light receiving surface of the CCD 12, and color filters composed of, for example, R, G, and B color filter elements are provided on the upper surface of the photoelectric conversion elements. Each photoelectric conversion element corresponds to one pixel. The subject image is converted into an electric signal corresponding to a predetermined color by each photoelectric conversion element, and A
It is input to the / D converter 13. In the configuration of FIG. 1, C
Although there is only one CD 12, it is also possible to have a configuration in which two or more CCDs are provided.

The signal A / D converted by the A / D converter 13 is converted into a luminance signal Y by a signal processing circuit (not shown).
Is converted into color difference signals Cb and Cr and input to the image memory 14. The image memory 14 is divided into mutually independent memory areas for storing the luminance signal Y and the color difference signals Cb, Cr, and each memory area has a storage capacity for one image.

The luminance signal Y and the color difference signals Cb and Cr read from the image memory 14 are input to the DCT processing circuit 21 for data compression processing. In the DCT processing circuit 21, the original image data such as the luminance signal Y is discrete cosine transformed (hereinafter referred to as DCT). That is, in this embodiment, the DCT transform is used as the orthogonal transform of the original image data. Although the DCT processing circuit 21 is shown as one processing circuit in FIG. 1, an independent DCT processing circuit is actually provided for each of the luminance signal Y and the color difference signals Cb and Cr.

The image compression apparatus includes a DCT processing circuit 21, a quantization processing circuit 22, a Huffman coding processing circuit 23, a spatial frequency data amount setting unit 24, and a quantization table generation unit 25.
Etc. DCT processing circuit 21, quantization processing circuit 22
In the Huffman coding processing circuit 23, the image data such as the luminance signal Y is divided into a plurality of blocks for one screen and processed in block units. Each block is 8x
It consists of 8 pixel data.

The DCT coefficients of the luminance signal Y and the color difference signals Cb and Cr obtained by the DCT processing circuit 21 are input to the quantization processing circuit 22, respectively. Similarly to the DCT processing circuit 21, the quantization processing circuit 22 is also provided for each signal. The DCT coefficients of the luminance signal Y and the color difference signals Cb and Cr input to the quantization processing circuit 22 are quantized by the quantization table Q1 composed of 8 × 8 quantization coefficients. This quantization is a linear quantization, ie each DCT coefficient is divided by the corresponding quantization coefficient.

In this embodiment, the quantization table Q1 for quantizing the DCT coefficient of the luminance signal Y and the DCT of the color difference signals Cb and Cr are based on the JPEG algorithm.
Although different from the quantization table Q1 for quantizing the coefficients, the same quantization table Q1 may be used for each signal. As will be described later, these quantization tables Q1 are generated optimally by the spatial frequency data amount setting unit 24 and the quantization table generating unit 25 according to the characteristics such as the spatial frequency distribution of the original image data.

The quantized DCT coefficients of the luminance signal Y and the color difference signals Cb and Cr output from the quantization processing circuit 22 are input to the Huffman coding processing circuit 23, and the Huffman table Q is supplied.
2 is used, and Huffman coding is performed by a predetermined algorithm.

The image signal (compressed image data) obtained by Huffman coding is recorded on a recording medium such as an IC memory card.

FIG. 2 shows, as an example, image data P (Y) xy of a block of 8 × 8 pixels, DCT coefficient S (Y) uv, quantized DCT coefficient R (Y) uv, and quantization table Q. (Y) uv is shown.

The image data P (Y) xy in FIG. 2 (a) is subjected to a two-dimensional DCT transformation to obtain 8 × 8 = 64 shown in FIG. 2 (b).
Are converted into individual DCT coefficients S (Y) uv. These DCT
Of the coefficients, the DCT coefficient S (Y) _{00 at the} position (0,0)
Is a DC component, and the remaining 63 DCT coefficients S (Y) uv are AC components. The AC component indicates from the coefficient S (Y) ₀₁ or the coefficient S (Y) ₁₀ to the coefficient S (Y) ₇₇ how much higher spatial frequency component is in the image data of the 8 × 8 pixel block. There is. The DC component represents the average value (DC component) of the pixel values of the entire 8 × 8 pixel block. That is, each DCT coefficient S (Y) uv corresponds to a predetermined spatial frequency.

FIG. 2D shows an example of the quantization table Q (Y) uv used in the quantization processing circuit 21. Such a quantization table Q (Y) uv may be different for the luminance signal Y and the color difference signals Cb and Cr as described above. The quantization table Q (Y) uv is the quantization table Q (Y) uv used at the position corresponding to each signal when the JPEG format image data is recorded on the recording medium. The contents are recorded.

The equation for quantizing the DCT coefficient S (Y) uv using the quantization table Q (Y) uv is defined as follows. R (Y) uv = round (S (Y) uv / Q (Y) uv) {0≤u, v≤
7} Round in this formula means an approximation to the nearest integer. That is, the quantized DCT coefficient R (Y) uv as shown in FIG. 2C is obtained by dividing and rounding off each element of the DCT coefficient S (Y) uv and the quantization table Q (Y) uv. To be

The quantized DCT coefficients R (Y) uv, R (Cb) uv, R obtained in the quantization processing circuit 22 in this way.
(Cr) uv is input to the Huffman coding processing circuit 23.

Next, the Huffman coding in the Huffman coding processing circuit 23 will be described with reference to FIGS. In the following description, the quantized DC coefficient means a quantized DC component, and the quantized AC coefficient means a quantized AC component.

Quantized DC coefficient R (Y) ₀₀ and quantized AC coefficient (quantized DCT coefficient R other than quantized DC coefficient R (Y) ₀₀ )
(Y) uv) has a different encoding method. Encoding of the quantized DC coefficient R (Y) ₀₀ is performed as follows.

[0023] First, the difference between the current quantized DC coefficient of a block to be coded R (Y) ₀₀ and one previously coded blocks quantized DC coefficient R (Y) ₀₀ is obtained. It is determined which of the categories shown in FIG. 3 the difference value belongs to, and the code word representing the category is obtained from the code table (DC component coding table) shown in FIG. For example, the quantization DC coefficient R of a block to be currently coded (Y) ₀₀ is "16", the quantization of the coded block before one DC coefficient R (Y) ₀₀ is "25" At some time, the difference value is "-9", and therefore the category to which the difference value = -9 belongs is determined to be "4" from the category table of Fig. 3, and the codeword of that category is determined from the code table of Fig. 4. It is judged as "101".

Next, in the category table of FIG. 3, the number of values in the category is shown by an additional bit. For example, since the difference value = −9 is the seventh smallest from the smallest in the group of category = 4, the additional bit is “0110”. That is, the Huffman code word of the quantized DC coefficient R (Y) ₀₀ of the block currently being encoded is “1010110”.

On the other hand, the coding of the quantized AC coefficient is performed by the processing routine shown in FIG. First step 12
At 0, 63 quantized AC coefficients are zigzag scanned in the order shown in FIG. 6 and rearranged into one-dimensional array data. The quantized AC coefficients will be referred to as scans AC ₁ , AC _2, ... AC _{63 in} the order of zigzag scanning. Next, at step 122, it is judged whether or not each quantized AC coefficient arranged in one dimension is "0". When the quantized AC coefficient is “0”, step 12
In 4, the number of consecutive quantized AC coefficients that are “0” is counted. As a result, the length of continuous "0", that is, the run length is obtained.

On the other hand, when it is determined in step 122 that the quantized AC coefficient is not "0", step 1
At 26, the same grouping as the quantized DC coefficients is performed and additional bits are determined. Unlike the grouping of the quantized DC coefficient, the grouping of the quantized AC coefficient is performed for the quantized AC coefficient itself. That is, when the quantized AC coefficient is "4", for example, the category "3" is obtained by referring to the table shown in FIG. Further, since the quantized AC coefficient “4” is fifth from the smallest in the group of category = 3, the additional bit is “100”.

In step 130, the AC code table (FIG. 8) of the Huffman table is referred to. For example, when the run length of the data immediately before the quantized AC coefficient “4” is “0”, this run length and category = Based on 3 and 3, the codeword "100"
Is obtained. And this code word "100" and step 1
The two-dimensional Huffman code word “100100” is obtained by combining the additional bits “100” obtained in step 26.

The result of Huffman coding the quantized DCT coefficient shown in FIG. 2C is shown as coded data HF in FIG.

FIG. 10 shows the encoded data HF shown in FIG. 9 again. Such encoded data HF is obtained for each block, and when one screen is composed of 5400 blocks, only 5400 encoded data HF is obtained. As described above, the encoded data HF includes the encoded data for one quantized DC coefficient and 63 quantized A
And C coded data.

The coded data relating to the quantized DC coefficient is
The code word FA0 of the category and the additional bit FB0. The encoded data relating to the quantized AC coefficient is the run length
It consists of a codeword of the category and additional bits. next,
The encoded data regarding the quantized AC coefficient will be described in more detail.

In the example of FIG. 9, the code word FA1 of the run length / category and the additional bit F are coded data based on the scan AC ₁ of 4 and the run length of 0.
B1 is generated, the scan AC ₂ is -7, and the coded data based on the run length is 0, the additional bit FB2 is generated code word FA2 run length category. Since the scan AC ₃ is 0, after the additional bit FB2, the code word FA4 of the run length / category and the additional bit are coded data based on the scan AC ₄ being 3 and the run length being 1. FB4 has been generated. Similarly, the run length / category code word F
A5, additional bit FB5, code word F of run length / category
A8, additional bit FB8, run length / category code word F
A9 and additional bit FB9 are generated respectively. The end data (EOB) indicates that all 0s continue after the scan AC ₁₀ .

Next, the generation of the quantization table Q1 will be described. As shown in FIG. 1, the quantization table Q1 is generated based on the set total code amount and the DCT data statistic. The set total code amount is the total number of bits of the encoded data HF for one screen recorded on the recording medium, for example, 5
It is 24288 bits (64 Kbytes). The DCT data statistic is obtained based on the output data of the DCT processing circuit 21. The output data of the DCT processing circuit 21 is obtained by performing DCT conversion on the original image data, and a quantization table Q1 (hereinafter, referred to as a default quantization table) in which all 8 × 8 quantization coefficients are “1” is used. It is equivalent to what is quantized. The output data was zigzag scanned (see FIG. 6) to obtain the run length and category (see FIG. 5).
Then, the DCT data statistic is obtained by referring to the table shown in FIG. 24 to obtain the bit length for each spatial frequency. That is, this DCT data statistic is a category distribution as shown in FIG. 11 and a code amount distribution for each scan as shown in FIG. 12, as described below.

FIG. 11 shows a predetermined spatial frequency (for example, scan AC ₁ ) when the default quantization table is used.
3 shows an example of the category distribution in, that is, the number of blocks classified into each category. The category distribution as shown in this figure is generated in one image for DC components and 63 AC components, that is, only 64 category distributions are generated in total. In the present embodiment, the original image data is divided into 5400 blocks, and in the example of FIG. 11, the number of blocks whose category is 0 is 602, and the number of blocks whose category is 1 is 1088.

FIG. 12 shows an example of the distribution of the number of bits of the coded data HF in each scan when the default quantization table is used. This is the total for each spatial frequency for all blocks of one image. Indicates the number of bits. For example, in the case of the DC component, the total number of bits of the code word FA0 of the category and the additional bit FB0 (see FIG. 10) is 42.
238 (reference DB0). For scan AC ₁ ,
The total number of bits of the run length / category code word FA1 and additional bit FB1 is 34,090 (code DB1), scan A
In the case of C ₂ , the total number of bits of the run length / category code word FA2 and the additional bit FB2 is 33833 (code DB2)
Is. In the example of FIG. 10, but not present code word and additional bit run length category in the scan AC _3, because of the presence data of the scan AC ₃ is In another block, the total number of bits becomes 25 920 (reference numeral DB3) ing. In this way, data up to the total bit number 20909 (reference DB ₆₃₎ of the scan AC ₆₃ is generated.

The set total code amount and the DCT data statistic are
It is input to the spatial frequency data amount setting unit 24. The spatial frequency data amount setting unit 24 sets the distribution of the code amount for each spatial frequency (that is, for each scan) as shown in FIG. 13, for example. That is, the code amount of the DC component is 44
240 bits (code SB0), the code amount of scan AC ₁ is 33008 bits (code SB1), the code amount of scan AC ₂ is 35629 bits (code SB2), and the code amount of scan AC ₆₃ is set (code SB63). ).
The total value of these code amounts, that is, the set total code amount is set to a predetermined value (524288 bits in the example of FIG. 13).

This code amount (data amount) is a target value,
In this image compression device, the quantization table Q is set so that the code amount for each spatial frequency becomes the target value as described later.
1 is generated. That is, the distribution of the code amount is the target value of the code amount for each spatial frequency in the Huffman encoded data when the finally obtained quantization table Q1 is used, and for example, it is desired to cut high frequency components. In this case, the code amount for the high frequency component is set to be relatively small. The code amount at each spatial frequency is obtained, for example, by trial and error so that the reproduced image becomes closer to the original image.

In the spatial frequency data amount setting unit 24,
The upper limit of the code amount for each spatial frequency may be determined using the distribution of the code amount for each scan of the DCT data statistic (see FIG. 12). For example, although the total number of bits of scan AC ₄ is set to 28844 (symbol SB4) in FIG. 13, it may be limited to 28461 bits based on the code amount distribution of DCT data statistics (the symbol of FIG. 12). See DB4).

The quantization table generation unit 25 generates each quantization coefficient that constitutes the quantization table Q1 based on the DCT data statistics and the input data from the spatial frequency data amount setting unit 24.

A method of obtaining the quantization coefficient for the DC component will be described. First, using the default quantization table (quantization table in which all quantization coefficients are 1),
The DC components of all blocks are quantized. Then, the quantization D of the block for which the quantization coefficient is currently obtained
The difference value between the C coefficient and the quantized DC coefficient of the immediately preceding block is obtained.

It is determined which of the categories shown in FIG. 3 the difference value belongs to, and the code word representing the category is obtained from the code table (DC component coding table) shown in FIG. Further, the number of additional bits corresponding to the difference value is obtained from the category table of FIG. For example, when the category of the difference value is “2”, the code length is 3 bits and the number of additional bits is 2 bits. Therefore, the code amount of the quantized DC coefficient whose category of difference value is “2” is 5 bits.

In this way, the code amount of the quantized DC coefficient is obtained for each block, and the total code amount (the number of bits) of these is obtained. This total code amount is D shown in FIG.
If the code amount of the C component (code SB0) or less, the quantized coefficient at that time is determined as the final one. If the total value is larger than the code amount (code SB0) in FIG. 13, the quantization coefficient is changed to 2, and the total number of bits as described above is examined using this quantization coefficient. Be seen.

In order to obtain the total number of bits when this new quantized coefficient is used, it is necessary to predict the number of blocks corresponding to each category. That is, FIG.
It is necessary to obtain a category distribution table as shown in FIG. An example of how to obtain this category distribution table is shown in FIG.
Will be described next.

For example, the difference value belonging to the category “2” is
-3, -2, 2, and 3. When the quantization coefficient becomes 2,
Since the difference value “2” is 2/2 = 1, the category “1”
Move on to. The same applies to the difference value "-2". On the other hand, since the difference value “3” is 3/2 = 1.5≈2, it remains the category “2”. The same applies to the difference value "-3". Therefore, in this example, of the difference values that belonged to the category “2” when the quantization coefficient was 1, half changed to the category “1” and half remained the category “2”. In this way, the category distribution when the quantization coefficient changes is predicted, and the total code amount is obtained by using the table of this category distribution.

Next, how to obtain the quantized coefficient for the AC component will be described. Regarding the AC component, the Huffman-encoded data includes data relating to the run length (for example, symbols FA1, FA2, FA4 in FIG. 10), and the run length changes when the quantization coefficient is changed. Therefore, the data amount of each scan when the quantized coefficient is changed cannot be predicted based on only the data amount of the scan before the quantized coefficient is changed. Therefore, in the present embodiment, as will be described below, the data amount of each scan when the quantization coefficient is changed is predicted using the category distribution table, and the quantization coefficient is determined based on this prediction value. ing.

FIG. 14 is an example of a table simultaneously showing the category distribution shown in FIG. 11 for each scan. That is, this category distribution table shows the number of blocks classified into each category at each spatial frequency when the default quantization table is used. In FIG. 14, the DC component and scan AC ₁ to scan AC _{11 are} shown, and scan AC ₁₂ to scan AC ₆₃ are omitted. Also, in FIG. 14, the numbers at the top indicate categories. For example, in scan AC ₁ ,
The number of blocks in category "0" is 602, and the category is "1"
The number of blocks is 1088.

Unlike FIG. 14, FIG. 15 shows a category distribution table when a quantization table in which all the quantization coefficients are “16” is used. As can be seen from FIG. 7, the AC component values of the categories “0” to “3” when the quantized coefficient “1” is used become 0 when the quantized coefficient “16” is used. The category of the AC component value is “0”. That is, the scan AC in FIG.
Blocks of 3479 categories “0” of ₁ are shown in FIG.
602 and 108 of categories “0” to “3” in
It corresponds to each block of 8, 1184, and 605.

Similarly, the AC component value of the category "4" when the quantized coefficient "1" is used becomes -1 or 1 when the quantized coefficient "16" is used, so that the category of the AC component value is It becomes "1". On the other hand, when the quantization coefficient “1” is used, the AC component value of the category “5” is the quantization coefficient “1”.
When "6" is used, there are one that becomes 1 and one that becomes 2.
Therefore, when the quantization coefficient “16” is used, the AC component value category of some blocks is “1”, and the AC component value categories of other blocks are “2”. That is, the 866 blocks of category “1” of scan AC ₁ in FIG. 15 correspond to the blocks of 529 of category “4” and some blocks of category “5” in FIG. 14. In this way, when the quantized coefficient is changed, a block belonging to a certain category does not necessarily move to another category as it is, but may move to a different category.

Next, the prediction of the number of blocks in which the categories of the scans AC _i-1 and AC _i are both "0" will be described with reference to FIG.
This will be described with reference to FIGS.

16 to 19, C

[0] is the number of blocks of category “0” in the scan, C [1
~] Is the number of blocks of category “1” or more in the scan. Z [k] is a predicted value of the number of blocks whose run length is k. C '

[0] is the number of blocks of category “0” in the previous scan, Z ′

[0] is the number of blocks whose run length is 0 in the immediately preceding scan.

FIG. 16 shows the category "0" of scan AC ₁ .
And the number of blocks of category “1” or more. In scan AC ₁ , as shown in FIG.
Number of blocks in category "0" C

[0] is 3479, and the number of blocks C [1] of category “1” and above is 19
21 (= 5400-3479). Therefore, in scan AC ₁ , the number Z of blocks having a run length of 1
[1] is 3479, and the number of blocks Z whose run length is 0

[0] is 1921.

FIG. 17 shows the number of blocks whose run lengths up to scan AC ₂ are ₂ , 1, 0. Scan A
In C ₂ , as shown in FIG. 15, the number of blocks of category “0” C

[0] is 3619, and the number of blocks C [1] of category “1” or higher is 1781. Therefore, the number of blocks Z whose run length is 0

Although it is clear that [0] is 1781, the number of blocks Z with a run length of 2
[2] and the number of blocks Z [1] having a run length of 1 are not clear. Therefore, in this embodiment, the number of blocks Z
The ratio of [2] and the number of blocks Z [1] is calculated by scanning AC ₁
Number of blocks with run length 1 in Z [1] and run length 0
The number of blocks Z '

The number of blocks Z [2] and the number of blocks Z [1] are predicted on the assumption that they are equal to the ratio with [0]. That is, in the present embodiment, it is assumed that the number of blocks whose category is “0” in both scan AC ₁ and AC ₂ is related to the number of blocks of category “0” in scan AC ₁ .

The number of blocks Z thus obtained
[2] corresponds to the scan AC ₂ when the categories of the scan AC ₁ and AC ₂ are both “0”. The number of blocks Z [1], when the scan AC ₁ category is other than "0", corresponds to the scan AC ₂ category is "0".

FIG. 18 shows the number of blocks whose run lengths up to the scan AC ₃ are 3, 2, 1, 0. In the scan AC ₃ , as shown in FIG. 15, the number of blocks C in the category “0” is C

[0] is 4366, and the category is "1".
The above block number C [1] is 1034. Therefore, the number of blocks Z

[0] is 1034. The block number Z [1] is the block number Z ′ in scan AC _2.

[0]
It is assumed that Z [1] = 4366 × 1781/5400 = 1440, assuming that it depends on the ratio of On the other hand, assuming that the ratio of the block numbers Z [2] and [3] is equal to the ratio of the block numbers Z [1] and [2] in the scan AC ₂ , Z [2] = 4366 × 1287/5400 = 1041 Z [3] = 4366 × 2332/5400 = 18885.

This block number Z [3] is the scan AC
It corresponds to scan AC ₃ when the categories of ₁ , AC ₂ , and AC ₃ are all “0”. Further, the number of blocks Z [2] is the scan AC whose category is “0” when the category of the scan AC ₁ is other than “0” and the category of the scan AC ₂ is “0”.
Corresponds to ₃ . The number of blocks Z [1] is scan A
It corresponds to the scan AC ₃ whose category is “0” when the category of C ₂ is other than “0”.

FIG. 19 shows a block in which the run length up to scan AC ₄ is _{4, 3} , 2, 1, 0. The above-described processing is performed even after the scan AC ₄ , and the number of blocks Z [k] having a run length of k is obtained.

When the run lengths up to the scan AC ₆₃ are obtained in this manner, next, a run length / category table as shown in FIGS. 20 to 23 is created for each category.

FIG. 20 is a table of run lengths / categories of scan AC ₁ , and this table shows category distributions with run length 0 and categories other than category “0”. Scan A
In C ₁ , the number of blocks other than the category “0” is shown in FIG.
As shown in Fig. 6, 1921 (= Z

[0]) Yes. The category distribution is, as shown in FIG. 15, a category “1”,
"2" ... "5" in the order of 866, 519, 300,
212 and 24, and the number of blocks of category “6” and above is zero. By posting this number as it is,
The table of FIG. 20 is created.

FIG. 21 is a table of scan AC ₂ run lengths / categories. This table shows category distributions for run lengths 0 and 1 other than category "0". In Scan AC ₂ , the number of blocks other than category “0” is
As shown in FIG. 17, a total of 1781 (= Z

[0]) Yes. As shown in FIG. 15, the category distributions are 850, 487, in the order of categories “1”, “2” ...
315, 119 and 10, and the number of blocks in the category “6” and above is 0. The block of 850 in the category “1” is the ratio of Z [1] and Z [2] shown in FIG. 17, that is, the number of blocks Z ′ having a run length of 0 in scan AC _1.

Assuming that the ratio is [0] and the number of blocks Z [1] having a run length of 1, the number of blocks having a run length of 0 is 302,
The number of blocks having a run length of 1 is set to 548. Similarly, regarding the block of 487 in the category “2”,
The number of blocks with a run length of 0 is 173, and the number of blocks with a run length of 1 is 314. In this way, FIG.
A table of 1 is created.

FIG. 22 is a table of run lengths / categories of scan AC ₃ , and this table shows category distributions of run lengths 0, 1 and 2 other than the category “0”. In the scan AC ₃ , the number of blocks other than the category “0” is 1034 (= Z in total as shown in FIG.

[0])
is there. As shown in FIG. 15, the category distributions are 675, 220, in the order of categories “1”, “2” ... “4”.
121 and 18, and the number of blocks of category “5” or higher is 0. The number of blocks in each category is distributed to the number of blocks with run lengths 0, 1 and 2 according to the ratio of Z [1], Z [2] and Z [3] shown in FIG.

FIG. 23 is a table of scan AC ₄ run lengths / categories. This table is also shown in FIG.
It is created with reference to FIG. In this way, a table of run lengths and categories up to the scan AC ₆₃ is created.

Next, the amount of data (the number of bits) in each scan is calculated with reference to the run length / category tables as shown in FIGS. For this calculation,
The code length table shown in is used. Each value in this table is J
The code length of each code word of the Huffman table recommended by PEG is shown. For example, the code length of a codeword having a run length of 1 and a category of 2 is 5.

The scan AC ₁ is shown in FIGS.
4, the number of bits of the codeword of the run length / category in all blocks is 866 x 2 + 519 x 2 + 300 x 3 + 212 x 4 + 24 x 5 = 4638. On the other hand, with regard to additional bits, the number of additional bits is 866 x 1 + 519 x 2 + 300 x 3 + 212 x 4 + 24 x 5 = 3772, as the number of the category corresponds directly to the number of bits. Become. Therefore, the expected data amount of scan AC ₁ is 4638 + 3772 = 8410 (bits).

For scan AC ₂ to scan AC ₆₃ , the data amount is obtained by the same calculation as for scan AC ₁ .

The data amount in each scan thus obtained must be less than or equal to the set code amount shown in FIG. For example, for scan AC ₁ 3300
Must be kept below 8 bits. In the above example, since the quantization coefficient is set to "16" for the description of the embodiment, the data amount is 8410 bits, which is considerably smaller than the set code amount. Therefore, in practice, a value smaller than "16" is suitable for the quantization coefficient. That is, when the quantization coefficient is “1”, the expected data amount is 3300
If it is larger than 8 bits, the quantized coefficient is increased by 1 and the expected data amount is recalculated. While performing such an operation, the quantized coefficient that is equal to or less than the set code amount is obtained.

The generation of the above quantization table is shown in FIG.
This will be described with reference to the flowchart of. Note that this flowchart shows the generation of the quantized coefficients for the scans AC _{2 to} AC ₆₃ , and it is assumed that the DC component and the quantized coefficient of the scan AC ₁ have already been obtained.

In step 101, the parameter i is set to 2. In step 102, the quantization coefficient q is set to 1 as an initial value.

In step 103, the category distribution of the scan AC _i as shown in FIG. 15 is generated using FIG. For example, the parameter i is 3 and the quantization coefficient q
Is 16, this category distribution is 4366, 675, 220, 121, 18, 0, 0, 0, 0, 0, 0 in the example shown in FIG.

Before the step 103 is executed, the number of blocks in the category "0" of the scan AC _i-1 is obtained, as indicated by the symbol P11. For example, the parameter i is 3 and the quantization coefficient q of the scan AC ₂ is 16
, The number of blocks of category “0” of scan AC ₂ is 3619 in the example of FIG. In addition, the symbol P
As shown in 12, the category of scan AC _i-1 is “0”.
The distribution of the number of blocks of is required. For example, the parameter i is 3 and the quantization coefficient q of the scan AC ₂ is 16
17 in the scan AC ₂ , the number of blocks Z [2] is 2332, and the number of blocks Z is Z32.
[1] is 1287, the number of blocks is Z

[0] is 1781.

[0069] At step 104, the category distribution of the scan AC _i determined at step 103, the scan AC _i-1 category number of blocks "0" (code P1
1) and the distribution of the number of blocks of category “0” of scan AC _i−1 (reference P12), based on the scan AC _i
The distribution of the number of blocks of category “0” in is predicted. For example, the parameter i is 3 and the quantization coefficient q is 1.
6, when the scan AC ₃ in the example of FIG.
Based on the data of FIG. 15 and FIG. 17, the number of blocks Z
[3] is 1885, the number of blocks Z [2] is 1041, and the number of blocks Z [1] is 1440.

At step 105, the predicted data amount of the compressed data in the scan AC _i is obtained. For example, when the parameter i is 3 and the quantization coefficient q is 16,
The predicted data amount D _i in the scan AC ₃ is 223 x 2 + 73 x 2 + 40 x 3 + 6 x 4 + 161 x 4 from the run length / category table of FIG. 22 and the code length table of FIG. + 52 x
5 + 29 x 7 + 4 x 9+ 291 x 5 + 95 x 8 + 52 x 10 + 8
x 12+ (223 + 161 + 291) + (73 + 52 + 95) x 2+ (40
+ 29 + 52) x 3 + (6 + 4 + 8) x 4 = 6260 (bits).

At step 106, it is judged if the predicted data amount D _i calculated at step 105 is the set code amount S _i as shown in FIG. In the example of the explanation of step 105 described above, since the quantization coefficient q is 16, the predicted data amount D _i is 6260 bits, which is considerably smaller than the set code amount S ₃ = 24042 (code SB3) in FIG. However, in reality, since the quantized coefficient q starts from 1, the predicted data amount D is initially set.
_i is larger than the set code amount S _i . Therefore, after it is confirmed in step 107 that the quantized coefficient q has not reached 255, the quantized coefficient q is incremented by 1 in step 108, and the step 1 is repeated.
Return to 03.

Then, steps 103 to 105 are executed again, and when it is determined in step 106 that the predicted data amount D _i is less than or equal to the set code amount S _i , the quantization coefficient q is fixed. That is, the routine proceeds from step 106 to step 110, and it is judged whether or not the parameter i has reached 63. When the parameter i has not reached 63, in step 111, the set code amount S _{i + 1} for the next scan is added by the difference between the set code amount S _i and the predicted data amount D _i . That is, the set code amount S _{i + 1 for} the next scan is allocated to the amount not used in the previous scan set code amount S _i .

Next, at step 112, the parameter i
Is incremented by 1 and the process returns to step 102,
Steps 102 to 108 are executed for the next scan to obtain the quantized coefficient.

In step 110, the parameter i is 6
When it is determined that the number has reached 3, the process proceeds to step 113, a quantized coefficient is generated based on the quantized coefficient q, and this program ends.

The first method of predicting the data amount of each scan AC _i described with reference to FIGS. 16 to 23 is a relatively simple example, and is not necessarily a highly accurate prediction. . Therefore, next, a more highly accurate second prediction method will be described with reference to FIGS. 26 to 28.

FIG. 26 shows the distribution of "00" data, that is, the number of blocks in which both scan AC _i-1 and scan AC _i are in category "0", and corresponds to the category distribution table of FIG. For example, scan AC ₁ , AC ₂
, The number of blocks in the category “0” is 120 (reference J ₁₂ ), and the scans AC ₂ and AC ₃ are both in the category “0”.
Block number is 159 (reference numeral J ₂₃ ), scan AC ₃ ,
The number of blocks of category "0" for both AC ₄ is 186 (reference J ₃₄ ), and the number of blocks of category "0" for both scans AC ₆₂ and AC ₆₃ is 327 (reference J ₆₂₆₃ ).

If the number of blocks in which the scans AC _i-1 and AC _i are in the category "0" respectively changes, scan A
The number of blocks in which both C _i-1 and AC _i are in the category “0” is the change rate of the number of blocks in scan AC _i-1 and scan A.
It is assumed to be proportional to the product of C _{i and} the rate of change. For example, if the number of blocks of the scans AC _i−1 and AC _i is doubled, the number of blocks of which the scans AC _i−1 and AC _i are both “0” is quadrupled. Under this assumption, based on the number of blocks of category "0" Scan AC _i-1 and AC _i, Scan AC _i-1 after quantization,
The quantized scans AC _i both predict the number of blocks of category “0”.

The distribution of "00" data is normalized by setting the reference value of the number of blocks to 3500 and stored in the memory. For example, the number of blocks whose scans AC ₁ and AC ₂ are both in the category “0” is
It is converted to 3823 as shown in FIG. FIG. 28 shows
The table of the distribution of the normalized “00” data is shown. For example, the “00” data of the scan AC ₁ and AC ₂ is 382.
3 (reference K ₁₂ ), scan AC ₂ and AC ₃ of “00” data is 3577 (reference K ₂₃ ), scan AC ₃ and AC ₄
"00" data is 3280 (symbol K _34), scan A
The data of “00” of C ₆₂ and AC ₆₃ is 2262 (code KJ
₆₂₆₃ ).

In FIG. 29, the number of blocks of the scan AC _i-1 of category "0" is plotted on the horizontal axis, and scan A
6 is a graph in which the vertical axis represents the number of blocks (that is, “00” data) in which both C _i−1 and AC _i are of category “0”. In this graph, the maximum value of the abscissa is the total number of blocks (5400), and S _low is the reference value 350 described above.
0, S _mid is 4800, S _high is 5150, for example.
Is. The maximum value O _max of the ordinate is variable, and the maximum value O _{max
When max} is the reference value 3500, O _high is 340, for example.
3, O _mid is 3144, O _low is 2463, for example.
Is. The break points f, g, and h are obtained by trial and error, for example, so that the quality of the reproduced image is close to that of the original image, but the break point b is scanned AC _i−1 , as described below.
It changes according to the number of blocks of the category "0" of AC _i .

The method of obtaining the bending point b will be described by taking the case of FIG. 15 as an example. For example, the number of blocks of category “0” of scan AC ₃ is 4366, and the number of blocks of scan AC _{4 is}
The number of blocks in the category “0” is 4108.

First, referring to the table of FIG. 28, scan A
As C ₃ and AC ₄ “00” data, K ₃₄ = 3280
Is obtained. Next, in FIG. 29, when the ordinate corresponding to S _low (3500) is read in the state where the maximum value O _max is set to the reference value 3500, the point c is obtained. Origin a and point c
A straight line L1 connecting the two and a straight line L extending laterally from O _low
When the intersection b with 2 is obtained, a polygonal line abfgh is obtained.

Next, the maximum value O _max of the ordinate is scanned for scan A.
If the number of blocks in C ₄ is 4,108, O _high is 39.
94, O _mid is 3690, and O _low is 2891. In this state, when the ordinate of the point P3 on the polygonal line bf corresponding to the block number 4366 of the scan AC ₃ in the abscissa is read, 3530 is obtained as “00” data of the scan AC ₃ and AC ₄ . This value is Z in FIG.
This corresponds to the sum of [2], Z [3], and Z [4] (= 3321), which is larger than the example of FIG. That is, according to the second prediction method, the number of “00” data of scans AC ₃ and AC ₄ is larger than that in the case of the first prediction method.

The "00" data obtained in this way is converted into Z [2], Z [3] and Z by the method described below.
It is distributed to [4].

For this distribution, Z relating to "00" data of scans AC ₂ and AC ₃ as shown in FIG.
The values of [1], Z [2], Z [3] are required, and
In order to obtain these Z [1], Z [2], and Z [3], the scan AC ₁ and AC ₂ shown in FIG.
The values of Z [1] and Z [2] for "0" data are required.

Scan AC ₁ , AC ₂ Z [1], Z
[2] is obtained by creating a polygonal line abfgh in the same manner as in FIG. 29 and reading the points on this polygonal line.
Z [2] is calculated directly from the polygonal line, Z [1] is C

It is obtained as the difference between [0] and Z [2]. Scan ac
₂ and AC ₃ , the sum of Z [2] and Z [3] is obtained by creating a polygonal line abfgh and reading the point on this polygonal line. Z [1] is the same as Z [2] and Z [2]. The sum of [3] is C

It is obtained by subtracting from [0]. Z [2]
And the values of Z [3] are Z of scan AC ₁ , AC ₂ .
It is distributed according to the ratio of [1] and Z [2]. That is,
Z [1], Z [2], for scan AC ₂ , AC ₃
Each value of Z [3] is obtained.

Similarly, Z of scan AC ₃ and scan AC ₄
The values of [2], Z [3], and Z [4] are for the scan A
C ₂ and AC ₃ are distributed according to the ratio of Z [1], Z [2], Z [3]. As a result, a run length / category table as shown in FIG. 23 is obtained, and the predicted data amount of compressed data in scan AC ₄ is calculated. Similar calculations are performed for other scans to obtain the predicted data amount.

Next, the third prediction method of the data amount of the scan AC _i will be described. In scan AC _i-1 , the number of blocks with run length k is Z [k], scan AC
_Let Z be the total number of blocks that are in category "0" in _i-1 .
_{Let ALL be} CT _j be the number of blocks that are in category “0” in scan AC _i−1 and in category “j” in scan AC _i . Further, the correction coefficient is ZW.

The number of blocks B [k, j] having a run length k and a category j in scan AC _i is B [k, j] = CT _j × Z [k] / Z _ALL + (1-CT _j / Z _ALL ) × Z [k] × ZW / 1000 (1) The number of blocks Z [k + 1] whose run length is (k + 1) in the scan AC _i (that is, the number of blocks whose run length is k and whose category is 0) is: Z [k + 1] = CT ₀ × Z [k] / Z _ALL + (1-CT ₀ / Z _ALL ) × Z [k] × ZW / 1000 (2)

The correction coefficient ZW can take a value from 0 to 1000. For example, when ZW = 0, Z [k + 1] = CT _j × Z [k] / Z _ALL , which has the same result as the second prediction method. ZW
= 1000, Z [k + 1] = Z [k], which means that all the blocks of the scan AC _{i−1 having} the run length k are in the category “0” in the scan AC _i . In the example described below, ZW = 5
It is set to 0.

Here, in scan AC ₃ , Z

[0] = 1034, Z [1] = 1440, Z [2] = 1
041, Z [3] = 1888, and scan AC
It is assumed that a table as shown in FIG. 30 is obtained as the category distribution in ₄ . That is, in scan 4366 block in which scan AC ₃ is category “0”, scan A
In C ₄ , the number of blocks CT ₀ of the category “0” is 332
1, the block number CT ₁ of the category “1” is 655, the block number CT ₂ of the category “2” is 276, the block number CT ₃ of the category “3” is 101, the block number CT ₄ of the category “4” is 13, It is assumed that the number of blocks in the category “5” and above is 0.

As described below, the number of blocks is obtained from the data having a run length of 1, and the run length / category distribution table as shown in FIG. 31 is obtained.

In scan AC ₄ , the number of blocks having a run length of 1 and a category of “1” can be calculated from the equation (1) as follows: B [1,1] = CT ₁ × Z [1] / Z _ALL + (1-CT
₁ / Z _ALL ) × Z [1] × ZW / 1000 = 655 x 1440/436
6 + (1-655/4366) x 1440 x 50/1000 = 216 + 61 =
It becomes 277. Similarly, the number of blocks having a run length of 1 and a category “2” can be calculated from the equation (1) as follows: B [1,2] = CT ₂ × Z [1] / Z _ALL + (1-CT
₂ / Z _ALL ) × Z [1] × ZW / 1000 = 91 + 67 = 158.

Regarding the number of blocks having a run length of 1 and a category of “3”, from the equation (1), B [1,3] = CT ₃ × Z [1] / Z _ALL + (1-CT
₃ / Z _ALL ) × Z [1] × ZW / 1000 = 33 + 71 = 104, but as shown in FIG. 30, CT ₃ = C [3] =
101. Therefore, it is forcibly set to 101 instead of 104.

Similarly, B [1,4] is forcibly set to 13 for the number of blocks having a run length of 1 and a category of "4".

In scan AC ₄ , the number of blocks Z [2] having a run length of 2 (that is, the number of blocks having a run length of 1 and a category of 0) is Z [2] = Z [1]-( 277 + 158 + 101 + 13) = 89
Becomes 1.

Regarding the number of blocks having a run length of 2 and a category of "1", B [2,1] = CT ₁ × Z [2] / Z _ALL + (1-CT
₁ / Z _ALL ) × Z [2] × ZW / 1000 = 655 x 1041/436
6 + (1-655/4366) x 1041 x 50/1000 = 156 + 44 =
It will be 200. For the number of blocks with run length 2 and category “2”, B [2,2] = 66 + 49 = 115.

The number of blocks B [2,3] whose run length is 2 and category "3" is 0 according to the result of B [1,3], and whose run length is 2 and category The number of blocks B [2,4] that is "4" is B [1,4]
It is 0 according to the result of.

In scan AC ₄ , the number of blocks Z [3] having a run length of 3 (that is, the number of blocks having a run length of 2 and a category of 0) is Z [3] = Z [2]-( 200 + 115 + 0 + 0) = 726.

For the number of blocks having a run length of 3 and a category "1", B [3,1] = CT ₁ × Z [3] / Z _ALL + (1-CT
₁ / Z _ALL ) × Z [3] × ZW / 1000 = 655 x 1885/436
6 + (1-655/4366) x 1885 x 50/1000 = 283 + 80 =
It becomes 363. However, the number of blocks CT in the category "1"
_Since the total number of 1s is 655, B [3,1] = 655−277−200 = 178.

For the number of blocks having a run length of 3 and a category "2", B [3,2] = CT ₂ × Z [3] / Z _ALL + (1-CT
₂ / Z _ALL ) × Z [3] × ZW / 1000 = 119 + 88 = 207. However, the number of blocks CT in the category “2”
_Since the total number of ₂ is 276, B [3,2] = 276−158−115 = 3.

The number of blocks having a run length of 3 and category "3" is 0 according to the results of B [1,3] and B [2,3], and the run length is 3 and The number of blocks in the category “4” is B [1,4], B [2,
It is 0 according to the result of [4].

The number of blocks Z [4] having a run length of ₄ in scan AC ₄ (that is, the number of blocks having a run length of 3 and a category of 0) is Z [4] = Z [3]-( 178 + 3 + 0 + 0) = 1704.

With respect to the number of blocks having a run length of 0, the number of blocks of categories "1" to "4" is 1 according to FIG.
55, 65, 24, and 3. Further, the number of blocks Z [1] having a run length of 1 (that is, the number of blocks having a run length of 0 and a category of 0) is Z [1] = 787 from FIG.

FIG. 32 is a table of run length / category distributions thus obtained, and corresponds to FIG. 23 in the first prediction method. Further, as shown in FIG. 31, the number of blocks having a run length of 4 (run length is 3,
And the number of blocks whose category is 0) is 1704, which is obtained by the first prediction method Z [4] =
It is larger than 1434. That is, it is understood that the run length is extended according to the third prediction method.

Next, in the category distribution table as shown in FIG. 30, the category of scan AC _i-1 is "0".
In addition, a method of obtaining the number of blocks whose scan AC _i category is “1” or more using the polygonal line abfgh in FIG. 29 will be described.

In this case, in FIG. 29, the horizontal axis represents the number of blocks of category “0” in scan AC _i−1 , and the vertical axis represents the sum of the number of blocks of category “0” and category “1” in scan AC _i . Show. First, let O _max on the vertical axis be C

Sum of [0] and C [1] (= 4108 + 810 = 491
8), and O _high , O _mid , and O _low are calculated accordingly. Next, when a point (for example, P3) on the polygonal line corresponding to the abscissa 4366 is read and the ordinate corresponding to this point is read, this value is a block of category “0” and category “1” in the scan AC ₄ . It is the sum of numbers. Therefore,
By subtracting the value of the already obtained category "0" from this sum, the number of blocks with the category of scan AC _i-1 being "0" and the category of scan AC _i being "1" can be obtained. Thereafter, in the same manner, the number of blocks in which the category of scan AC _i-1 is “0” and the category of scan AC _i is “2” or more is obtained.

[0107]

As described above, according to the present invention, it is possible to obtain the effect that image compression can be achieved according to the image quality of each image.

[Brief description of drawings]

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

FIG. 2 is image data P (Y) xy, DCT conversion coefficient S (Y) uv.
, Quantized DCT coefficient R (Y) uv, quantization table Q (Y) uv
It is a figure which shows the example of.

FIG. 3 is a diagram showing categories of difference values of DC coefficients.

FIG. 4 is a diagram showing a coding table of DC coefficients.

FIG. 5 is a flowchart showing a processing routine for encoding quantized AC coefficients.

FIG. 6 is a diagram showing zigzag scanning performed when Huffman encoding AC coefficients.

FIG. 7 is a diagram showing categories of AC coefficients.

FIG. 8 is a diagram showing a Huffman table recommended by JPEG.

FIG. 9 is a diagram showing an example of encoded data by Huffman encoding.

FIG. 10 is a diagram showing constituent elements of encoded data.

FIG. 11 is a diagram showing an example of category distribution at a predetermined spatial frequency when a quantization table in which all the quantization coefficients are “1” is used.

FIG. 12 is a diagram showing an example of the distribution of the number of bits of encoded data in each scan when a quantization table in which all the quantization coefficients are “1” is used.

FIG. 13 is a diagram showing an example of a distribution of code amounts for each spatial frequency (that is, for each scan) set in the spatial frequency data amount setting unit.

FIG. 14 is a diagram showing the category distribution shown in FIG. 11 simultaneously for each scan.

FIG. 15 is a diagram showing a category distribution table when a quantization table in which all the quantization coefficients are “16” is used.

FIG. 16 is a diagram showing the number of blocks of category “0” and the number of blocks of category “1” or more in scan AC ₁ .

FIG. 17 is a diagram showing the number of blocks whose run lengths up to scan AC ₂ are ₂ , 1, 0;

FIG. 18 shows run lengths up to scan AC ₃ of 3, 2, 1,
It is a figure which shows the number of blocks which is 0.

FIG. 19 shows run lengths up to scan AC ₄ of 4, 3, 2,
It is a figure which shows the number of blocks which are 1 and 0.

FIG. 20 is a diagram showing a table of run lengths / categories of scan AC ₁ .

FIG. 21 is a diagram showing a table of run lengths / categories of scan AC ₂ .

FIG. 22 is a diagram showing a table of run lengths / categories of scan AC ₃ ;

FIG. 23 is a diagram showing a table of run lengths / categories of scan AC ₄ ;

FIG. 24 is a diagram showing the code length of each codeword of the Huffman table recommended by JPEG.

FIG. 25 is a flowchart of a program for generating a quantization table.

FIG. 26 is a diagram showing a distribution of “00” data.

[FIG. 27] “00” of scan AC ₁ and scan AC ₂
It is a figure which shows the normalization of data.

FIG. 28 is a diagram showing a table of distribution of normalized “00” data.

FIG. 29 is a diagram showing the relationship between the "00" data of the number of blocks of scan AC _i-1 of the category "0" and the scan AC _i-1 and Scan AC _i.

FIG. 30 is a diagram showing an example of category distribution in scan AC ₄ .

FIG. 31 is a diagram showing an example of the distribution of run lengths / categories obtained by the third prediction method.

FIG. 32 shows a table of distribution of run lengths / categories of FIG.
It is the figure expressed in the format corresponding to 3.

[Explanation of symbols]

 Mic memory card

Claims (9)

[Claims] 1. An orthogonal transform means for performing an orthogonal transform on original image data to obtain an orthogonal transform coefficient for each spatial frequency, and an orthogonal transform coefficient which is quantized orthogonally by quantizing the orthogonal transform coefficient by a quantization table composed of predetermined quantization coefficients. Quantization means for obtaining a transform coefficient, and the quantized orthogonal transform coefficient having a predetermined 1 with respect to a spatial frequency.
After rearranging into the dimensional array data, encoding means for obtaining encoded data by performing encoding based on the quantized orthogonal transform coefficient, and a target value of the amount of encoded data for each spatial frequency are set. Means, means for predicting the data amount of the encoded data of the second spatial frequency based on the statistic of the encoded data of the first spatial frequency, and the predicted data amount of the second spatial frequency is An image compression apparatus, comprising: a quantization coefficient calculation unit that determines a quantization coefficient corresponding to the spatial frequency so as to be equal to or less than a target value. 2. The quantized orthogonal transform coefficient corresponding to the second spatial frequency is adjacent to the quantized orthogonal transform coefficient corresponding to the first spatial frequency in the one-dimensional array data. The image compression apparatus according to claim 1. 3. The encoding means classifies the quantized orthogonal transform coefficient into a category having a predetermined data amount based on the value of the quantized orthogonal transform coefficient, and The run length is obtained based on a continuous number, and the predicting unit predicts the data amount of the encoded data of the second spatial frequency based on the category and the run length. The image compression device described. 4. The orthogonal transform means divides the original image data into a plurality of blocks to obtain orthogonal transform coefficients, and the predicting means predicts a data amount of the second spatial frequency by using a first 4. The image compression apparatus according to claim 3, wherein the number of blocks of category "0" corresponding to a quantized orthogonal transform coefficient of 0 at the spatial frequency of is used. 5. The predicting means predicts the number of blocks that are both category “0” in the first and second spatial frequencies based on the number of blocks in category “0” in the first spatial frequency. The image compression apparatus according to claim 4, wherein 6. The predicting means is a category “0” in the first spatial frequency and a category “0” in the second spatial frequency based on the number of blocks in the category “0” in the first spatial frequency. The image compression apparatus according to claim 4, which predicts the number of blocks of category "j" (j is a positive integer). 7. The orthogonal transforming unit divides the original image data into a plurality of blocks to obtain an orthogonal transform coefficient, and the predicting unit uses the quantized coefficient “1” for the first and second transforms. Based on a category distribution table indicating the number of blocks in which the second spatial frequencies are both category "0", both the first and second spatial frequencies when using a predetermined quantization coefficient are category "0". The image compression apparatus according to claim 4, which predicts the number of blocks that are 8. The orthogonal transforming unit divides the original image data into a plurality of blocks to obtain an orthogonal transform coefficient, and the predicting unit uses the quantized coefficient “1” for the first and second transforms. Based on a category distribution table showing the number of blocks in which the second spatial frequencies are both category "0", the first spatial frequency is category "0" when a predetermined quantization coefficient is used, and The image compression apparatus according to claim 4, wherein the number of blocks in which the second spatial frequency is category “j” (j is a positive integer) is predicted. 9. The orthogonal transforming unit divides the original image data into a plurality of blocks to obtain orthogonal transform coefficients, and the predicting unit, at each spatial frequency when a quantized coefficient “1” is used, The number of blocks classified into each category at each spatial frequency when a predetermined quantization coefficient is used is determined based on the number of blocks classified into each category. Image compression device.