JPH07143485A - Image data coder - Google Patents

Abstract

PURPOSE:To reduce data quantity of generated codes and to decrease the storage capacity of a coding table as required. CONSTITUTION:A difference of DCT coefficients in the unit of a block is coded. When a comparator 12 decides the difference of the DCT coefficients to be smaller than a minimum coefficient, a selector 16 is used to replace the difference of the DCT coefficients with the minimum coefficient. When a comparator 14 decides the difference of the DCT coefficients to be larger than a maximum coefficient, the difference of the DCT coefficients are replaced with the maximum coefficient. Since a range of DCT coefficients to be coded is made narrow, the information quantity allocated to codes is reduced and generated data quantity is decreased.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image data encoding device for frequency component image data converted from two-dimensional image data, and more particularly to a data amount of a code generated from frequency component image data. The present invention relates to an image data encoding device capable of reducing the amount of hardware to be used, such as reducing the storage capacity of a memory that stores an encoding table to be used, if necessary.

[0002]

2. Description of the Related Art In a communication system for still images such as a facsimile and a videotex, standardization of compression and encoding of image data is in progress for the purpose of convenience of mutual communication. For example, CCITT SGVIII (International Telegraph and Telephone Consultative Committee Study Group VIII) Laporta Group CCIC (common conponent for image communi
cation) and ISO JTC1 / SC29 / WG10 joint group, namely JPEG (joint photographic exp)
ert group).

The coding algorithm for still image data, which is being studied by this group, is usually called the JPEG algorithm. In this JPEG algorithm,
Image compression techniques that are becoming more and more familiar in recent years are used. This is for transmitting still images, storing image data in a database, etc. By reducing the data amount of the image data, it is possible not only to shorten the transmission time and access time but also to store it in the database. As a result, the required storage capacity can be reduced.

Regarding the JPEG algorithm, etc.,
Interface magazine Dec. 1991, pp. 160-18
A detailed explanation is given on page 2. The following JPEG
The description of the algorithm is based on this article.

FIG. 6 is a block diagram showing the configuration of a conventional image data encoding / decoding device based on the JPEG algorithm.

The image data encoding / decoding apparatus shown in FIG. 6 is provided with the DCT base encoder 50,
The original image 82, which is two-dimensional image data, is encoded to generate encoded data 84. The DCT base encoder 50
The encoding of DCT (discrete cosine tran) is mainly used.
sform) transformation, quantization, entropy coding, etc.
It is to generate compressed image data.

The generated code data 84 is transferred, for example, via a transmission line 86. Alternatively, in the case of a database, it is stored in a predetermined storage device.

On the other hand, for example, the coded data 88 received from the transmission path 86 is reproduced by the DCT base decoder 60 as a reproduced image 92 which is two-dimensional image data. The DCT-based decoder 60 obtains the reproduced image 92 by performing entropy decoding, inverse quantization, and inverse DCT (IDCT) on the coded data 88 that is image compressed data. .

These DCT-based encoder 50 and DC
The T-based decoder uses a common quantization table and coding table. The quantization table and the encoding table are stored in the quantization table memory 72 or the encoding table memory 74, respectively.

The DCT base encoder 50 is mainly composed of a DCT converter 52, a quantizer 54, and an entropy encoder 56.

The DCT converter 52 first divides the original image 82 into blocks each of which has (8 × 8 = 64) pixels. Further, the DCT converter 52 is
Two-dimensional DCT transformation is performed on each divided block. By this two-dimensional DCT transform, for each block,
Conversion from two-dimensional image data to frequency component image data is performed. For example, the original image 82 as shown in FIG.
The two-dimensional image data Pxy of a certain block in the inside is converted into the frequency component image data Sxy as shown in FIG. 8 by such two-dimensional DCT conversion for each block.

The quantizer 54 includes the DCT converter 5
In order to improve the coding efficiency of the frequency component image data Sxy obtained in 2, the processing for reducing the data amount is performed. Thereby, for example, the frequency component image data Sxy as shown in FIG. 8 is converted into the quantized frequency component image data Rxy as shown in FIG. At this time, the quantization table recorded in the quantization table memory 72 is used.

This quantization table is, for example, as shown in FIG. The quantization table Qxy shown in FIG. 9 is an example corresponding to the frequency component image data Sxy shown in FIG. 8, and is used for conversion into the quantized frequency component image data Rxy shown in FIG. It is an example of what is used.

The entropy encoder 56 is intended to encode the quantized frequency component image data Rxy generated from the quantizer 54 while further reducing the data amount. In the JPEG algorithm, the Huffman coding method is used in the entropy encoder 56. The entropy encoder 56
Encoding the quantized frequency component image data Rxy output from the quantizer 54, for example, shown in FIG.
The code data 84 as shown in FIG. 11 is generated.

On the other hand, the DCT-based decoder 60 is mainly composed of an entropy decoder 62, an inverse quantizer 64, and an IDCT converter 66.

The entropy encoder 56 uses the table stored in the encoding table memory 74 at the time of such encoding. First, the entropy decoder 62 uses the same encoding table memory 74 as that used by the entropy encoder 56, and converts the code data 88 into quantized frequency component image data. For example, the code data 88 as shown in FIG.
Is converted into quantized frequency component image data Rxy as shown in FIG.

The dequantizer 64 dequantizes the quantized frequency component image data generated from the entropy decoder 62 to generate frequency component image data. For example, the quantized frequency component image data Rxy as shown in FIG. 12 is converted into frequency component image data Sxy. In such a conversion, the inverse quantizer 64 uses the same quantization table memory 72 as that used by the quantizer 54.

The IDCT converter 66 converts the frequency component image data generated by the inverse quantizer 64 into two-dimensional image data. For example, the frequency component image data Sxy as shown in FIG. 13 is converted into the two-dimensional image data Pxy as shown in FIG. The two-dimensional image data Pxy corresponds to one block in the reproduced image 92.

FIG. 15 shows the entropy encoder 56.
FIG. 3 is a block diagram showing a configuration particularly related to DC coefficient of FIG. That is, what is shown in FIG. 15 shows a part of the configuration of the entropy encoder 56 shown in FIG. 6, and shows a configuration related to the DC coefficient.

The entropy encoder 56 mainly
A block delay circuit 56a, a DC differentiator 56b, a grouper 56c, a one-dimensional Huffman encoder 56d,
It has a DC encoding table memory 56e.

The DC differencer 56b outputs the quantized frequency component output from the block delay circuit 56a, which is one block before the individual coefficient of the quantized frequency component image data of the input block. The difference from the corresponding coefficient of the image data is obtained, and this is set as the DC difference value.

The grouping unit 56c converts the DC difference value obtained by the DC difference unit 56b into a group number and an additional bit value according to the correspondence shown in FIG. The group number and the additional bit value thus obtained are stored in the DC code table memory 56.
While using e, it is converted into a DC code by the one-dimensional Huffman encoder 56d.

FIG. 16 is a block diagram showing the configuration of the portion of the entropy encoder 56, particularly regarding AC coefficients. That is, FIG. 16 shows the configuration of the portion of the entropy encoder 56 shown in FIG. 6, particularly regarding the AC coefficient.

The entropy encoder 56 includes a zigzag scan selection circuit 56k, a decision device 56m, a run length counter 56n, a grouper 56p, a two-dimensional Huffman encoder 56q, and an AC encoding table memory 5.
6r and.

The zigzag scanner 56k selects the target coefficient in zigzag with respect to the quantized frequency component image data generated by the quantizer 54. For example, for the quantized frequency component image data Rxy having (8 × 8 = 64) coefficients as shown in FIG. 10, the target coefficient is subjected to a zigzag scan as shown in FIG. Are sequentially selected.

The judging device 56m judges whether or not the coefficient values sequentially selected by the zigzag scan selecting circuit 56k are zero. If it is determined to be zero, the run length counter 56n sequentially counts the continuous lengths even though the coefficient value is zero. This result is
It is output as the run length NNNN. On the other hand, when it is judged by the judging device 56n that it is not zero, the grouping device 56p judges that the A-value as shown in FIG.
The value of the corresponding coefficient is converted into the corresponding group number and additional bit value according to the correspondence table between the C coefficient value and the group number and additional bit value. The obtained group number is input to the two-dimensional Huffman encoder 56q.

The two-dimensional Huffman encoder 56q uses the AC code table memory 56r to generate an AC code as shown in FIG.

According to the JPEG algorithm as described above, it is possible to achieve image compression of a two-dimensional image used in facsimile, videotex, etc., and reduce the amount of data transmitted through the transmission path 86, for example. It is possible to reduce the amount of data of the two-dimensional image stored in the database or the like.

[0029]

However, although image compression can be achieved by the JPEG algorithm as described above, it is desired that the degree of compression be further improved. Further, in the image data encoding device using such a JPEG algorithm, it is desired that the required hardware be simpler.

For example, in the case of a transmission line having a fixed data transfer rate, by reducing the amount of data to be transferred,
It is possible to reduce the transmission time of the two-dimensional image. Further, in a database or the like, by reducing the data amount of the two-dimensional image to be stored, it is possible to improve the number of two-dimensional images that can be stored and to reduce the storage capacity of the storage device. .

The present invention has been made to solve the above-mentioned conventional problems, and is generated from frequency component image data in an image data encoding device for frequency component image data converted from two-dimensional image data. It is possible to provide an image data encoding device capable of reducing the amount of coded data to be used and, if necessary, reducing the amount of hardware to be used, such as reducing the storage capacity of a memory that stores an encoding table to be used. To aim.

[0032]

According to the present invention, in an image data encoding device for frequency component image data converted from two-dimensional image data, quantization is performed in a quantization process for the frequency component image data. A quantizer that obtains frequency component image data, a quantization table memory that stores a quantization coefficient used in the quantizer, and at least the magnitude of the value of each coefficient that constitutes the quantized frequency component image data. Accordingly, the coefficient is coded, and when the value of the coefficient is smaller or larger and the deviation is larger, the information assigned to the code of the coefficient. The object is achieved by the provision of an entropy encoder whose quantity is reduced.

In the image data encoding device,
The entropy encoder is a coefficient range determiner for determining whether or not the coefficient of the quantized frequency component image data to be encoded is outside the range indicated by the minimum coefficient value and the maximum coefficient value, If the coefficient range determiner determines that the coefficient is out of the range, the coefficient to be coded is smaller than the minimum coefficient value, the coefficient is replaced with the minimum coefficient value, and the coefficient is larger than the maximum coefficient value. A coefficient value replacer for replacing the coefficient with the maximum coefficient value, wherein the entropy encoder encodes each coefficient of the quantized frequency component image data according to the output of the coefficient value replacer. The storage capacity of the encoding table used can be further reduced.

In the image data encoding device,
The encoding performed by the entropy encoder uses an encoding table showing the relationship between the magnitude of the coefficient value of the quantized frequency component image data and the corresponding code, The object is achieved by further comprising a coding table memory for storing the coding table in which the allocated information amount is reduced when the value of the coefficient is increased. At the same time, the configuration can be made almost the same as the conventional one, and the application of the present invention to the conventional image data encoding device is facilitated.

[0035]

Focusing on the quantized frequency component image data output from the quantizer 54 shown in FIG. 6, for example, regarding the quantized frequency component image data Rxy shown in FIG. , There is a certain concentration distribution in the value. For example, each coefficient shown in FIG.
Most of the values are “0”. Also, the absolute values of the other values are limited to the range of about "10".

However, with respect to the coefficient of the quantized frequency component image data Rxy having such characteristics, the encoding performed by the entropy encoder 56 is uniform.

For example, as described above, the DCT coefficient, that is, the value of the DC difference value shown in FIG. 18 is, for example, in the extremely limited range whose absolute value is about "10" or less. , Assuming that the absolute value would be very large, more bits were allocated. For example, in FIG. 18, the absolute value of the DC difference value is assumed to be in the range of “32767”. For this reason, the maximum value of the group number is set to a larger value and the number of additional bits is also increased corresponding to the DC difference value having a large absolute value that hardly occurs. For this trend,
The same applies to the AC coefficient as shown in FIG. 19, for example.

Focusing on such a point, in the present invention, the coefficient (DCT coefficient, that is, DC difference value or AC coefficient) of the quantized frequency component image data output from the quantizer 54 is Non-linear coding is used.

Conventionally, when encoding the quantized frequency component image data, one group number is assigned to each DC difference value or each AC coefficient value as shown in FIG. 18 and FIG. 19, for example. And additional bit values were assigned. However, according to the present invention, the DC difference value and the AC difference value, which hardly cause such a value and have a larger deviation from the average value, are obtained.
As for the coefficient value, multiple adjacent values are grouped together,
For example, one group number or one additional bit value is assigned.

As described above, even if the DC difference value and the AC coefficient value whose values are hard to occur and whose deviation is large with respect to the average value are put together, the deterioration of the two-dimensional image to be encoded. The inventor has found that is relatively small. Further, in this way, by processing the magnitude of the value of each coefficient in the coding of the coefficient of the quantized frequency component image data in a non-linear manner, it is possible to reduce the obtained data amount. Further, this can also reduce the hardware used.

The present invention does not specifically limit such non-linear handling, and may be handled stepwise as in the embodiment described later, or may be handled logarithmically. That is, when the deviation of the target coefficient becomes large, the amount of information to be allocated may be reduced.

[0042]

Embodiments of the present invention will be described in detail below with reference to the drawings.

FIG. 1 is a block diagram showing a configuration of an application part of the present invention in an embodiment of an image data encoding device to which the present invention is applied.

In this embodiment, the DC shown in FIG. 6 is used.
Based on the T-base encoder 50, the entropy encoder 56 has a configuration as shown in FIG. Therefore, the operation of the other parts is the same as that described above with reference to FIGS. 6 to 17.

In the present embodiment, when the quantized frequency component image data obtained from the quantizer 54 is encoded by the entropy encoder 56, the entropy encoder 56 is provided with Each coefficient is converted by the configuration shown in FIG. That is, for example, in FIG.
For each value of the quantized DCT coefficient as shown in FIG. 2, the value is replaced by the configuration shown in FIG.

In FIG. 1, the comparators 12 and 14 included in the entropy encoder 56 and the selector 16 are included.
And 18.

First, in FIG. 1, the difference value of the DCT coefficient to be processed is each coefficient of the quantized frequency component image data output by the quantizer 54. For example, the (8 × 8 = 64) Ds shown in FIG.
C coefficient. The minimum coefficient value and the maximum coefficient value are set according to the distribution of the difference value of the DC coefficient, and for example, the range of the value centering on the average value of the difference value of the DCT coefficient is set. is there. For example, in the following description, the minimum coefficient value is "-128" and the maximum coefficient value is "128".

In FIG. 1, first, the comparator 12
Compares the difference value of the DC coefficient with the minimum coefficient value. When the difference value of the DCT coefficient is smaller than the minimum coefficient value, the signal LT becomes “1”.

The comparator 14 compares the difference value of the DCT coefficient with the maximum coefficient value. When the difference value of the DCT coefficient is larger than the maximum coefficient value, the signal GT becomes “1”.

The selector 16 operates according to the signal LT output from the comparator 12. That is, the selector 1
When the signal LT becomes "0", the selector 6 selects the difference value of the DCT coefficient and outputs it. On the other hand, the signal L
When T becomes "1", the selector 16 outputs the minimum coefficient value, for example, "-128".

The selector 18 operates according to the signal GT output from the comparator 14. That is, the signal G
When T is "0", the selector 18 selects the signal output by the selector 16. On the other hand, when the signal GT is "1", the selector 18 causes the maximum coefficient value,
For example, "128" is selected.

In FIG. 1, the comparators 12 and 14 constitute a coefficient range judging device of the present invention.
Further, the selectors 16 and 18 constitute the coefficient value replacing device of the present invention.

According to the coefficient range deciding device and the coefficient value replacing device as shown in FIG. 1, when the difference value of the DCT coefficient becomes equal to or less than the minimum coefficient value, the difference value of the DCT coefficient is uniformly changed. Replaced by the minimum coefficient value. For example, when the minimum coefficient value is "-128" and the difference value of the DCT coefficients is "-128" or less, the DCT coefficient difference value is replaced with "-128".

On the other hand, when the difference value of the DCT coefficient becomes larger than the maximum coefficient value, the difference value of the DCT coefficient is replaced with the maximum coefficient value.

FIG. 2 is a circuit diagram showing an example of the comparator 14.

In the present embodiment, when the value of the maximum coefficient value is "128" and the bit width of the difference value of the DCT coefficient is 16 bits, the comparator 12 is as shown in FIG. A 5-input OR logic gate 32 can be used. The OR logic gate 32 has X7 to X11 among the bits X0 to X15 of the difference value of the DCT coefficient.
Is entered.

FIG. 3 is a circuit diagram showing an example of the comparator 12.

In this embodiment, the minimum coefficient value is "-
128 "and the bit width of the difference value of the DCT coefficient is 16 bits, as shown in FIG.
It can be constituted by a 5-input NAND logic gate 34. The bits of the difference value of the DCT coefficient are X0 to X1.
5, the NAND logic gate 34 has X7-X.
11 is input and X0 is input to the NAND logic gate 36.
~ X6 is input.

FIG. 4 is a diagram showing the relationship between the DC difference value and the group number / additional bit value in this embodiment. FIG. 5 is a diagram showing the relationship among the AC coefficient value, the group number and the number of additional bits in this embodiment.

As is apparent by comparing FIGS. 4 and 5 with FIGS. 18 and 19 described above, in the present embodiment, the groups with group numbers 8 and above are different from the conventional ones. That is, conventionally, a total of nine groups, that is, group numbers 8 to 16 are grouped into one group number 8 in this embodiment.
In addition, regarding the group number 8 collected in this way,
The DC difference value is "-128" or less (that is, "-32767").
To "-128") or "12"
8 "or more (that is," 128 "to" 32767 ", or
Only "128" to "32768") are distinguished, and the number of additional bits for this is only 1 bit.

As described above, according to the first embodiment, by providing the configuration of FIG. 1 in the entropy encoder 56, as shown in FIG. 4 and FIG. The absolute value of the DC difference value or the AC coefficient value is 1
For 28 or more, one group is set, and for such absolute values of 128 or more, the amount of information to be assigned is reduced. As a result, the data amount of the code generated from the frequency component image data can be reduced. Further, as shown in FIG. 4 and FIG. 5, the coding table can be reduced, the storage capacity of the memory for storing the coding table can be reduced, and the hardware to be used can be reduced. Can be reduced.

In applying the present invention, instead of providing the structure shown in FIG. 1, the structure shown in FIG.
Alternatively, it is possible to deal with the data setting of the encoding table shown in FIG. That is, the correspondence between the group number and the additional bit value corresponding to the DC difference value shown in FIG. 18 and the correspondence between the group number and the additional bit value corresponding to the AC coefficient value shown in FIG. This is achieved by changing the table data in FIGS. 18 and 19. That is, when the absolute value of the DC difference value or the DC coefficient value is 128 or more, the table data is set so that the group number corresponds to 8. Such data setting can be performed without changing the conventional hardware configuration. Therefore, the present invention can be more easily applied to an image data encoding device having a conventional configuration. The application of the present invention by setting the encoding table in this way is referred to as a second embodiment.

As described above, in the first embodiment, the D
When the absolute value of the C difference value is 128 or more and the absolute value of the AC coefficient value is 128 or more, one group number, that is, the group number 8 is assigned. The following modifications are similar to this.

That is, the DC difference value is "-32767".
Up to "-128" have the same group number as the value of "-127", that is, group number 7, and have the same additional bit value. Further, the DC difference value is "12.
8 "to" 32767 "are assigned the same group number 7 and the same additional bit value as the DC difference value" 127 ".

Regarding the AC coefficient value, "-32"
767 "to" -128 "are the AC coefficient values"-"
The same group number 7 and the same additional bit value as those of 127 "are assigned. Also, the AC coefficient value" 128 ".
It is also conceivable that the same group number 7 and the same additional bit value as the AC coefficient value “127” are assigned to the “˜32768”.

When such an allocation is performed, the minimum coefficient value input to the comparator 12 shown in FIG. 1 is "-127", and the maximum coefficient value input to the comparator 14 is the maximum coefficient value. The coefficient value is “127”. In addition, such a thing is hereafter called a 3rd Example.

According to the third embodiment, the required group number is only "0" to "7", and the bit width can be represented by 3 bits. It can be shortened by 1 bit compared to. Therefore,
According to the third embodiment, the data amount of the code generated from the frequency component image data is
It is possible to reduce more.

[0068]

As described above, according to the present invention,
In an image data encoding device for frequency component image data converted from two-dimensional image data, the data amount of a code generated from the frequency component image data is reduced, and an encoding table to be used is used as necessary. It is possible to obtain an excellent effect that it is possible to reduce hardware to be used, such as reduction in storage capacity of a memory to be stored.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration used in a first embodiment of an image data encoding device to which the present invention is applied.

FIG. 2 is a circuit diagram showing a comparator 12 used in the first embodiment.

FIG. 3 is a circuit diagram showing a comparator 14 used in the first embodiment.

FIG. 4 is a diagram showing a grouping of difference values of DC coefficients of the encoding table used in the first embodiment.

FIG. 5 is a diagram showing a grouping of AC coefficients, particularly of AC coefficients, in the coding table used in the first embodiment.

FIG. 6 is a block diagram showing the configuration of a conventional image data encoding / decoding device using a JPEG algorithm.

FIG. 7 is a diagram showing an example of data of a block of a two-dimensional image used in the conventional image data encoding / decoding device.

FIG. 8 is a diagram showing an example of one block of DCT coefficients after DCT conversion obtained from the conventional image data encoding / decoding device.

FIG. 9 is a diagram showing an example of a quantization table for one block used in the conventional image data encoding / decoding device.

FIG. 10 is a diagram showing an example of DCT coefficients for one block obtained by the quantizer of the conventional image data encoding / decoding device.

FIG. 11 is a diagram showing an example of one block of code data generated by an entropy encoder of the conventional image data encoding / decoding device.

FIG. 12 is a diagram showing an example of a quantized DCT coefficient of one decoded block obtained by an entropy decoder of the image data encoding / decoding device.

FIG. 13 is a diagram showing an example of DCT coefficients for one block generated by an inverse quantizer of the conventional image data encoding / decoding device.

FIG. 14 is a diagram showing an example of decoded two-dimensional image data of one block obtained by the IDCT converter of the conventional image data encoding / decoding device.

FIG. 15 is a block diagram showing a configuration related to a DC coefficient used in an entropy encoder of the conventional image data decoding / encoding device.

FIG. 16 is a block diagram showing a configuration related to AC coefficients used in the entropy encoder of the image data encoding / decoding device.

FIG. 17 is a diagram showing a zigzag scan performed by the entropy encoder of the conventional image data encoding / decoding device.

FIG. 18 is a diagram showing a portion related to DC coefficients of an encoding table of the conventional image data decoding / encoding device.

FIG. 19 is a diagram showing a portion related to AC coefficients of an encoding table used in the image data encoding / decoding device.

[Explanation of symbols]

12, 14 ... Comparator 16, 18 ... Selector 32 ... 5-input OR logic gate 34 ... 5-input NAND logic gate 36 ... 7-input NAND logic gate 38 ... OR logic gate 50 ... DCT base encoder 52 ... DCT converter 54 Quantizer 56 Entropy coder 56a Block delay circuit 56b DC differencer 56c Grouper 56d One-dimensional Huffman coder 56e DC coding table memory 56k Zigzag scan selection circuit 56m Judger 56n ... Run length counter 56p ... Grouper 56r ... AC encoding table memory 56q ... Two-dimensional Huffman encoder 60 ... DCT base decoder 62 ... Entropy decoder 64 ... Inverse quantizer 66 ... IDCT converter 72 ... Quantization table memory 74 ... Encoding table memory 82 ... Two-dimensional Image data of one block of image (original image) 84, 88 ... Code data 86 ... Transmission line 92 ... Image data of one block of two-dimensional image (reproduced image)

Claims (3)

[Claims] 1. An image data encoding device for frequency component image data converted from two-dimensional image data, wherein quantization is performed to obtain quantized frequency component image data by a quantization process on the frequency component image data. And a quantization table memory for storing the quantization coefficient used in the quantizer, and for each coefficient constituting the quantized frequency component image data, the coefficient is encoded according to at least the magnitude of the value. In addition, the amount of information assigned to the code of the coefficient is reduced when the value of the coefficient becomes smaller or larger in the encoding and the deviation becomes larger. An image data encoding device comprising an entropy encoder. 2. The entropy encoder according to claim 1, wherein the coefficient of the quantized frequency component image data to be encoded is outside a range indicated by a minimum coefficient value and a maximum coefficient value. A coefficient range determiner for determining whether or not the coefficient range determiner determines that the coefficient is out of range, and when the coefficient to be encoded is smaller than the minimum coefficient value, replaces the coefficient with the minimum coefficient value, A coefficient value replacer that replaces the coefficient with the maximum coefficient value when the coefficient is greater than the maximum coefficient value, and the entropy coder according to the output of the coefficient value replacer, the quantized frequency component image data. An image data encoding device, which encodes each coefficient of. 3. The code according to claim 1, wherein the encoding performed by the entropy encoder indicates the relationship between the magnitude of the coefficient value of the quantized frequency component image data and the corresponding code. And a coding table memory for storing the coding table in such a relationship that the amount of information to be allocated is reduced when the value of the coefficient is increased. An image data encoding device provided with.