JPH08204970A - Image data compressing and expanding device - Google Patents

Abstract

PURPOSE: To attain the target compressibility through a single compression processing by quantizing the block image data which underwent the discrete cosine transformation(DCT) based on the quantization tables that satisfy the necessary conditions and applying the entropy coding to the transformation coefficients to compress them. CONSTITUTION: The multilevel image data are divided at a block division part 1, undergo the DCT through a DCT part 2, and are quantized at a quantization part 3 based on plural quantization tables 4. The DCT coefficients are compressed at an entropy coding part 7 and outputted. At the same time, the quantity of compressed data measured at a compressed data quantity measurement part 9 is compared with the target compressed data quantity of a target compressed data quantity storage device 10 to select one of those corresponding tables 4 at a comparison processing part 11. Thus the target compressibility is attained by a single compression processing not by two operations carried out for every block.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a digital image processing apparatus. For example, the present invention relates to an apparatus for compressing image data by encoding multi-valued image data read from an image reading apparatus such as a scanner, and an apparatus for decompressing the compressed image data.

[0002]

2. Description of the Related Art Printing devices capable of printing high-resolution full-color data and display devices capable of displaying have become popular in recent years. However, since high-resolution full-color data has a large amount of information, storage devices A large amount of storage area is required to save in.

For example, with a paper size of A3, the pixel color is composed of four colors of yellow, magenta, cyan and black, and the resolution in which each pixel color is represented by 8 bits is 400 dpi (d).
The area for storing 128 Mbytes is required for full-color image data of
If this is to be stored in the memory, a very large capacity memory is required, resulting in an increase in manufacturing cost.

Therefore, there has been conventionally devised a method of reducing the cost by encoding and compressing the image to reduce the information amount of the image and the area for storing the image.

Particularly, as an encoding method for compressing an image,
As an international standardization method, JPEG (Joint Photo)
There is a proposed baseline system at ograph Experts Group). In this method, image data is subjected to discrete cosine transform (hereinafter referred to as DCT transform).
In this method, the quantized coefficient after the Huffman coding is performed to reduce the data amount. This will be described in detail.

FIG. 10 is a block diagram schematically showing a compression method proposed by the conventional JPEG. First, the image data is read in units of 8 × 8 pixels. This is pixel
(X, y) (where x = 0 to 7, y = 0 to 7).
An example of this image data is shown in FIG. Next, D in FIG.
In the CT calculation unit 2, DCT calculation represented by the following equation (1) is performed, and DCT (i,
j) is obtained (however, i = 0 to 7, j = 0 to 7).

[0007]

[Equation 1] The result of this DCT calculation is shown in FIG. Where DC
Due to the nature of the T calculation, the higher the frequency, the higher the frequency becomes. Where origin DC
T (0,0) is called a DC component, and the other components are called AC components.

Due to the nature of human vision, low-frequency components are more important because they are easily perceived, and high-frequency components are less perceptible. Therefore, set the high frequency component to 0,
A method of compressing the image by leaving the low frequency component is conceivable. In JPEG, quantization is performed so that the high frequency component of DCT (i, j) (that is, the component with larger i + j) is set to 0 as much as possible. Q is the quantized result
Let V (i, j) be the quantization table Quantum
If (i, j), then QV (i, j) = Round [DCT (i, j) / Qu
Quantum (i, j)]. However, Round [] is a rounding function. FIG. 13 shows an example of the quantization table. FIG.
As can be seen from the above, the higher the frequency (that is, the larger i + j is), the larger the table value is. That is, the QV obtained by dividing the DCT component by the value in the quantization table.
Creates a table so that the higher the frequency component, the closer to zero. FIG. 14 shows the value of the DCT coefficient after being quantized. It can be seen from FIG. 14 that the high frequency component is almost zero. For this quantized data QV, Hu
By performing entropy compression processing such as ffman encoding, image compression of 1/10 to 1/100 can be performed.

In order to restore the compressed image, the quantized data is dequantized to derive the DCT coefficient.
However, since the DCT coefficient is rounded to a value divided by the value in the quantization table during compression, the DCT coefficient is not completely restored. Although this causes deterioration of image quality, the larger the value of the quantization table is, the larger the rounding error is, and thus the deterioration of image quality becomes larger. In inverse proportion to this, the larger the value of the quantization table, the larger the compression ratio. The compressed data is restored to the original image by performing the inverse DCT operation shown in the following equation (2) on the restored DCT coefficient.

[0010]

[Equation 2] As can be seen from the above example, the compression rate is determined by the property of the image data and the value of the quantization table. Considering the nature of the image data, when compressing a blue sky without clouds, most of the images have low frequency components, and there is no high frequency component, so the compression rate is high. On the contrary, in the case of an image in which the color changes bit by bit, there are many high frequency components, and the compression rate is relatively low.

Further, as described above, if the value of the quantization table is increased, the compression rate becomes relatively high as much data is rounded and discarded. Instead, the image quality is greatly degraded. Tables 1 and 2 show the relationship between this compression and image quality.

[0012]

[Table 1]

[Table 2] Here, consider a case where a target compression rate is realized in order to store color image data in a storage device having a certain storage capacity using the JPEG compression method. The problem at this time is the relationship between Table 1 and Table 2. One thing to note is that the compression rate varies greatly depending on the input image data. The second is that if the value of the quantization table is increased in order to increase the compression rate, the image quality will deteriorate. In order to achieve a certain target compression rate, if you want to select a quantization table that satisfies all image data, you have to select the quantization table that satisfies the image data with the worst compression rate. It will be a sacrifice. On the contrary, if the image quality is prioritized, the compression rate may not be satisfied.

Therefore, a conceivable method is to satisfy the image quality and the compression rate by changing the quantization table according to the input image. For example, an image with mostly low frequency components is selected with a small value in the quantization table, and an image with many high frequency components is selected with a large value in the quantization table. However, the problem in this case is that it is impossible to know what the compression ratio will be until the image data is compressed, so it is difficult to know which quantization table can be used for optimal quantization. That is not the case.

In order to avoid this problem, the image data is first compressed using a certain average quantization table and the compression ratio is measured. If the compression ratio is not satisfactory, a good value is re-established. It is conceivable to select a quantization table that is predicted to obtain In this way, if it takes time to compress twice, the target compression rate can be almost reached.

However, since it takes a lot of time and effort to perform the compression process, the compression time is lost. Therefore, it is desirable that the target compression rate be reached once. A method in which the target compression rate is reached by such one-time processing is disclosed in Japanese Patent Laid-Open No. 6-
22152 publication. An example of the image processing apparatus described in the publication is shown in FIG. With this method,
First, the image data is divided by the block division unit 1 into several blocks of the same size. The unit to be divided may be arbitrarily set depending on the device, but the minimum unit is 8 × 8 JPE.
It must be greater than or equal to the unit block for G compression. Also,
It must be a continuous block in the order of compression. Here, for the sake of convenience, B (1), B (2), and B (1), B (2), and B
(3) ,. ． ． ． ． ． , B (16). The image data is divided into a plurality of blocks by the block division unit 1, and the DCT operation unit 2 performs DCT change on a block-by-block basis to generate DCT coefficients. The DCT coefficient is quantized using one of the plurality of quantization tables 4,
It is encoded by the entropy encoding unit 7. Further, the cumulative measurement unit 8 measures the accumulation of the compressed data for each block, and the quantization table 4 is dynamically switched based on the measurement result. Hereinafter, this quantization table switching control will be described with reference to FIG.

When the target compressed data amount is M when the entire image is compressed, the average one block data amount is M / 16. Total compressed data amount M _T of B (1) to B (N) when compressed to the Nth block
(N) is (M / 16) × N = M _T (N) (see solid line A)
If the compression ratio is controlled so that the final data amount becomes the target data amount. In order to control the amount of compression, it is necessary to switch the quantization table 4 appropriately. Because of this,
The cumulative measurement unit 8 measures the cumulative amount of data after compression, and when the compression of the Nth block is completed, the target cumulative compression amount M
_T (N) is compared with the actual total compressed data amount M (N) of B (1) to B (N), and if M _T (N)>
If it is M (N) (see the broken line B), the compression amount is smaller than the target value, so that a quantization table that reduces the compression rate is selected to control the compression rate of the next N + 1th block. . On the contrary, if M _T (N) <M (N) (see the alternate long and short dash line C), the compression amount is larger than the target value. Therefore, a quantization table that gives a higher compression rate is selected and N
The compression rate of the + 1st block is controlled.

In this way, by comparing the preset degree of accumulation of compressed data with the actual degree of accumulation of compressed data and controlling the compression rate of the next block according to the comparison result, Finally, the amount of data after compression becomes the target amount of compressed data.

In the above-described conventional compression rate control method, the cumulative compression amount of a block is compared with the target compressed data amount, and the quantization table for the next block is switched based on the comparison result. Therefore, the quantization table is determined irrespective of the compression rate of the block currently being compressed, and there is the inconvenience that the compression process suitable for the content of the image being processed cannot be performed.

For example, it is assumed that there is image data such that an image having many high frequency components of 30 blocks and an image having many low frequency components are alternately generated as shown in FIG. FIG.
Reference numeral 8 represents switching of the quantization table used for compressing this image. table0 is a quantization table used when the compression rate is good, and the image quality is relatively good.
Table 2 is a quantization table used when the compression rate is poor, and the image quality is relatively poor. table
Is an intermediate quantization table between table2 and table0.

Ideally, table 2 is used for a block having a large amount of high frequency components and a small compression ratio, and table 0 is used for a block having a large amount of low frequency components and a high compression ratio, so that the amount of data compressed in one block is equal. The better the image quality, the better.

However, as can be seen from FIG. 18, when the compression rate is controlled by the accumulated data, the quantization table is calculated based on the accumulated compressed data amount regardless of the compression rate of the currently compressed data. Since the selection is made, the compression processing suitable for the content of the image being processed cannot be performed, and there is a problem that the image quality is deteriorated.

That is, in the examples of FIGS. 17 and 18, the first few blocks have a large amount of data after compression because the frequency component of the image data is high, and the quantization table table2 having a high compression rate is used. Is selected. However, even if the block having the high frequency component ends, if the cumulative amount of compressed data is larger than the target amount of compressed data, table 2 continues to be selected for a while even if the block has the low frequency component. For this reason, a block of a low frequency component having a small amount of data is compressed at a high compression rate, and the image quality of the image at the time of decompression deteriorates.

[0023]

Therefore, the present invention provides an image compression and decompression device capable of achieving a target compression rate with a single compression process and preventing deterioration of image quality. The purpose is to

[0024]

In order to achieve the above-mentioned object, the image data compression apparatus of the present invention divides a multi-valued image into a plurality of blocks and inputs the divided data into each block. A means for transform-encoding the value image to obtain transform coefficients, a plurality of quantization tables having mutually different quantization data, and a transform coefficient selected from the plurality of quantization tables. Means for quantizing, means for entropy-encoding the quantized transform coefficients to compress, means for measuring the amount of compressed data of one divided block, and means for holding a target amount of compressed data for each block, And a means for comparing the compressed data amount of the one block with a target compressed data amount per one block and selecting the quantization table according to the comparison result. It is characterized in.

The image data decompression device of the present invention further comprises means for inputting the compressed data divided into a plurality of blocks for each block, and means for obtaining the quantized transform coefficient by entropy inverse encoding the input compressed data. A plurality of quantization tables having mutually different quantized data, and means for dequantizing the quantized transform coefficients by using a selected quantization table among the plurality of quantization tables to obtain transform coefficients. , Means for inversely transform-encoding transform coefficients into a multi-valued image, means for holding a target compressed data amount per block, means for measuring the compressed data amount of one divided block, and the one block Comparing the amount of compressed data of and the target amount of compressed data per block,
Means for selecting the quantization table according to the comparison result.

[0026]

The image data divided into a plurality of blocks is transform-coded for each block to generate transform coefficients,
The transform coefficient is quantized using a quantization table selected from a plurality of quantization tables, and the quantized transform coefficient is entropy coded and compressed. At the time of this compression processing, the data amount after compression is measured for each block and compared with a preset target compressed data amount per block. When the amount of data after compression is larger than the target amount of compressed data, a quantization table with a high compression rate is selected, and in the opposite case, a quantization table with a low compression rate is selected. As a result, compression is performed block by block using a quantization table suitable for the nature of the image.

Further, when decompressing the compressed data, a process reverse to the above-described compression process is performed, and the compressed data is decompressed in block units using a quantization table suitable for the nature of the image, and little deterioration occurs. The image is restored.

[0028]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The features of the present invention will be specifically described below based on embodiments with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of the image data compression apparatus of the present invention. The image data compression apparatus shown in FIG. 1 divides a multi-valued image into a plurality of blocks and inputs each block, and a block division unit 1 that performs DCT coding on the input multi-valued image to obtain transform coefficients.
T calculation unit 2, a plurality of quantization tables 4 having different quantized data, and a quantization processing unit that quantizes a transform coefficient using a quantization table selected from the plurality of quantization tables 4. 3, an entropy coding unit 7 for entropy coding and compressing the quantized transform coefficient, a compressed data amount measuring unit 9 for measuring the compressed data amount of one divided block, and a target compressed data amount per one block. The target compressed data amount holding unit 10 to hold and the comparison processing unit 11 that compares the compressed data amount of one block and the target compressed data amount of one block, and selects one of the quantization tables according to the comparison result. It has and. In addition,
Portions corresponding to those of the image data compression apparatus shown in FIG. 15 are designated by the same reference numerals.

The operation of the image data compression apparatus shown in FIG. 1 will be described below. In the present embodiment, first, the multi-valued image data is divided into a plurality of blocks having the same size by the block division unit 1. The unit of division may be arbitrarily set depending on the device, but when JPEG is adopted as the compression method, the minimum unit must be equal to or larger than a unit block of JPEG compression of 8 × 8 pixels. Also, the blocks must be consecutively arranged in the order of compression. Here, for the sake of convenience, the image data is divided into 16 sections of 4 × 4, and in order of compression, B (1), B (2),
B (3) ,. ． ． ． ． ． ． , B (16).

When the target compressed data amount is M when the entire image is compressed, the average one block data amount is M / 16. The average amount of compressed data per block is mt. Then, the compressed data amount when the Nth block B (N) is compressed is set to m (N).

The image data divided into a plurality of blocks by the block dividing unit 1 is discrete cosine transformed by the DCT computing unit 2 according to the equation (1) to generate transform coefficients, and the transform coefficients are stored in one of the quantization tables 4. To quantize.
Further, the quantized transform coefficient is coded by the entropy coding unit 7 using a Huffman code or the like and compressed for each block. Further, the compressed data amount measuring unit 9 measures the compression amount for each block. The target compressed data amount storage device 10 stores the target compressed data amount mt, and the comparison processing unit 11 compares the measured value by the compressed data amount measurement unit 9 with the value of the target compressed data amount storage device 10. Then, based on the comparison result of the comparison processing unit 11,
The quantization table 4 is switched, and the N + 1-th block is compressed using the switched quantization table.

Assuming that the quantization table 4 has five types, for example, the table having the highest compression rate is used in the order of Table.
0, Table1, Table2, Table3, Ta
ble4. The relationship between the compression rate and the quantization table is shown in FIG.
, The quantization table with the higher compression rate is selected for the image block with many high frequency components and the low compression rate, and the image table with the high compression rate with many low frequency components is selected.
Select a quantization table with a low compression rate to prevent image deterioration as much as possible. That is, the quantization table may be selected under the following conditions.

When mt + β <m (N), Table0 is selected. When mt + α <m (N) ≦ mt + β, Table1 is selected. When mt−α ≦ m (N) ≦ mt + α, Table2 is selected mt-β ≦ m (N). <Table 3 is selected when mt-α is selected. Table 4 is selected when m (N) <mt-β is set. However, α is a distance from the block average compression amount (target value) mt to the upper and lower limits of the compressed data amount that Table 2 handles. Where β is the distance from mt to the upper limit of the compressed data amount that Table1 is responsible for and the lower limit of the compressed data amount that Table3 is responsible for. Regarding the values of α and β, the optimum values may be arbitrarily selected according to each device. The comparison processing unit 11 performs the above comparison and selects a table according to the comparison result.

In the prior art, the quantization table is determined by comparing the cumulative amount of compressed image data with the cumulative amount of the target compressed data amount. In this way, compare the amount of compressed data in each block with the target amount of compressed data,
The quantization table to be used next is determined. In other words, in this method, the quantization table used in the (N + 1) th block is determined on the basis of the compression rate in the immediately preceding Nth block. Since it is estimated that the properties of images between adjacent blocks are similar to each other, the quantization table of the N + 1th block is determined based on the compression rate of the image of the Nth block to determine the image quality of the image. Appropriate compression according to the content is performed, and an image with little deterioration can be reproduced when decompressing the compressed data. The first block is
Since there is no previous compressed block, compression is performed using Table2, which is an intermediate quantization table. However, the quantization table of the first block may be arbitrarily determined by the device. In the present embodiment, the target compression rate can be achieved while minimizing the image deterioration by the above method. FIG. 3 is a graph showing a quantization table selected in the present invention for the image data shown in FIG.

For compressed data, a normal JP
In addition to the EG data format, a quantization table is added.
In the case of normal JPEG data, there is one quantization table, but in this embodiment, since a plurality of quantization tables are used, all quantization tables are added. Other than that, there is no apparent difference from normal JPEG data. FIG. 4 shows an example of the data format in the present invention when N quantization tables are used.

5 and 6 are flow charts showing the above-mentioned compression operation. Image data from a storage device (not shown) is divided into a plurality of blocks by the block division unit 1, and N
Read the image data of the block (step 10
1). If the read block is the first block (step 102), Tabl is used as the quantization table.
e2 is selected (step 103). Next, the DCT calculation unit 2 performs DCT calculation in units of 8 × 8 pixels to generate DCT coefficients (step 104). Next, in the quantizer 3, using the previously selected quantization table, 8 × 8
Quantize the DCT coefficient in pixel units (step 1
05). Next, in the entropy coding unit 7, 8 ×
Huffman coding is performed in units of 8 pixels, and the compressed data is stored in a memory (not shown) (step 106).
When the compression process in units of 8 × 8 pixels is completed, it is determined whether or not the process for one block is completed (step 107),
If not, return to step 104, 8 ×
The compression process is repeated in units of 8 pixels. When the processing for one block is completed (step 107), it is judged whether or not the processing for the entire image data is completed (step 108). If not completed, the processing returns to step 101 to read the next block. The above compression process is repeated.

After the second block, the process proceeds from step 102 to step 109, a table is selected according to the compressed data amount measured by the compressed data amount measuring unit 9 (steps 109 to 117), and the compression process is performed similarly. . When the processing of the entire image data is completed, the processing ends.

Next, the expansion device will be described. FIG.
It is a block diagram which shows the Example of the image data expansion apparatus of this invention. The image data decompression device according to the present exemplary embodiment includes a block input unit 21 that inputs compressed data divided into a plurality of blocks for each block, and Huff the input compressed data.
An entropy inverse encoding unit 22 that obtains a quantized transform coefficient by performing entropy inverse encoding using a Man code and the like, and a plurality of quantization tables 2 having mutually different quantized data.
4, a dequantization unit 23 that dequantizes a quantized transform coefficient using a quantization table selected from a plurality of quantization tables 24 to obtain a transform coefficient, and a transform coefficient that is inverse-transform-coded to a large number. Inverse DCT operation unit 25 for restoring the value-value image, compressed data amount holding unit 27 for holding a target compressed data amount per block, and compressed data amount measuring unit 26 for measuring the compressed data amount of one divided block. And a comparison processing unit 28 that compares the compressed data amount of one block with the target compressed data amount per one block and selects the quantization table 24 according to the comparison result. The quantization table 24 corresponds to the quantization table 4 of the image data compression apparatus shown in FIG. 1, and is the table shown in FIG.
It has the characteristics of 0 to Table 4.

The process of decompressing the compressed data is the reverse of the above-mentioned compression process. This image data expansion processing will be described below with reference to the block diagram of FIG. 7. The compressed data is B (1), B (2), B
(3) ,. ． ． ． ． ． ． It is assumed that the block is divided into 16 blocks of B (16).

The compressed data divided into a plurality of blocks is input for each block by the block input unit 21, and the input compressed data is entropy inversely encoded by the entropy inverse encoding unit 22 to obtain a quantized DCT coefficient. Next, the quantized DCT coefficient is inversely quantized by the inverse quantization unit 23 using the quantization table selected from the plurality of quantization tables 24, and the DCT coefficient DCT before being quantized
Get (i, j).

In this inverse quantization, the first block is
As the quantization table, the central Table2 is selected, and after the second block, the size of the compressed data of the immediately preceding block is compared with the target compressed data amount of the one block, and Table0 to Table4 are compared. Select one. The criteria to be selected are as described in the compression example.

In the above-described inverse quantization, the quantized result is set to QV (i, j), and the quantization table is set to Quantum.
If (i, j), then DCT (i, j) = QV (i, j) × Quantum
It is represented by the formula (i, j).

Next, the inverse DCT operation shown in equation (2) is inverse D
The original image data is restored by the CT operation unit 25.

8 and 9 are flow charts showing the above-described expansion operation. The block input unit 21 reads the Nth block of compressed data from a storage device (not shown).
(Step 201). If the read block is the first block (step 202), Table2 is selected as the quantization table (step 203). Next, the inverse entropy encoding unit 22 performs inverse Huffman encoding in units of 8 × 8 pixels to generate quantized DCT coefficients (step 204). Next, the inverse quantization unit 23 uses the previously selected quantization table to perform inverse quantization in units of 8 × 8 pixels to obtain DCT coefficients before quantization (step 205). Then the inverse DCT
The operation unit 25 performs an inverse DCT operation in units of 8 × 8 pixels to obtain decompressed data, and stores the decompressed data in a memory (not shown) (step 206). When the decompression process in units of 8 × 8 pixels is completed, it is determined whether or not the process for one block is completed (step 207). If not completed, the process returns to step 204 and the unit of 8 × 8 pixels is completed. Repeat the compression process with. When the processing for one block is completed (step 207), it is judged whether or not the processing for the entire image data is completed (step 208).
Returning to step 201, the next block is read and the above decompression processing is repeated. When the processing of the entire image data is completed, the processing ends.

[0046]

As described above, in the present invention, the compressed data amount in each block is compared with the target compressed data amount to determine the quantization table to be used next. There is. As a result, in one compression process, appropriate compression is performed for each block according to the content of the image, and an image with little deterioration can be reproduced when decompressing the compressed data.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of an image data compression device of the present invention.

FIG. 2 is an explanatory diagram showing a relationship between a compression rate and a quantization table.

FIG. 3 is a graph showing a quantization table selected in the present invention for the image data shown in FIG.

FIG. 4 is an explanatory diagram showing an example of a data format in the present invention when N quantization tables are used.

FIG. 5 is a first part of a flowchart showing a compression operation.

FIG. 6 is the second part of the flowchart showing the compression operation.

FIG. 7 is a block diagram showing an embodiment of an image data decompression device of the present invention.

FIG. 8 is a first part of a flowchart showing a decompression operation.

FIG. 9 is a second part of the flowchart showing the decompression operation.

FIG. 10 is a block diagram schematically showing a compression method proposed by a conventional JPEG.

FIG. 11 is an explanatory diagram showing an example of image data of 8 × 8 pixels.

FIG. 12 is 8 × 8 obtained from the pixel data shown in FIG.
5 is an explanatory diagram showing a DCT coefficient matrix of FIG.

FIG. 13 is an explanatory diagram showing a quantization table.

14 is an explanatory diagram showing data after being quantized using the quantization table shown in FIG. 6. FIG.

FIG. 15 is a block diagram showing a conventional image processing apparatus that employs a method in which a target compression rate is reached in a single process.

FIG. 16 is a graph showing the relationship between the target cumulative compressed data amount and the actually measured cumulative compressed data amount.

FIG. 17 is a graph showing a difference in frequency component of image data for each block.

18 is a graph showing a quantization table selected in the conventional example for the image data shown in FIG.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Block division part, 2 ... DCT calculation part, 3 ... Quantization part, 4 ... Quantization table, 5 ... DC component coding part, 6 ...
AC component encoding unit, 7 ... Encoding unit, 8 ... Measuring unit, 9 ... Compressed data amount measuring unit, 10 ... Target compressed data amount storage unit, 1
DESCRIPTION OF SYMBOLS 1 ... Comparison processing part, 21 ... Block input part, 22 ... Decoding part, 23 ... Dequantization part, 24 ... Quantization table, 25
Inverse DCT calculation unit, 26 ... Compressed data amount measuring unit, 27 ...
Target compressed data volume storage

Claims (2)

[Claims] 1. A means for dividing a multi-valued image into a plurality of blocks and inputting each block, a means for transform-encoding the input multi-valued image for each block to obtain transform coefficients, and different quantized data from each other. A plurality of quantization tables having; a means for quantizing a transform coefficient using a quantization table selected from the plurality of quantization tables; a means for entropy coding and compressing the quantized transform coefficients; , A means for measuring the amount of compressed data of one divided block, a means for holding a target amount of compressed data for one block, and a comparison of the amount of compressed data for one block with the amount of target compressed data for one block And a means for selecting the quantization table in accordance with the comparison result, the image data compression apparatus. 2. A unit for inputting compressed data divided into a plurality of blocks for each block, a unit for entropy-decoding the input compressed data to obtain a quantized transform coefficient, and different quantized data. A plurality of quantization tables, means for dequantizing the quantized transform coefficients using the quantization table selected from the plurality of quantization tables to obtain transform coefficients, and inverse transform coding the transform coefficients A unit for restoring a multi-valued image, a unit for holding a target compressed data amount per block, a unit for measuring the compressed data amount of a divided one block, a compressed data amount for the one block and the one block And a means for comparing the target compressed data amounts and selecting the quantization table according to the comparison result. Place.