JP2935320B2 - Image compression / decompression device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image compression / expansion apparatus for encoding and compressing and decompressing an amount of data, and more particularly to an image compression / expansion apparatus using an interlaced scanning camera and display.

[0002]

2. Description of the Related Art Since an image signal has a very large amount of data, the amount of data is often recorded or transmitted after compressing the amount of data. This compression of the amount of data is called encoding. Decompressing data to restore an image is called decoding. As such an encoding / decoding method, a method using a discrete cosine transform (hereinafter referred to as DCT) is known. For example, “Overview of International Standard for Color Still Image Coding (JPEG)” by Takao Omachi and Fumitaka Ono ( 1) ”(1991, Journal of the Institute of Image Electronics Engineers of Japan, Vol. 20, No. 1,
No.).

On the other hand, as an image pickup apparatus and a display, those using interlaced scanning, such as a television camera and a receiver, are widely used. As an example of a method of compressing and expanding an image using such an imaging device, a display, an encoder, and a decoder, there is a method shown in a block diagram of FIG. This is a television camera (hereinafter T
This is a method of taking one image with a V camera, compressing the image, storing it on a recording medium such as a magnetic disk, extracting it when necessary, expanding it, and displaying it on a display.

In FIG. 1, an NTSC TV camera 1 is shown.
Are output in the form of a composite signal, and the TV camera 1 sequentially outputs the captured single image into two fields by interlaced scanning. The screen shown in FIG. 3A is 525 in the NTSC system.
The number of scanning lines present is schematically represented by 16 lines. In the interlaced scanning, these scanning lines are output twice every other line from the top to form two fields. The scanning lines with diagonal lines are the first field, and the remaining scanning lines are the first field. Configure two fields.

[0005] The output of the TV camera 1 is sampled and quantized by an analog-to-digital converter (hereinafter referred to as an A / D converter) 2 and converted into a digital signal. The digitized composite signal is converted into an R signal, a G signal, and a B signal (hereinafter collectively referred to as RGB signals) by the RGB synthesizing unit 3. next,
The first sent RGB signal of the first field is stored in the first field memory 4 and the RGB signal of the next second field is further stored.
The signal is written into the second field memory 5. Therefore, the contents written in the first field memory 4 and the second field memory 5 become RGB signals obtained by converting composite signals taken out every other scanning line as shown in FIG. 3B.

[0006] The encoder 6 compresses the RGB signal of one field written in the first field memory 4 and compresses the RGB signal in the first field.
Field compressed data is generated, and one field RGB signal written in the second field memory 5 is further compressed to extract the second field compressed data, and the decoder 8 converts the field RGB signal into one field RGB signal. The data is expanded and written in the first field memory 9. Next, the compressed data of the second field is extracted, and the RG of one field is similarly extracted.
The signal is decompressed into a B signal and is written into the second field memory 10. The data is read out from the RGB signals in the first field memory 9 and converted into analog signals by a digital-to-analog converter (hereinafter referred to as a D / A converter) 11.
Output to 2.

Subsequently, the RGB of the second field memory 10
The signals are sequentially read out, converted to analog signals by the D / A converter 11 in the same manner, and output to the display 12. Also the second
When the reading of the field memory 10 is completed, the process returns to the reading of the first field memory 9 again, and the reading of the first field memory 9 and the second field memory 10 alternately is repeated. Thus, the expanded RGB signals of the first field and the second field are repeatedly output to the display 12.

[0008] The display 12 is configured to display the screen shown in FIG.
Signals are displayed every other line from the first scanning line, and RGB signals of the next second field are displayed every other line from the second scanning line. The image captured by the TV camera 1 in this manner is divided into two fields by interlaced scanning and output, and is separately compressed in the order of the first field and the second field and recorded on the recording medium 7.

At the time of display, the compressed data of the first field and the compressed data of the second field are decompressed in this order and written into the first field memory 9 and the second field memory 10, respectively, and these are alternately read. You.
The display 12 displays these while performing interlaced scanning. In the NTSC system, both the first and second fields are read out about 30 times per second, so that they are viewed as one image.

Next, the configuration of the encoder 6 will be described. It comprises a block reading section 20, a DG section 21, a quantization section 22, and a variable length coding section 23. As shown in FIG. 4A, the block reading unit 20 divides a block obtained by dividing the original image by 8 pixels in the horizontal direction × 8 pixels in the vertical direction from the upper left, one block in the horizontal direction from the upper left block of the original image.
The DCT unit 21 reads out the data sequentially from the line to the lower right block.
Output to

Specifically, first, the R signal written in the first field memory 4 is regarded as the original image shown in FIG. 4A, and the addresses of the R signals included in the block to be read are generated one after another. Each block of signals is read out sequentially. After the reading of the R signal of the first field memory 4 is completed, the G signal of the first field memory 4 is similarly read and the DCT unit 2
Output to 1.

Then, the R signal, the G signal, and the B signal of the second field memory 5 are sequentially read for each block in the same manner as in the first field, and output to the DCT unit 21. Here, the number of pixels in the horizontal direction and the vertical direction of the original image are each assumed to be a multiple of 8, but if the number of pixels in the horizontal direction of the original image is not a multiple of 8, the rightmost block needs the rightmost pixels. It can be configured by repeatedly reading out only the pixels that do not exist and substituting them. Similarly, when the number of pixels in the vertical direction of the original image is not a multiple of 8, the lowermost block can be configured by repeatedly reading the lowermost pixels for the missing pixels and substituting them.

The DCT unit 21 performs a two-dimensional discrete cosine transform on each of the input blocks at eight points in the horizontal direction and eight points in the vertical direction. The block input below is a two-dimensional integer array f
These processes are specifically described as (i, j) (i and j are coordinates in the horizontal direction and the vertical direction, respectively, and take integer values from 0 to 7 as shown in FIG. 4B). .

F (i, j) is, for example, -128 for the signal level of the darkest pixel and 1 for the signal level of the brightest pixel.
Take an 8-bit integer value of 27. f (i, j) is converted into a DCT coefficient F (u, v) according to the following equation (1). Here, u and v represent the horizontal and vertical frequencies in the two-dimensional discrete cosine transform, respectively, and take integer values from 0 to 7 similarly to i and j. Also, sq of the equation (1)
rt (2) is the square root of 2. F (u, v) is a two-dimensional integer array. If f (i, j) is an 8-bit integer, it takes 3 bits and takes an 11-bit integer from -1024 to 1023.

[0015]

The DCT has the property of collecting image information into low frequency components, that is, frequencies at which u or v is small. Generally, the absolute value of F (u, v) is such that the low frequency components are large and the high frequency components are small.

In the next quantization unit 22, this F (u, v)
Is quantized as shown in the following equation (2) using a quantization matrix Q (u, v) taking a generated numerical value recorded in the table 30 and converted into a two-dimensional integer array C (u, v). This table 30 is composed of, for example, a read-only memory.

[0018]

Here, the division is an integer division, and the fractional part is discarded. By setting the value of the quantization matrix small for low frequency components and large for high frequency components as shown in the following equation (3), C (u,
The high frequency component of v) can be almost zero.

[0020]

[0021] Since these high frequencies are difficult to discern to the human eye, information can be compressed without any noticeable deterioration of the decoded image. The next variable length coding unit 23 codes C (u, v) using a variable length code such as a Huffman code. The variable-length code is a code that compresses the entire information amount by assigning a code word with a short bit length to a symbol with a high appearance frequency and assigning a code word with a long bit length to a symbol with a low appearance frequency. In the coding of C (u, v), C (u, v) is zigzag-scanned from a low frequency as shown below and rearranged into one-dimensional data.

C (0,0), C (1,0), C (0,
1), C (0,2), C (1,1), C (2,0), C
(3,0), C (2,1), C (1,2), C (0,
3), C (0, 4), C (1, 3), C (2, 2), C
(3,1), C (4,0), ..., C (7,6), C
(6,7), C (7,7).

Next, the one-dimensional data is examined from C (0,0), and the number of continuous 0, that is, the number of runs of 0, and the following non-zero coefficient are C (0,0), C (0,0). , 1), C
In the case of only (1, 3), C (u, v) is represented as follows.

(0, C (0, 0)), (1, C (0,
1)), (4, C (2, 1)), (3C (1, 3)),
EOB.

Then, the set of (0 run number, non-zero coefficient) and the end-of-block EOB are each encoded as a symbol with a variable length code. The correspondence between the symbol and the code word is determined by a correspondence table, and the symbol can be faithfully restored from the compressed data. In general, it can be expected that the greater the number of C (u, v) in 0, the greater the number of 0 runs. Therefore, the number of sets of (0 run number, non-zero coefficient) decreases, and it can be expected that the compressed data of the block decreases.

The encoder 6 compresses each block in this way. Therefore, first, the R signal of the first field, G
Signal, and then compresses the B signal in this order, generates compressed data of the first field from these signals, and then generates the R signal of the second field,
The G signal and the B signal are compressed in this order, and compressed data of the second field is generated from these. As described above, T
Compressed data of an image captured by the V camera 1 is generated. The layer bit number of the compressed data obtained in this way is considerably smaller than the total bit number of the original image, and a typical compression ratio is 1/1.
About 0 to 1/50 is obtained.

Next, the configuration of the decoder 8 will be described. This includes the variable length decoding unit 24, the inverse quantum unit 25, and the inverse DCT unit 2
6, a block writing unit 27. Here is TV
It is assumed that the number of pixels in the horizontal and vertical directions of the image captured by the camera 1 is known at the time of decoding. In addition, for simplicity of explanation, these are all multiples of eight. That is, the number of horizontal blocks and the number of vertical blocks of the first and second fields are known at the time of decoding. It is also assumed that the symbol-codeword correspondence table used in the variable-length encoding unit 23 is known at the time of decoding. However, even if these are not fixed, it is sufficient to provide a header portion at the beginning of the compressed data, put this in the header portion, and read it before decoding.

First, the compressed data of the first field recorded on the recording medium 7 is extracted and decoded by the variable length decoding unit 24, the inverse quantization unit 25 and the inverse DCT unit 26 for each block. The compressed data of the first field is an R signal, a G signal, and a B signal.
Since the signals are arranged in the order of the signals, the R signal is decoded first. Then, in the block writing unit 27, the position of this block in the first field is obtained from the number of decoded blocks and the number of blocks in the horizontal direction of the first field, and the address of the R signal contained in this block is continuously obtained. And writes them into the first field memory 9.

When decoding of the block at the bottom right of the R signal is completed, the address of the G signal included in the block is similarly generated from the next block as a G signal, and these are written in the first field memory 9. When the decoding of the block of the G signal at the lower right is completed, the address of the B signal included in the block is similarly generated from the next block as the B signal,
Write to the first field memory 9. Thus, the decompression of the compressed data of the first field is performed.

Subsequently, the compressed data of the second field is taken out, decompressed in the order of the R signal, the G signal, and the B signal and written in the second field memory 10 as in the first field. Here, the number of pixels in both the horizontal and vertical directions of the original image is assumed to be a multiple of 8, but even when the number of pixels in the horizontal direction is not a multiple of 8, as described above, the rightmost block is encoded by the encoder 6. If you put the rightmost pixel for the missing pixel,
By not writing these in the first field memory 9 and the second field memory 10, one field can be expanded. Similarly, even when the number of pixels in the vertical direction of the original image is not a multiple of 8, the supplemented pixels in the lowermost block need not be written.

The decoding of one block is performed by the reverse process of the encoding as follows. The variable length coding unit 24 converts the codeword included in the compressed data into a corresponding symbol by referring to a symbol-codeword correspondence table. As a result, a set of (0 run number, non-zero coefficient) or EOB is restored. These are substituted into C (u, v) from (0,0) in the zigzag scan order described above. That is, C
The value 0 is substituted for the number of 0 runs detected from (0, 0), and the value of the detected non-zero coefficient is substituted for the next C (u, v). The same is repeated from the next C (u, v) again using the next restored (0 run number, non-zero coefficient). If the converted symbol is EOB, 0 is substituted for the remaining C (u, v). After substituting C (7,7) in this way, the coefficient C (u, v) for one block is there.
Are aligned.

Next, the inverse quantization unit 25 uses the quantization matrix Q (u, v) stored in the table 30 to convert the DCT coefficient F 'from one block of C (u, v) according to the equation (4). Calculate (u, v).

F ′ (u, v) = C (u, v) Q (u, v) (4) F (u, v) and F ′ (u, v) calculated by the DCT unit 21
And the difference between the absolute values is 1 / Q (u, v).
2 or less. Next, the inverse DCT unit 26 divides F ′ (u, v) into 8 points in the horizontal direction × 8 points in the vertical direction according to the following equation (5).
Performs a dimensional inverse discrete cosine transform. Where c (u), c
(V) is the same constant as equation (1).

[0034]

The two-dimensional inverse discrete cosine transform is the inverse of the two-dimensional discrete cosine transform, and if F ′ (u, v) and F ′
If (u, v) are equal, f '(i, j) is equal to block f (i, j) of the original image. Actually, as described above, since the absolute value of F ′ (u, v) and F (u, v) has a difference equal to or smaller than half of the quantum or the matrix Q (u, v), the decoded f ′ (I, j) is the block f (i, j) of the original image
And slightly different. However, the quantization matrix Q (u, v) can be selected so that the deterioration caused by this encoding / decoding method is not so noticeable to human eyes. The decoder 8 performs decoding for each block in this manner. In this conventional example, an image picked up by the TV camera 1 is compressed and recorded on a recording medium 7, extracted and expanded when necessary, and displayed on a display 12.

[0036]

The above-mentioned TV camera 1
When compressing and decompressing the image captured by the above, in order to reduce the total number of bits of the compressed data, it is necessary to increase the values of the elements of the quantization matrix stored in the table 30, and as a result, the image quality of the decompressed image Deteriorates. Conversely, in order to suppress the deterioration of the image quality, it is necessary to reduce the values of the elements of the quantization matrix, and as a result, the total number of bits of the compressed data increases.

An object of the present invention is to take advantage of the fact that an image is divided into two fields by interlaced scanning and separately encoded, thereby suppressing the deterioration of an image seen by human eyes while suppressing the totality of compressed data. An object of the present invention is to provide an image compression / decompression device capable of reducing the number of bits.

[0038]

According to the structure of the present invention, a captured image is interlaced and scanned by N fields (N is 2).
A camera that divides and outputs the N fields, and an encoder that compresses the N fields one by one to generate compressed data, and decompresses the compressed data to generate N decoded fields. In an image compression / decompression device that displays a decoded image of the captured image by a decoder and a display that interlace scans and displays the N decoded fields, the encoder encodes the one field. Each block is divided into two-dimensional blocks, and these blocks are sequentially subjected to two-dimensional discrete cosine transform to obtain transform coefficients. The transform coefficients are quantized using a threshold determined for each coefficient position, and the quantized transform coefficients are used to obtain the transform coefficients. The compressed data of the block is generated, and these are collected to generate the compressed data of the one field. The decoder is configured to generate the compressed data of the one field from the compressed data of the one field. The quantized transform coefficient of the block is taken out, multiplied by the threshold value, and then subjected to two-dimensional inverse discrete cosine transform to generate a decoded block. These are collected to generate the single decoded field. The compression ratio is increased by changing at least one of the threshold values equally in the compression and decompression of the second and subsequent fields.

[0039]

FIG. 1 is a block diagram for explaining a first embodiment of an image compression / decompression apparatus according to the present invention. In this case, an image captured by the TV camera 1 is divided into two fields and compressed separately as in FIG. The difference between this embodiment and the conventional example is that
The compression of the RGB signal of the first field is performed using the quantization matrix stored in the table 28, and the R signal of the second field is compressed.
By using the quantization matrix stored in the table 29 for compressing the GB signal, the operation is otherwise the same as that of the conventional example, and the description is omitted.

Here, the table 28 has, for example, the same quantization matrix Q (u, v) (u,
v is an integer from 0 to 7), and the table 29 stores 2Q (u, v) obtained by doubling each Q (u, v).
v) is stored. Therefore, RGB of the first field
The signal is compressed using the same quantization matrix as in the prior art. That is, coarse quantization is performed in the second field than in the first field. By the way, since the RGB signal of the first field and the RBG signal of the second field are signals of adjacent scanning lines, when the first field and the second field are considered as different images, these two images are generally quite similar.

Therefore, if the first field and the second field are encoded by the same quantization matrix as in the conventional example,
The total bits of the compressed data of the first field and the compressed data of the second field are not so different. However, when each block of the second field performs coarser quantization than the first field as in the present embodiment, the quantized DCT coefficient C (u, v) of each block of the second field is used.
Can be expected to contain more zeros than in the first field. Therefore, the compressed data of the second field is more suitable for the first field.
It is expected that the total number of bits will be smaller than the compressed data of the field, that is, the total number of total bits of both fields is expected to be smaller than in the conventional example.

Further, since the second field performs coarse quantization, the image quality of the expanded second field is inferior to the image quality of the expanded first field. However, the display 12 displays the first field and the second field by interlaced scanning as shown in FIG. That is, since each scanning line of the second field is sandwiched between the scanning lines of the first field with good image quality, the image displayed on the display appears to have much deteriorated image quality as compared with the conventional example. Absent.

Here, an example is given in which the elements of the quantization matrix stored in the table 29 are twice as large as the elements of the quantization matrix stored in the table 28. Assuming that the quantization matrix shown in (3) is Q (u, v), the quantization matrix stored in the table 28 is Q (u, v) when the horizontal frequency u is 3 or less, and 2Q when the horizontal frequency u is 4 or more. (U, v), the quantization matrix stored in the table 29 is expressed by Q (u, v) when the vertical frequency v is 3 or less, and 2Q when the frequency is 4 or more.
(U, v).

In this case, as compared with the conventional example, the RGB signal of the first field is roughly quantized in the horizontal direction, and the RGB signal of the second field is roughly quantized in the vertical direction.
Therefore, compared to the conventional example, the first field has a deteriorated high frequency in the horizontal direction, and the second field has a deteriorated high frequency in the vertical direction. However, since these two fields are interlaced and displayed on the display, deterioration of the expanded image is less noticeable than in the conventional example. Further, since quantization is coarser in both fields than in the conventional example, the total number of bits of the compressed data can be expected to be smaller than in the conventional example.

FIG. 2 is a block diagram of another embodiment of the present invention. This is to separate the composite signal output from the TV camera 1 into a luminance signal (hereinafter referred to as Y signal) and two color difference signals (hereinafter referred to as Cb signal and Cr signal), and to compress these Y signal and Cb signal Cr signal. . It is known that the resolution for the Cb signal and the Cr signal of the human eye is lower than the resolution for the Y signal.
Also in the system, the band of the Cb signal and the band of the Cr signal are narrower than the band of the Y signal. Therefore, the quantization matrix used in the encoder 6 is also divided into a Cb signal and a Cr signal
The signal, the Cr signal, and the Cr signal perform coarser encoding than the Y signal. In this embodiment, a quantization matrix used for compression and decompression of a Y signal is replaced by a first field and a second field.

An image picked up by the TV camera 1 is output in the form of a composite signal. This is sampled and quantized by the A / D converter 2, and then separated by a luminance signal / chrominance signal separation unit (hereinafter referred to as Y / C) 13 into a Y signal, a Cb signal, and a Cr signal. Then, the Y signal, Cb signal, C
The r signal is written to the first field memory 14, and the Y signal, Cb signal, and Cr signal of the second field are written to the second field memory 15. Then, the Y signal for one field written in the second field memory 14 by the encoder 6,
The Cb signal and the Cr signal are compressed in order in
Generate compressed data for the field.

Here, the quantization matrix stored in the table 31 is used for quantizing the DCT coefficient of the Y signal block, and the quantization matrix stored in the table 32 is used for quantizing the DCT coefficient of the Cb signal block and the Cr signal block. Using a conversion matrix. The quantization matrix stored in the table 31 is, for example, described in the example of the conventional table 30 (3).
The quantization matrix shown in the equation is used. As the quantization matrix stored in the table 32, for example, a quantization matrix Q ′ (u, v) of the following equation (6) is used.

[0048]

Further, the Y signal, Cb signal, and Cr signal for one field written in the second field memory 15 are similarly compressed in order to generate compressed data of the second field. Here, the quantization matrix stored in the table 31 is used to quantize the DCT coefficient of the Y signal block, and Cb
The quantization matrix stored in the table 33 is used for quantizing the DCT coefficients of the signal block and the Cr signal block. In the table 32, each element is Q '(u, v) of (6).
Is stored.

The compressed data of the first field and the second
The compressed data of the field is sequentially recorded on the recording medium 7.
When displaying, first, the compressed data of the first field is taken out, decompressed by a decoder 8 into a Y signal, a Cb signal, and a Cr signal of one field.
Write in 6. Here, the quantization matrix stored in the table 31 is used for the quantization of the DCT coefficient of the Y signal block, and the quantization matrix stored in the table 32 is used for the quantization of the DCT coefficient of the Cb signal block and the Cr signal block. Is used.

Next, the compressed data of the second field is extracted, and the Y signal, Cb signal, Cr
The signal is decompressed and written into the second field memory 17. Here, the quantization matrix stored in the table 31 is used to quantize the DCT coefficient of the Y signal block, and Cb
The quantization matrix stored in the table 33 is used for quantizing the DCT coefficients of the signal block and the Cr signal block.

Next, the Y signal of the first field memory 16
The signal is read from the Cb signal and the Cr signal.
The signal is converted into a signal, a G signal, and a B signal, and is converted into an analog signal by the D / A converter 11 and output to the display 12. Subsequently, similarly, the Y signal, the Cb signal, and the Cr signal of the second field memory 17 are sequentially read out, and
After being converted into an R signal, a G signal, and a B signal in 8, they are converted into analog signals by a D / A converter 11 and output to a display 12.

When the reading of the second field memory 17 is completed, the process returns to the reading of the first field memory 16 again.
Reading the field memory 16 and the second field memory 17 alternately is repeated. Thus, the display 12
Is the expanded RG of the first field and the second field.
The B signal is repeatedly output. Since these are displayed while performing interlaced scanning, they appear as one image to human eyes.

In this embodiment, the Cb and Cr signals of the second field are quantized more coarsely than the Cb and Cr signals of the first field.
The signal b is deteriorated as compared with the expanded first field. However, the Cb signal occupying the compressed data in the first field,
It can be expected that the total number of bits of the compressed data of the Cb signal and the Cr signal in the compressed data of the second field is smaller than the compressed data of the r signal. Therefore, by using the same quantization matrix stored in the table 32 for the first field, the second field Cb signal, and the Cr signal as in the related art, it can be expected that the total number of bits of the compressed data is reduced. Further, since the display 12 displays these by interlaced scanning, it does not appear that the display 12 is much deteriorated as compared with the conventional example.

[0055]

As described above, the image compression / decompression apparatus of the present invention compresses and decompresses the first field and the second field using different quantization matrices, so that the image is expanded and decompressed as compared with the conventional example. There is an effect that the total number of bits of the compressed data can be reduced while suppressing the deterioration of the image quality.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating an embodiment of an image compression / decompression method according to the present invention.

FIG. 2 is a block diagram illustrating a second embodiment of the present invention.

FIG. 3 is a block diagram illustrating a conventional image compression / expansion method.

FIG. 4 is a display diagram for explaining the interlaced scanning in FIG. 3;

FIG. 5 is a schematic diagram illustrating an original image and blocks.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 TV camera 2 A / D converter 3 RGB synthesis part 4 1st field memory 5 2nd field memory 6 Encoder 7 Recording medium 8 Decoder 9 1st field memory 10 2nd field memory 11 D / A converter Reference Signs List 12 display 13 Y / C separation unit 14 first field memory 15 second field memory 16 first field memory 17 second field memory 18 RGB synthesis unit 20 block reading unit 21 DCT unit 22 quantization unit 23 variable length encoding unit 24 Variable length decoding unit 25 Inverse quantization unit 26 Inverse DCT unit 27 Block writing unit 28-33 Table

Claims (1)

(57) [Claims] 1. A camera for interlaced scanning a captured image and outputting it in N fields (N is an integer of 2 or more) and compressing the N fields one by one to generate compressed data A decoder that expands the compressed data to generate N decoded fields, and a display that interlace-scans and displays the N decoded fields and displays the decoded image of the captured image. In the image compression / decompression device that displays
The one field is divided into two-dimensional blocks, and these blocks are sequentially subjected to two-dimensional discrete cosine transform to obtain transform coefficients, which are quantized by a threshold determined for each coefficient position. Compressed data of the block is generated by using a transform coefficient, and these are collected to generate compressed data of the one field. The decoder determines the quantum of each block from the compressed data of the one field. Transform coefficients, multiplying the transformed coefficients by the threshold value, and performing a two-dimensional inverse discrete cosine transform to generate a decoded block, and collecting these to generate the single decoded field; and At least one of the thresholds
An image compression / expansion apparatus wherein the compression ratio is increased by equally changing compression and decompression of the second and subsequent fields.