JPH04329765A - Data conversion circuit - Google Patents

Abstract

PURPOSE:To simplify the data conversion circuit by using an input data as an address so as to designate the address when a control signal is a prescribed value thereby outputting a conversion data corresponding to the address and outputting the input data when the control signal is not the prescribed value. CONSTITUTION:In the case of coding a data with a coding ROM 11A, a data at the outside of a coding object and a coding object data are inputted thereto. When a control signal represents the normal mode, the data is coded to output a variable length coding data and its code length. Moreover, when the control signal represents the path mode, an address is outputted as it is as a fixed length data and a code length is outputted based on the control length. On the other hand, in the case of decoding a data by a decoding ROM 12, a code data string in which fixed length and variable length data are in existence in mixture is inputted thereto. When the control signal represents the normal mode, a data decoding a head code data of the input code data string into the original data and the decoded code length are outputted and when the control signal represents the path mode, the data is outputted as it is.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data conversion circuit for encoding data into code data such as a Huffman code, and for decoding code data.

0002

2. Description of the Related Art In recent years, for example, in the medical field, there has been an active movement to digitize and handle medical images including X-ray images. This digitized X-ray image has a large amount of data, and in order to store, search, or transmit this X-ray image, it is necessary to efficiently reduce the amount of data and prevent the deterioration of the X-ray image. Image compression technology that reduces the number of images is becoming extremely important. As such a compression technique, for example, DPCM (dif
ferential pulsecode module
ation) method, and DCT (di
(screte cosine transform)
There are methods, etc. In such a DPCM method, compression/
Although the image does not deteriorate during decompression, the compression ratio is, for example, about 1/2. In addition, in the DCT method, although the image data is irreversible, the compression rate is, for example, 1/5 to 1/50.
degree and high compression ratio.

[0003] When compressing image data using the DPCM method, the difference between two adjacent pixels is determined and code data is assigned to the value of the difference. Although there are many types of code data for this compression/expansion, relatively good results can be obtained in terms of efficiency by using variable length code data. As this variable length code, for example, a Huffman code, a run length (hereinafter referred to as RL) code, etc. are well known. Compressed data is obtained by data packing this variable length code data together with the pixel value at the starting point. Furthermore, when decompressing compressed data into image data, the process inverse to the above process may be performed.

[0004] When compressing image data using the DCT method, the image data is converted into the frequency domain by discrete cosine transformation, and after truncating unnecessary components and performing quantization processing, the image data is compressed into a variable length Compressed data is obtained by encoding the encoded data and packing the encoded data. In addition, when expanding this compressed data into image data, the reverse process to the above process may be performed.

In FIG. 4A, which shows an encoding ROM 1 that encodes code data as a conventional data conversion circuit,
When encoding data to be encoded, the first bit of the data to be encoded of k bits is assigned to port A1-1 or A1.
-k is input as an address value to ports A1-1 to A1-k of the encoding ROM 1 via the address bus. The encoding target data input to the encoding ROM 1 is, for example, DPC.
This data is difference data between pixels in the M method, or quantized transform coefficient data or zero run length in the DCT method, and is encoded into variable length code data. Also encoded ROM
1, all the data to be encoded input to ports A1-1 to A1-k are used as addresses, and each address contains variable length code (Huffman code or RL code) data and variable length code data. Data indicating the code length (hereinafter referred to as code length) is written in advance. Therefore, an address is specified in the data to be encoded inputted to ports A1-1 to D1-k, and code data such as a Huffman code written in advance to the specified address is transferred to ports D1-1 to D1-r.
The code length, which is the effective number of bits of code data, is output from ports D2-1 to D2-s.

Furthermore, in FIG. 4B, which shows a decoding ROM 2 for decoding code data as a conventional data conversion circuit, when decoding from code data, the leading bit is transferred to port A3-1 or A3- of the decoding ROM 2. An r-bit code data string (or code string) is input as code data to ports A3-1 to A3-r. This code data string is output from a shift circuit (not shown) that can shift arbitrary bits, and is a continuous data string of packed variable-length code data (sometimes fixed-length code data is mixed). . Decryption ROM2 has port A3-1.
All the code data strings input to ~r are used as addresses, and the decoded data of the variable length code at the beginning of the code data string and the code length of the variable length code are written in advance in each address. Therefore, the code data string input to ports A3-1 to A3-r is specified as an address, and the decoded data of the first code of the code data string written in advance to the specified address is output from ports D3-1 to D3-k.
The code length of variable length code data corresponding to the decoded data is output from ports D4-1 to D4-s.

Next, an encoding circuit that performs encoding using such encoding ROM 1 will be explained based on FIG. 5. In FIG. 5, the aforementioned encoding ROM 1 inputs data to be encoded, outputs encoded data to a multiplexer (hereinafter referred to as MUX) 3, and outputs code length data to MUX 4. MUX 3 inputs code data from encoding ROM 1, and also inputs the pixel value itself (in the DPCM method, for example).
data not to be encoded, which is fixed-length code data such as data of the first pixel) or DC component data itself in the DCT method and which is not converted to variable-length code data, and inputs the data to be encoded based on the control signal. One data is selected from the data and the data not to be encoded, and the data to be shifted is outputted to the shift circuit 5. In addition, MUX 4 inputs the code length output from the encoding ROM 1 and also inputs the non-encoding data length of the non-encoding data, and uses the control signal to combine the code length and non-encoding data length into one piece of data. is selected and the shift number, which is the code length, is output to the shift circuit 5. The shift circuit 5 then packs the data based on the input data to be shifted and the number of shifts, and outputs the packed data to a code buffer (not shown).

Next, the operation will be explained. When data is encoded and packed, for example in the DPCM method, the difference data between pixels is encoded into a variable length code such as a Huffman code and packed, but for example, the first pixel value itself of each line is not encoded. Packed as fixed-length code data. In addition, in the DCT method, transform coefficient data after quantization, zero-run length data, etc. are similarly encoded and packed into variable-length codes such as Huffman codes, but one of the Huffman codes is assigned to an escape code, and There are cases where data is packed as code+fixed length data, or where DC component data itself is packed. Therefore, the DPCM method, DC
In both the T methods, there is a mixture of data packed with variable length code and data packed with fixed length code data. Therefore, for example, in the DPCM method, difference data between pixels, quantized transform coefficient data, zero run length data, etc. in the DCT method are output to the encoding ROM 1 as encoding target data and encoded into variable length code data. After that, it is output to MUX3, and on the other hand, the encoded R
The code length of the variable length code output from OM1 is MUX4
Output to. On the other hand, the pixel value itself in the DPCM method, the DC component data itself in the DCT method, the fixed length data following the escape code, etc. are output to the MUX 3 as non-encoding data, and the code length (or data length) is output to MUX4. Then, MUX 3 selects one piece of data from the coded data and non-encoding target data and outputs it to shift circuit 5. The code length of the selected data is selected and output from MUX 4 as a shift number to shift circuit 5.
Then, the fixed length data and variable length data are packed and data compression is performed.

[0009]

[Problems to be Solved by the Invention] By the way, in conventional data conversion circuits, fixed-length data that is not coded and variable-length data that is coded are mixed, so coded data and non-coded data are mixed. Two MUX3 to select
, 4 are required, making the circuit configuration complicated.

Furthermore, the conventional data conversion circuit is provided with one conversion table, and variable length code data (decoded data in the case of a decoding ROM) and code length written in advance are encoded and encoded according to this conversion table. It has been decrypted. However, the overall properties and local properties of input image data vary in various ways, such as in medical images, but if encoding and decoding are performed based on one conversion table in this way, the properties of input image data will change. There is a risk that they will not be able to respond to changes.

The present invention has been made in view of the above-mentioned conventional problems, and an object of the present invention is to provide a data conversion circuit whose circuit configuration is simplified. Another object of the present invention is to provide a data conversion circuit that can select an optimal conversion table when the nature of input image data changes in various ways, such as in medical images.

0012

[Means for Solving the Problems] Therefore, the present invention has an input address port and an output data port, inputs fixed length data as input data to the input address port, encodes it, and outputs encoded data and output data from the output data port. At least one of the code lengths of the code data is output as output data, or a code data string in which the code data is continuous is sequentially input to an input address port as input data and decoded to combine the fixed length data before encoding and the code data. In a data conversion circuit having a function of outputting at least one of the code length of the code data to be decoded out of the code data string from the output data port as output data, the address on the input side is for data input. It consists of an address and an address for control signal input, which have at least two different characteristics, and the output data is switched depending on the control signal value input to the control input address.

[0013] Furthermore, when the control signal is a predetermined value, the data input to the input address is converted and output, and when the control signal is not a predetermined value, the input data is output as is as output data. Good too. Or it can have multiple conversion tables for input data and output data,
One of the plurality of conversion tables may be specified by the control signal, and input data may be converted into output data based on the conversion table.

[0014] Alternatively, the control signal may be based on the image size, the density range of the image, the number of bits per pixel of the image, the type of image, the photographing conditions of the image, the statistical properties of the whole or local part of the image, the quantization width, etc. The conversion table may be generated according to one or a combination of conditions related to the compression ratio, and the conversion table may be specified by the generation condition.

Alternatively, one or more of the conversion tables may be a non-conversion table, and when the non-conversion table is designated, the input data may be output as is as output data. This data conversion circuit may be applied to a data compression/expansion device. Further, this data conversion circuit may be applied to data conversion processing of medical images.

[0016]

[Operation] According to the above structure, when converting data, the output data is switched according to the control signal value input from the control signal input address port. In the case where the control signal value is switched between a predetermined value and when it is not a predetermined value,
When the control signal value is a predetermined value, input data input from the data input address port is addressed as an address value, and converted data corresponding to the designated address is output. Furthermore, when the control signal value is not a predetermined value, the input data is output as is. This makes it possible to switch between data to be converted and data not to be converted by a control signal, thereby eliminating the need for a conventional multiplexer.

[0017] Furthermore, in the case where a conversion table is specified by a control signal value, multiple conversion tables can be selected according to the characteristics of the image, so that high conversion efficiency (encoding efficiency) can be maintained even if the characteristics of the image quality change. It becomes possible to obtain. Further, when the control signal value is generated based on the overall or local statistical properties of the input image data, the conversion table is specified by the generation condition.

[0018] Furthermore, in the case where a control signal is generated according to conditions related to compression ratio such as quantization width, a change table including quantization and inverse quantization functions can be created, and the circuit size can be reduced. . Furthermore, in the case of a non-conversion table, input data is output as output data, so a wider range of selection is possible.

By applying this data conversion circuit to a data compression/expansion device and further to data conversion processing in the medical field, it becomes possible to efficiently process a large amount of data.

[0020]

[Embodiments] Hereinafter, embodiments of the present invention will be explained based on FIGS. 1 to 3. Incidentally, the same elements as those in FIGS. 4 and 5 are given the same reference numerals, and the description thereof will be omitted. In FIG. 1 showing an encoding ROM 11A as a data conversion circuit of the first embodiment, the encoding ROM 11A is a conventional data input port A.
In addition to 1-1 to k, code data output ports D1-1 to r, and code length output ports D2-1 to s, for example, n bits for distinguishing between data to be encoded and data not to be encoded. control signal input address ports A2-1 to A2-n for inputting control signals. In order to distinguish between data to be encoded and data not to be encoded, the number of bits n for inputting a control signal may be "1", but n is set to a plurality, for example, when processing image data in the medical field. This is to cope with the case where data with a different number of bits, such as 8, 10, or 12 bits, is input so that it can be specified by a control signal. By setting the number of bits to n, (2n −1
) fixed-length data with different code lengths can be handled. Then, the code length registered as the data length to be encoded based on the value of the input control signal is determined by the port D2-1.
~s is output. In the encoding ROM 11A, all the data to be encoded input to the ports A1-1 to A1-k are used as addresses, and each address corresponding to the normal mode, which will be described later, has a corresponding variable length code. (Huffman code or RL code) data and data indicating the code length of variable length code data are written in advance. The address values of the data input ports A1-1 to A1-k and the code length determined by the control signal value are written in each address corresponding to a pass mode in which the control signal will be described later. If the value of the control signal input to the encoding ROM 11A is the normal mode (for example, a predetermined value "
0), input data is addressed as an address value, code data corresponding to the specified address is output, and the code length is output, and when the control signal value is pass mode (predetermined value 0 ''), the input address value itself and a predetermined code length determined by the control signal value are output.

[0021] Also, a decoding ROM as a data conversion circuit
In FIG. 1(B) showing the decoding ROM 12A, the decoding ROM 12A
In addition to the conventional code data string input ports A3-1~r, decoded data output ports D3-1~k, and code length output ports D4-1~s, the original data of the decoded data to be output is Control signal input port A4-1 to input a control signal of, for example, n bits to specify whether the data is to be encoded, that is, variable-length code data, or data that is not to be encoded, that is, fixed-length data. It has n. Port A3
The code data string (or code string) input to −1 to r is output from the shift circuit as before, and is a mixture of packed variable-length code data and some fixed-length data. is a continuous code data string. As in the case of the encoding target ROM 11A, in order for the original data of the decoded data to specify the encoding target data and the non-encoding target data, the number of bits inputting the control signal may be "1", but n The reason why there are multiple numbers is to cope with different numbers of bits of decoded data. Decryption ROM12A has port A
All the data to be decoded input to 3-1 to r, that is, the code data string, is used as an address, and each address where the control signal indicates normal mode corresponds to the variable length code data at the beginning of the input code data string. The code length of the decoded data and the decoded variable length code data are written in advance. For each address where the control signal indicates the pass mode, the address value itself of the code data string input ports A3-1 to A3-r,
and the code length determined by the control signal value. If the value of the control signal input to the decoding ROM 12A is the normal mode described above, an address is specified using the input code data string, and decoded data corresponding to the specified address is output, and the decoded code is output. The code length of the data is output, and when the value of the control signal is the above-mentioned pass mode, the address value which is the input data to be decoded is output as is, and the code length determined by the control signal value is output.

Next, the operation will be explained. In FIG. 1A, when k-bit data is input as an address value to ports A1-1 to A1-k of the encoding ROM 11A with the first bit set to port A1-1 or A1-k via the address bus, port A2 When a normal mode control signal is input to −1 to n, the input data is, for example, DP
This data is the difference data between pixels in the CM method, the transform coefficient data after quantization, the zero run length data, etc. in the DCT method, and is the data to be encoded to be encoded into variable length code data, and is input to ports A1-1 to A1-k. The code data written in advance at the specified address is output from ports D1-1 to r, and the code length of the code data is output from ports D2-1 to s. Furthermore, when a pass mode control signal is input to ports A2-1 to A2-n, the input data is, for example, the pixel value itself in the DPCM method, or the DC value in the DCT method.
This is data that is not subject to encoding, such as component data itself or fixed-length data following an escape code, and port A
The non-encoding data values input to 1-1 to 1-k, that is, the address values of the encoding ROM 11A, are directly transferred to port D1-
The code length is specified by a control signal and output from ports D2-1 to D2-s.

In FIG. 1(B), the decoding ROM 12A
When r-bit data is input as an address value to ports A3-1 to A3-r via the address bus with the first bit set to port A3-1 or A3-r, port A4
When the normal mode control signal is input to −1 to n,
The first data of the input code data string is variable length code data, and the code data string input to ports A1-1 to A1-r is specified as an address value, and the code data written in advance at the specified address is It is output from ports D3-1 to D3-k. As in the past, the code length of coded data is unknown until decoding is completed, but the code length of the coded data, such as Huffman code or RL code, can be determined from the coded data. , and its code length is output from ports D4-1 to D4-s. Also port A4-1
When the pass mode control signal is input to ~n, the first data of the input code data string is fixed length data,
Address data input to ports A3-1 to A3-r is output as is from ports D3-1 to D3-k. At this time, the value registered as the code length of the data to be decoded is specified based on the value of the control signal, and the code length is
~s is output.

Next, an encoding circuit using such an encoding ROM 11A will be explained based on FIG. 2. The encoding ROM 11A inputs the above-mentioned data not to be encoded or data to be encoded and the control signal, and inputs to the shift circuit 5 the data to be shifted which is variable length code data or fixed length data and the shift number which is the code length thereof. Outputs . The shift circuit 5 packs data based on the data to be shifted and the number of shifts, and outputs the packed data to a code buffer (not shown), as in the conventional case.

According to this configuration, the encoding ROM 11A
When encoding, input data not to be encoded and data to be encoded, and when the value of the control signal is in normal mode, encode it and output the variable length code data and its code length, and the value of the control signal When in pass mode, the address value is output as is as fixed length data, and its code length is output based on the value of the control signal.Furthermore, when decoding with decoding ROM 12A, fixed length data and variable length data are When a mixed code data string is input and the control signal value is normal mode, the first code data of the input code data string is decoded to the original data and the code length of the decoded code is calculated. When the control signal value is in pass mode, the address value is output as the original data, and the code length is output based on the control signal value, thereby converting the data to be converted (variable length code data). Since the decoding method for data (fixed length data) not subject to data conversion can be switched by a control signal, a conventional multiplexer is not required, and the circuit configurations of the encoding circuit and decoding circuit can be simplified.

Next, a second embodiment will be explained. This data conversion circuit of the first embodiment, which inputs a control signal, is further provided with the ability to select a table number depending on the application, thereby performing optimal compression/expansion. In FIG. 3A showing the encoding ROM 11B, the encoding ROM 11B has ports A1-1 to A1-k for inputting data to be encoded and ports D1-1 to D1-1 for outputting encoded data.
r, ports D2-1 to s that output the code length of code data
, and control signal input ports A2-1 to A2-n, as well as t-bit ports A5-1 to A5-t for inputting table numbers. Set the number of bits of the port for inputting the table number to t
By doing so, it is possible to select 2t encoding tables with different characteristics. And encoding ROM11B
Code data of 2t encoding tables and their code lengths, which will be described later, are written to each address where the control signal corresponds to the normal mode, and in addition to the input data to be encoded and the control signal, If the control signal is in the normal mode, code data and its code length are output based on the table number inputted in ~t.

FIG. 3(B) also shows the decoding ROM 12B.
In the decoding ROM 12B, ports A3-1 to r input data to be decoded, ports D3-1 to k output decoded data, ports D4-1 to D4-s output the code length of encoded data, and a control signal input. In addition to ports A4-1 to A4-n, there is also a t-bit port A6-n for inputting the table number.
1 to t. Then, the control signal of the decoding ROM 12B is set to the above-mentioned 2t at each address corresponding to the normal mode.
The decoded data of the decoding table corresponding to the encoding table and the code length of the code data to be decoded are written, and in addition to the input data and control signal, the port A6-1
When the control signal is in normal mode, the decoded data and the code length of the code data to be decoded are output based on the table number inputted from ~t.

Next, the operation will be explained. The fineness of the image and the degree of change in the image data are determined by, for example, the type of image to be compressed (medical image, non-medical image), the image size (256 x 256, 512 x 51
2, 1k x 1k, 2k x 2k, 4k x 4k), and regarding medical images, imaging equipment (X-ray images, CT images,
MRI image, etc.), imaging site (chest image, abdominal image, head image, limb bone image, etc.), A/D density range (density 0 to 2
, density 0 to 3, density 0 to 4), the number of bits per pixel of image data (for example, 8 to 16 bits/pixel), and the degree of correlation of data within the image also changes. For example, when performing A/D conversion on X-ray film,
The wider the A/D density range, the larger the image size, and the smaller the number of bits per pixel, the stronger the correlation between adjacent pixels becomes. , the variance of the histogram of the DCT coefficients (
(spread of distribution) tends to decrease. Furthermore, if the degree of correlation between image data differs, the optimal code table will naturally change as well. Therefore, it is preferable to prepare a plurality of code tables corresponding to predicted changes in image data in advance and to select one of these tables.

Particularly when the DPCM method or the DCT method is used, the histogram of the difference values between pixels and the AC component coefficients of the DCT becomes a Laplace distribution or a Gaussian distribution with a peak at zero. By selecting several types of variance values (spreading of the distribution) of these distributions and preparing the optimal Huffman code table for each, the image characteristics shown above (image type, image size, etc.) can be achieved. It becomes possible to obtain optimal code data according to the requirements. In addition, these difference values, D
The histogram of AC component coefficients of CT varies from image to image and also changes depending on the local area of the image. For example, the variance values of these histograms become large in images with rapid density changes or in local areas of the image. Therefore, for example, the properties of the entire image or local parts may be investigated using, for example, the difference value, the sum of squares or the sum of absolute values of AC coefficient values, and the table may be selected based on this statistical property.

In FIG. 3(A), ports A1-1 to A1-k
, the data to be encoded and the control signal are input from ports A2-1 to A2-n, and the table number is input from ports A5-1 to A5-t to select a table. Optimal code data and code length are output from D1-1 to r and D2-1 to s. Also, Figure 3 (
In B), ports A3-1 to r, ports A4-1 to
The data to be decoded and the control signal are input from ports A6-1 to D3-k, respectively, and the table number is input from ports A6-1 to D3-t to select a table. Decoded data and code length are output from D4-1 to D4-s.

[0031] According to this configuration, by writing a table according to the application into the encoding ROM 11A and the decoding ROM 12B and inputting the table number, it is possible to perform optimal encoding and improve encoding efficiency. I can do it. In the second embodiment, the input data, control signal, and table number are input to the data conversion circuit, but only the table number may be input as the control signal. In this case, a non-conversion table is prepared in the data conversion circuit to output input data as is, and the non-conversion table can be selected. Furthermore, a quantization table may be added as the encoding table. When a quantization table is used, the data before quantization is input as a ROM address to ports A1-1 to A1-k of the encoding ROM 11B, and
Specify the quantization table number using all or part of -1 to t. Then, a quantization width is selected based on the specified quantization table number, quantization is performed, and after the quantization process, an encoding process is performed, and the obtained code data and its code length are output. By adding a quantization table in this way, it is possible to perform both quantization processing and encoding processing using the same encoding ROM. Similarly, the quantization table number may be used during decoding, and the decoding ROM may be configured to perform dequantization processing and decoding processing at the same time.

[0032] In the above first and second embodiments, all ROMs are used.
Although a case has been described in which a logic circuit is used, the present invention is not limited to this, and a logic circuit may also be used. That is, it can be realized by inputting an address value input to the ROM to a logic circuit, and converting it into a data value output by the ROM in the logic circuit.

[0033]

As explained above, according to the present invention, when converting data, when the control signal input from the control signal input address port has a predetermined value, the input from the data input address port Addresses data as an address value, outputs converted data corresponding to the specified address, and when the control signal is not a predetermined value,
By outputting input data as is, it is possible to distinguish between data subject to data conversion and data not subject to data conversion using a control signal, eliminating the need for a conventional multiplexer and simplifying the configuration of a circuit using this data conversion circuit. You can.

[0034] Furthermore, by specifying the number of a plurality of conversion tables (for example, a plurality of Huffman tables) using a part of the input address, the optimum conversion table can be used according to the nature of the image data. I can do it. Further, by combining both of these functions, a more preferable data conversion circuit can be constructed. Further, by incorporating a quantization or inverse quantization function, a quantization or inverse quantization circuit can be omitted.

[Brief explanation of the drawing]

[Fig. 1] An explanatory diagram showing a data conversion circuit according to a first embodiment of the present invention.

[Figure 2] Example of a circuit using the encoded ROM in Figure 1

[Fig. 3] An explanatory diagram showing a data conversion circuit according to a second embodiment of the present invention.

[Figure 4] Explanatory diagram showing a conventional data conversion circuit

[Figure 5
] Example of a circuit using the encoded ROM shown in Figure 4

[Explanation of symbols]

11A-B Encoding ROM 12A-B Decryption ROM 5 Shift circuit

Claims (7)

[Claims] Claim 1: A device having an input address port and an output data port, inputting fixed length data as input data to the input address port, encoding it, and outputting at least one of the coded data and the code length of the coded data from the output data port. Output the code data as output data, or sequentially input the code data string in which the code data is continuous into the input address port as input data and decode the fixed length data before encoding and the code data string to be decoded. In the data conversion circuit, the input side address is at least two of the data input address and the control signal input address. 1. A data conversion circuit comprising addresses having two different properties and switching output data depending on a control signal value input to a control input address. 2. When the control signal is a predetermined value, the data input to the input address is converted and output, and when the control signal is not the predetermined value, the input data is output as output data. The data conversion circuit according to claim 1, characterized in that: 3. A plurality of conversion tables for input data and output data are provided, one of the plurality of conversion tables is specified by the control signal, and the input data is converted into output data based on the conversion table. 2. The data conversion circuit according to claim 1, wherein the data conversion circuit is configured as follows. 4. The control signal includes image size, image density range, number of bits per pixel of image, type of image,
Image shooting conditions, statistical properties of the entire image or local areas,
Any one of the conditions regarding the image compression rate such as quantization width,
4. The data conversion circuit according to claim 3, wherein the conversion table is specified by the generation condition. 5. One or more of the conversion tables is a non-conversion table, and when the non-conversion table is designated, input data is output as is as output data. The data conversion circuit according to item 3. 6. The data conversion circuit according to claim 1, wherein the data conversion circuit is applied to a data compression/expansion device. 7. The data conversion circuit according to claim 1, wherein the data conversion circuit is applied to data conversion processing of medical images.