JP2001128176A - Image compressing and expanding device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To record and reproduce images at an arbitrary set compression factor. SOLUTION: In an image compressing device 10, a DCT computing section 11, an area segmenting and filtering section 12, a quantizing section 13, and a Haffman coding section 14 are provided. The DCT computing section 11 converts image data into DCT coefficients through DCT computation. The area segmenting and filtering section 12 compares the absolute values of the DCT coefficients with a threshold m related to a compression factor in a zigzag scanning direction in the block of the DCT coefficients. When the absolute values continuously become smaller than the threshold m, the section 12 uses one of the DCT coefficients as a boundary DCT coefficient and changes the values of the DCT coefficient preceding the boundary DCT coefficient in the zigzag scanning direction to zero. The quantizing section 13 converts the DCT coefficients into quantized DCT coefficients by quantizing the DCT coefficients. The Haffman coding section 14 generates compressed image data by encoding the quantized DCT coefficients into Haffman codes.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image compression and decompression apparatus which generates compressed image data by compressing digital image data and decompresses the original image data by decompressing the compressed image data. .

[0002]

2. Description of the Related Art Conventionally, JPEG has been used as a standard algorithm for encoding digital color still images.
(Joint Photographic Expert Group) method is standardized. According to the JPEG method, first, image data is converted into DCT transform coefficients by a DCT operation, the DCT transform coefficients are quantized, and compressed image data is generated by entropy coding such as Huffman coding. When decompressing the compressed image data, an operation opposite to the compression process is performed. That is, the entropy decoding process, the inverse quantization, and the inverse DCT transform are performed on the compressed image data, whereby the original image data is restored.

An image compression and decompression device for performing such image compression and image decompression processing is provided, for example, in a digital camera. A subject image captured by the digital camera is compressed as digital image data, and It is recorded on a recording medium such as a disk. When the compressed image data is restored by the decompression process, the subject image is reproduced and displayed on the image display device provided in the digital camera.

[0004] The compression ratio at the time of performing the compression process depends on the quantization table. That is, by changing the value of the quantization coefficient constituting the quantization table,
The compression ratio changes. Conventionally, a plurality of quantization tables are prepared in advance, and when a user selects a compression ratio by using a select button or the like, that is, a quantization table, image data is compressed at a compression ratio according to the quantization table.

[0005]

As described above, in an image compression apparatus provided in a conventional digital camera or the like, several predetermined compression ratios (quantization tables) are used.
, The digital image data is compressed. Therefore, conversely, the user cannot arbitrarily set the compression ratio and cannot compress the digital image data at the set compression ratio.

SUMMARY OF THE INVENTION An object of the present invention is to provide an image compression and decompression device capable of recording and reproducing an image at various arbitrary compression ratios.

[0007]

SUMMARY OF THE INVENTION An image compression apparatus according to the present invention is an image compression apparatus for performing compression processing for digital image data composed of a plurality of blocks of pixel data block by block. An orthogonal transform means for transforming data into orthogonal transform coefficients by orthogonal transform, and a block of orthogonal transform coefficients, which are formed by orthogonal transform coefficients along a zigzag scan direction, which have a value larger than a threshold corresponding to a compression ratio. Area dividing means for dividing the effective area and an invalid area composed of orthogonal transform coefficients having a value smaller than the threshold value, and, in a block of orthogonal transform coefficients, the orthogonal Filter processing means for setting a value of a certain orthogonal transform coefficient to 0, and quantizing means for converting the filtered orthogonal transform coefficient into a quantized orthogonal transform coefficient. The quantized orthogonal transform coefficients, characterized by comprising an entropy coding means for converting the compressed image data by entropy coding. When the threshold value is arbitrarily changed by such a compression processing device, image compression processing with various compression ratios is performed accordingly.

It is desirable that the image compression apparatus further has a recording unit for recording the compressed image data. By having such a recording means, image decompression processing becomes possible. .

The area dividing means sequentially determines along the zigzag scanning direction whether or not the absolute value of the orthogonal transform coefficient is less than or equal to a threshold value for a predetermined number of times in a block of the orthogonal transform coefficients, and continuously determines the absolute value of the orthogonal transform coefficient for a predetermined number of times. If it is less than or equal to the threshold value, one of the continuous orthogonal transform coefficients is defined as a boundary orthogonal transform coefficient, and an area composed of orthogonal transform coefficients ahead of the boundary orthogonal transform coefficient in the zigzag scan direction is defined as an invalid area. It is desirable that For example, the region dividing means determines whether or not the absolute value of the orthogonal transform coefficient is equal to or smaller than the threshold three times in succession, and among the three consecutive orthogonal transform coefficients that are equal to or smaller than the threshold, along the zigzag scan direction. The first orthogonal transform coefficient that is equal to or smaller than the threshold is defined as a boundary orthogonal transform coefficient. By such a region division process, an orthogonal transform coefficient having a value of 0 is generated according to a threshold value corresponding to the compression ratio, and compression processing is performed at a predetermined compression ratio.

It is preferable that the entropy coding means performs Huffman coding on the quantized orthogonal transform coefficients.

It is desirable that the image compression apparatus further has a position storage means for storing the position of the boundary orthogonal transform coefficient in the block of the orthogonal transform coefficient in a memory. In this case, the Huffman encoding means, based on the position of the boundary orthogonal transform coefficient stored in the position storage means, for the quantized orthogonal transform coefficient ahead of the position of the boundary orthogonal transform coefficient along the zigzag scan direction, It is desirable to output codes for ZRL (zero run length) and the remaining run length. Thereby, the time required for the compression process is reduced.

It is desirable that the threshold is determined according to the set compression ratio. Thereby, image data can be compressed at various compression rates.

The image data includes image data of a luminance signal,
It is preferable that the image data be composed of color difference signal image data. In this case, the image compression device has one quantization table for the image data of each signal.

The orthogonal transform means converts the image data into a DC
It is desirable to perform a T (Discrete Cosine Transformation) operation.

The image decompression device according to the present invention comprises: means for reading out compressed image data recorded in the recording means; and entropy decoding for restoring quantized orthogonal transform coefficients by performing entropy decoding on the compressed image data. Means,
Inverse quantization means for restoring orthogonal transform coefficients by performing inverse quantization on the restored quantized orthogonal transform coefficients using a quantization table, and inverse orthogonal transform for the restored orthogonal transform coefficients. And an inverse orthogonal transform for restoring image data. With such an image decompression device, the compressed image data is restored.

Preferably, the entropy decoding means performs Huffman decoding on the compressed image data. Further, the inverse orthogonal transform means applies an IDCT (Inverse Discrete Cosine Transformatio) to the restored orthogonal transform coefficient.
n) It is desirable to perform calculations.

An image compression apparatus according to the present invention includes an orthogonal transform unit for transforming digital image data composed of a plurality of blocks of pixel data into orthogonal transform coefficients by orthogonal transform, and for quantizing the orthogonal transform coefficients. And a quantization table inversion means for inverting a quantization table used for generating an inverse quantization table, and performing quantization along a zigzag scan direction on a block of orthogonal transform coefficients, thereby performing quantization orthogonal transform. A quantization means for generating a coefficient, wherein the orthogonal transform coefficient from the first position to a predetermined position along the zigzag scanning direction is quantized using an inverse quantization table, and the quantization is performed in the zigzag scanning direction. A quantization means for performing quantization using a quantization table for an orthogonal transform coefficient located ahead of a predetermined position along the Coefficients, characterized by comprising an entropy coding means for converting the compressed image data by entropy coding. With such an image compression device, even if a quantization table other than the quantization table for the non-compression mode is used, it is possible to substantially perform the lossless encoding process.

It is desirable that the quantization table inverting means inverts the quantization table represented by the matrix with the diagonal of the matrix as an axis, thereby generating an inverted quantization table satisfying the following equation. Q ′ _ji = Q _7−i7−j (1) where Q represents a quantization coefficient which is an element of a quantization table, and Q ′ is an inverse quantum which is an element of an inverse quantization table. Represents the conversion coefficient. The subscripts j and i represent the vertical and horizontal positions of the quantization table and the inverse quantization table, respectively. By such an inversion process, the value of the quantization coefficient corresponding to the orthogonal transform coefficient in the low-frequency component region can be converted to a small value.

In the block of orthogonal transform coefficients, the quantizing means is arranged so that the orthogonal transform coefficient at a position advanced by half the number of orthogonal transform coefficients constituting the block from the orthogonal transform coefficient at the first position in the zigzag scan direction. Up to the conversion factor,
It is desirable to quantize the orthogonal transform coefficients using an inverse quantization table. As a result, the orthogonal transform coefficients are hardly quantized.

[0020]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, an image compression and decompression device according to an embodiment of the present invention will be described with reference to the drawings. In the present embodiment, the image compression device and the image decompression device are provided in a digital camera.

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

The light reflected by the subject S is transmitted through a lens L
The subject image is formed on the light receiving surface of the image sensor 17 by capturing through Red (R), green (G), and blue (B) color filters (not shown) are provided on the light receiving surface of the image sensor 17, and an image signal corresponding to each color is generated by photoelectric conversion. I do. R,
The image signals corresponding to G and B are converted into digital signals in the signal processing circuit 18, and further converted into luminance data Y and color difference data Cb and Cr. The luminance data Y and the color difference data Cb, Cr are sent to the storage memory 19.

The storage memory 19 is provided with a memory area for each of the luminance data Y, the color difference data Cb, and Cr.
The luminance data Y and the color difference data Cb and Cr for the screen are
It is temporarily stored in each area of the storage memory 19.

The luminance data Y and the color difference data Cb, Cr
Are read from the storage memory 19,
It is input to the image compression device 10. The image compression device 10
It comprises a DCT operation unit 11, an area division / filter processing unit 12, a quantization unit 13, a Huffman encoding unit 14, a CPU 60, and a recording medium M. The CPU 60 controls the compressed image device 1
0 overall control.

In the image compression device 10, the image data of the luminance data Y and the color difference data Cb and Cr are
The image data is divided into pixel blocks each consisting of 8 × 8 = 64 pixel data, and image data for one screen is compressed for each block. Further, the luminance data Y and the color difference data Cb, Cr are separately compressed.

The DCT operation section 11 applies DCT (Discrete Cos) to the luminance data Y and the color difference data Cb and Cr.
ine Transformation) operation, that is, a discrete cosine transform is performed. The DCT operation is one of orthogonal transforms, and the luminance data Y and the color difference data Cb and Cr are decomposed for each spatial frequency component and are converted into DCT coefficients.

In the area dividing / filtering section 12, the luminance data Y and the color difference data Cb, Cr
Are subjected to region division processing and filter processing. Thus, with the boundary DCT coefficient as a boundary,
DCT coefficient values that are less than or equal to the threshold are converted to zero. The position of the boundary DCT coefficient is recorded in the position memory IM.

In the quantization unit 13, the filtered D
The CT coefficients are quantized, and quantized DCT coefficients are obtained. This quantization is a linear quantization, and 8 × 8
= Quantization table Q composed of 64 quantization coefficients
Is quantized using. As for the quantization table Q, a quantization table for luminance data Y and a quantization table for color difference data Cb and Cr (both not shown) are provided.

The luminance data Y and the color difference data Cb, Cr
Are subjected to Huffman coding in the Huffman coding unit 14 to generate compressed image data. At this time, the Huffman table, the group table, and the position information stored in the position memory IM sent from the CPU 60 are used. The generated compressed image data is recorded on the recording medium M. Note that the quantization table Q, Huffman table, and group table are stored in advance in CP
It is stored in a memory (not shown) in U60.

A compression process for one block (hereinafter referred to as a pixel block) of image data will be described with reference to FIGS. Here, the compression process is performed on the pixel block of the luminance data Y, but the same compression process is performed on the pixel block of the chrominance data Cb and Cr.

FIG. 2 shows an 8 × 8 pixel block P and DC
T coefficient block D and filter DCT coefficient block KD
FIG. Referring to FIG.
1. The processing in the region division / filter processing unit 12 (see FIG. 1) will be described. The elements of each block are
Pixel data P _yx , DCT coefficient D _ji , filter D
It represented a CT coefficient KD _ji.

In the pixel block P, a subscript y indicates a vertical position, and is 0, 1, 2,... 7 from the top. The subscript x indicates the position in the horizontal direction, and from the left, 0, 1,.
2 ... 7. For example, when y = 1 and x = 1, the pixel value P = 162. In the DCT coefficient D and the filter DCT coefficient KD, the subscript j indicates the vertical position as in the case of the subscript y, and the subscript i indicates the position in the horizontal direction as in the case of the subscript x.

First, each pixel data P of the pixel block P_yx
Is the DCT coefficient D_jiIs converted to
At position (0,0) in DCT coefficient block D
DCT coefficient D₀₀(= 260) is a DC (direct current) component
And the DCT coefficient D at the remaining position_jiIs AC (alternating current)
Minutes. The DC component is the pixel data in the pixel block P.
Data P_yxCorresponds to the average value of On the other hand, the AC component is
DCT coefficient D₀₁Or D _TenTo the DCT coefficient D₇₇In the direction of
The higher the spatial frequency component value is in the pixel block
Represents how many are in P. Thus, pixel block
Is decomposed for each spatial frequency component by DCT operation
Is done.

Next, the DCT coefficient block D, compared with a threshold value m and the DCT coefficients D _{ji are} applied sequentially along the zigzag scan direction indicated by an arrow in FIG. The threshold value m is a value corresponding to the compression rate in the image compression processing. By changing the threshold value m, the compression rate changes.
Here, the threshold value m = 10. Where the threshold m and DC
In comparison with the T coefficients D _ji, the absolute value of the DCT coefficients D _ji | D _ji | and threshold value m is compared.

The threshold values m and D along the zigzag scan direction
While sequentially comparing the CT coefficient _Dji with the CT coefficient _Dji , when the absolute value of the DCT coefficient _Dji is continuously smaller than the threshold value m,
One of the DCT coefficients _Dji is set as a boundary DCT coefficient _DT, and its position is detected. In the figure, the DCT coefficient D ₀₃ (=
5) becomes the boundary DCT coefficient _DT . Boundary DC detected
The position of the T coefficient _DT is stored in the position memory IM (see FIG. 1).

Once the boundary DCT coefficient D _T is determined, the DC
In the T coefficient block D, a filtering process is performed.
That is, the DCT coefficient D along the zigzag scan direction
₀₀ (= 260) DCT coefficients D _ji (hereinafter, DCT of coefficients D _ji in the effective area) in the range surrounded by a broken line from to the boundary DCT coefficients D _T (= 5) while the intact value of the boundary DCT coefficients D _T (= 5) along a zigzag scan direction ahead of the DCT coefficients D _ji (hereinafter, referred to as DCT coefficients D _ji in the invalid area) in the range enclosed by two-dot chain line from the position of all 0 Is converted to the value of

FIG. 3 is a diagram showing a DCT coefficient block D. Here, the region division processing will be described in detail.

As shown in FIG. 2, from the position of the DCT coefficient D ₀₀ (= 260) of the DC component, along the zigzag scanning direction, the absolute value of the DCT coefficient _Dji and the threshold value m (= 10)
Are sequentially compared. If the absolute value of the DCT coefficient _Dji is larger than the threshold value m (= 10), the DCT coefficient D at the next position along the zigzag scanning direction is compared with the threshold value m.

In the case of the DCT coefficient block D shown in FIG.
Threshold m and DCT coefficient D along zigzag scan direction_ji
Are compared, the DCT coefficient D₂₀(= 0)
Is initially less than or equal to the threshold value m (= 10). However
From the DCT coefficient D at the next position ₁₁(= 36) is the threshold m
(= 10), which is continuously less than or equal to the threshold value m
Therefore, the threshold value m and the DCT coefficient D_ji
Is compared with the absolute value of

Subsequently, the threshold value m and the DCT coefficient D_jiThe absolute value of
By comparison, this time the DCT coefficient D ₁₃(= 5) is the threshold m
(= 10) or less. In addition, zigzag scanning direction
DCT coefficient D at the next position along₁₂(= -2)
DCT coefficient D at the next position of_{twenty one}= Absolute value of (−8)
Is also equal to or less than the threshold value m (= 10). In the present embodiment, 3
DCT coefficient D_jiIs the threshold value m (= 10)
DCT in the first position of the three if
Coefficient D_ji(Here, the DCT coefficient D₀₃= 5) to the boundary D
CT coefficient D_TAnd Then, in the filtering process,
Boundary DCT coefficient D _TIn earlier position, ie invalid
DCT coefficient D in the region_ji(DCT coefficient D ₁₂, DCT
Number D_{twenty one}...) Are all converted to a value of 0.

As described above, the DCT coefficient block D
Is generated by decomposing a pixel block P for each spatial frequency component, and is a DCT that is a DC component.
Region of DCT coefficients D _ji in the coefficient D ₀₀ (= 260) side is a region of low frequency components, whereas, DCT coefficients D ₇₇
The region of the DCT coefficient _Dji on the side is the region of the high-frequency component. Normally, the value of the DCT coefficient _{Dji in} the low-frequency component area is larger than the value of the DCT coefficient in the high-frequency component area, and the value of the DCT coefficient _Dji gradually decreases along the zigzag scan direction. To go. From this, when the DCT coefficient _Dji becomes the threshold value m or less for three consecutive times, the probability that the DCT coefficient _Dji located further ahead in the zigzag scanning direction becomes the threshold value m or less is very high.

As described above, in the DCT coefficient block D, the threshold value m and the absolute value of the DCT coefficient _Dji are compared along the zigzag scanning direction, and the absolute value of the DCT coefficient _Dji is 3
When the number of times becomes equal to or less than the threshold value m successively, the DCT coefficient _DT at the first position among the three is set as the boundary DCT coefficient _DT .

FIG. 4 shows a filter DCT coefficient block KD
, A quantization table Q, and a quantized DCT coefficient block E
FIG. The quantization in the quantization unit 13 (see FIG. 1) will be described with reference to FIG. Note that the elements of each block are the filter DCT coefficients K
D _ji , quantization coefficient Q _ji , and quantization DCT coefficient E _ji .

The filter DCT coefficient block KD is quantized using a quantization coefficient table Q. That is, each filter DCT coefficient KD _ji is divided by the corresponding quantization coefficient Q _ji , and the remainder is rounded.
For example, the filter DCT coefficient KD ₀₁ (= 49) is divided by the quantization coefficient Q ₀₁ (= 11), and the remainder is rounded. Thereby, the quantized DCT coefficient E ₀₁ (= 4) is obtained.

Filter DCT coefficient KD having a value of 0
_ji has a value of 0 irrespective of the value of the quantization coefficient _Qji.
Is the quantized DCT coefficient E _ji . In the present embodiment, only one quantization table Q is stored in the memory in the CPU 60.

FIG. 5 is a diagram showing the Huffman encoding process in the Huffman encoding unit 14 (see FIG. 1).

First, a zigzag scan is performed on the quantized DCT coefficient block E, and 64 quantized DCT coefficients are obtained.
The coefficients _Eji are rearranged into one-dimensional columns. In the quantized DCT coefficients E _ji arranged in one column, Huffman coding is performed on a quantized DCT coefficient E ₀₀ (= 16), which is a DC component, using a DC component group table (not shown). Is applied. Then, the quantized DCT, which is the other AC component,
The coefficient E _{ji is} subjected to Huffman coding using a group table for AC components and a Huffman table (not shown).

In the Huffman encoding process for the quantized DCT coefficient E _ji of the AC component, the quantized DCT coefficient
If there is a coefficient _Eji, it is determined as an invalid coefficient, and a length of zero (zero run) is counted along the zigzag scan direction. On the other hand, for a quantized DCT coefficient (effective coefficient) _Eji that is not 0, the corresponding group number and additional bit are obtained by referring to the group table.

The code word is obtained by referring to the Huffman table based on the obtained group number and the counted zero run. Then, code data is generated by combining the code word and the previously obtained additional bits. When code data is sequentially generated for each quantized DCT coefficient _Eji , one block of compressed image data is generated. Note that such Huffman encoding processing is conventionally known.

In the Huffman encoding process, when the run length of the invalid coefficient, that is, the length of 0s along the zigzag scan direction exceeds 15, "ZRL" is continuously output as code data until it becomes 15 or less. . by the way,
The position of the quantized DCT coefficient E ₀₃ (= 1) is the same position as the boundary DCT coefficient D _T, and this position is stored in the position memory IM (see FIG. 1) in the area division processing. All quantized DCT coefficients E _ji at the tip position along the zigzag scan direction from the position of the quantized DCT coefficients E ₁₂ 0
Since it is, in the present embodiment, with respect to the quantized DCT coefficients E _ji located beyond the position of the quantized DCT coefficients E ₀₃ along a zigzag scan direction, without counting in each case 0, the ZRL Output. That is, the quantized DCT coefficient E _ji (= from the position of the next quantized DCT coefficient E _ji to the last position (quantized DCT coefficient E ₇₇ ) following the position stored in the position memory IM along the zigzag scan direction.
ZRL is output based on the number of 0).

For example, in the case of FIG. 5, the quantized DCT coefficient E
₀₃Earlier quantized DCT coefficients E _jiThe number of (= 0) is
Quantized DCT coefficient E₀₀From the quantized DCT coefficient E₀₃For up to
Quantized DCT coefficient E_jiIs 7 so that 64-
7 = 57. For the last 0, code data and
Output “EOB (End Of Block)”
In addition, ZRL is output every time 16 consecutive 0s are output.
From here, three ZRLs (16 × 3) are output
And the sign bits for the remaining eight zero runs
Is generated.

As described above, the pixel blocks P constituting the original image data are compressed according to the threshold value m (= 10).

FIG. 6 is a block diagram of the image decompression device according to the present embodiment.

The image decompression device 20 includes the Huffman decoding unit 2
1. Inverse quantization unit 22, IDCT operation unit 23. CPU6
0, a recording medium M. The compressed image data recorded on the recording medium M is read by the CPU 60 and sent to the Huffman decoding 21.

The Huffman decoding unit 21 performs Huffman decoding on the compressed image data. The Huffman decoding is the reverse operation of the Huffman coding, and the quantized DCT coefficients of the luminance data Y and the color difference data Cb and Cr are restored from the compressed image data. The quantized DCT coefficients are sent to the inverse quantization unit 22.

The inverse quantization unit 22 performs inverse quantization on the quantized DCT coefficient. As a result, the DCT coefficients are restored. The restored DCT coefficient is sent to the IDCT operation unit 23. The IDCT operation unit 23 calculates an IDCT (Inverse Discrete Cosine Transf
ormation) operation, ie, inverse discrete cosine transform. The IDCT operation is the reverse operation of the DCT operation,
Luminance data Y and color difference data C by IDCT operation
The image data of b and Cr is restored.

The restored image data of the luminance data Y and the color difference data Cb and Cr is sent to the image memory 19. The image data is stored in an image display device (not shown).
The recorded image is reproduced and displayed.

The image expansion process for one block will be described with reference to FIG.

FIG. 7 shows a quantized DC signal based on Huffman decoding.
FIG. 5 is a diagram showing a T coefficient block E (see FIG. 4), a DCT coefficient block D ′, and a pixel block P ′. Quantized DC
The T coefficient block E is reversibly restored by the Huffman decoding of the compressed image data and the formation of a matrix which is the inverse of zigzag scanning. Note that the elements of each block, quantized DCT coefficients respectively E _ji, DCT coefficient block D _'ji, the pixel data P' is expressed as _ji.

The quantized DCT coefficient block E is
, And is inversely quantized by the signal Q. That is, each quantum
DCT coefficient E_ji, The quantizer at the corresponding position
Number Q _jiIs multiplied. Quantization during the compression process
Irreversible compression / decompression (encoding) processing
You. Therefore, the DCT coefficient block generated by the inverse quantization
Lock D 'is the same as DCT coefficient block D shown in FIG.
I will not do it.

For example, when the quantized DCT coefficient E ₀₁ (= 4) is multiplied by the quantized coefficient Q ₀₁ (= 11), D
The CT coefficient D ′ ₀₁ (= 44) is obtained.

Then, the DC restored by the inverse quantization
An IDCT operation is performed on the T coefficient block D ′.
As a result, the pixel block P ′ is restored. By restoring all the pixel blocks P ', original image data is restored. As described above, the original image data is restored from the compressed image data by the expansion processing by the image expansion device 20.

FIG. 8 is a diagram showing an image compression process for one block when the value of the threshold value m is changed. Here, it is assumed that the threshold value m = 4 instead of the threshold value m = 10.

When the threshold value m = 4, the DCT coefficient block D
, The DCT coefficient D
₄₀ (= −2) is the boundary DCT coefficient D _T. This boundary D
The position of the CT coefficient _DT is different from the position when the threshold value m = 10. Then, the DCT coefficients _Dji located ahead of the zigzag scan direction are all converted to zero values by the filter processing. The position of the boundary DCT coefficient _DT is stored in the position memory IM, and in the Huffman encoding process, ZRL is output based on the position.

By changing the value of the threshold value m, the position of the boundary DCT coefficient D _T changes. Therefore, the number of DCT coefficients _Dji converted to 0 changes according to the value of the threshold value m, and the length of the zero run in the Huffman coding process changes, so that the compression ratio changes. That is,
The value of the threshold value m corresponds to the compression ratio. In the present embodiment, when the user sets the compression ratio by dial operation or the like, the value of the threshold value m is determined according to the set compression ratio. When the compression ratio is high, the value of the threshold value m increases, and when the compression ratio is low, the value of the threshold value m decreases. The value of the threshold m is
It is calculated in the CPU 60.

The non-compression processing, that is, the lossless encoding processing will be described with reference to FIGS. In the present embodiment, when the user selects the non-compression mode, that is, when image data is recorded without compression so that the original image data can be completely restored, the region division processing and the filter processing are not performed, and the quantization is not performed. By performing the inversion process on the table Q, the image data is recorded without compression.

FIG. 9 is a diagram showing the inversion processing of the quantization table.

The value of each quantization coefficient Q _{ji in} the quantization table Q corresponds to the value of each DCT coefficient D _ji in the DCT coefficient block D, and the DCT coefficient D _{ji in} the low frequency component region
The value of the quantization coefficient _Qji corresponding to _ji is large, and conversely, the quantization coefficient _Qji for the DCT coefficient _Dji in the high-frequency component region is large.
The value of _ji is small. Normally, quantization is performed on the DCT coefficient _Dji along the zigzag scan direction.
In the present embodiment, the DCT coefficient _Dji is quantized by selectively using both the inverted quantization table Q ′ obtained by inverting the quantization table Q and the quantization table Q.

As shown in FIG. 9, X-
The quantization table Q, which is a matrix, is inverted with respect to the X 'axis to generate an inverted quantization table Q'. At this time, 'inverted quantization coefficient Q _J'i is each element _of' the quantization coefficients Q _ji inverted quantization table Q satisfies the following relationship.
Here, the suffixes j and i indicate the vertical and horizontal positions in the inverse quantization table Q ′, respectively.

Q ′ _ji = Q _7-i7-j (1)

For example, when the quantization coefficient Q ₃₂ (= 4) is inverted, it becomes an inverted quantization coefficient Q ′ ₅₄ (= 4). In addition,
The generation of the inverse quantization table Q ′ is performed by the CPU 6
Performed by 0.

FIG. 10 is a diagram showing a DCT coefficient block D, a quantization table Q, and an inverse quantization table Q ′.

As described above, the DCT coefficient block D
In the zigzag scan direction, the DCT coefficient D
_jiAre sequentially quantized. Uncompressed mode is set
First, in the DCT coefficient block D, the zigzag
DCT coefficient D along the can direction₀₀(= 260) to 3
DCT coefficient D at second position₄₃Area up to (= 0)
Quantized DCT coefficient D in Q'R_jiIs the amount of reversal
The quantization is performed using the quantization table Q '. That is, 6
Four DCT coefficients D_jiDCT coefficient block D
Inverted quantization table up to half (32)
The quantization is performed using Q '. And zigzagski
33rd DCT coefficient D along the direction ₃₄(= 1)
DCT coefficient D in the destination region QR_jiFor the quantum
The quantization is performed using the quantization table Q.

As shown in FIG. 10, the quantization DCT in the area QY used for quantization in the quantization table Q
The value of the coefficient Q _ji, the value of _ji 'quantization factor Q in the region QX utilized for quantization in' the inversion quantization table Q is are both smaller. Therefore, the value of each quantized DCT coefficient _Dji of the DCT coefficient block D does not change so much even by the quantization, and the image data is recorded on the recording medium M substantially without compression.

As described above, according to the present embodiment, in the DCT coefficient block D, the value of the predetermined threshold value m is compared with the absolute value of the DCT coefficient _Dji along the zigzag scanning direction. Then, three consecutive DCT coefficients D
If the absolute value of _ji is equal to or less than the threshold value m, the first threshold value m the following become DCT coefficients D _ji and boundary DCT coefficients D _T, DCT coefficients D at the tip of the boundary DCT coefficients D _T along a zigzag scan direction Convert _ji to all zero values. The value of the threshold value m is determined according to an arbitrary set compression ratio. By changing the threshold value m, image compression processing at an arbitrary compression ratio can be performed using only one quantization table Q. . At this time, image compression processing at various compression rates can be performed using only one quantization table Q without preparing a plurality of quantization tables corresponding to the compression rates as in the related art. Then, the compressed image data generated by the compression processing is restored by the image decompression device 20.

In the DCT coefficient block D, if the DCT coefficient _Dji becomes equal to or less than the threshold value m consecutively three times in the zigzag scanning direction, the position of the boundary DCT coefficient _DT is recorded in the position memory IM. Then, in the Huffman encoding process, to the boundary DCT coefficients D quantized DCT coefficients E _ji (= 0) at the tip along the zigzag scan direction than the quantized DCT coefficients E _ji corresponding to the position of _T, 0 , And ZRL is output. Thereby, the time required for the compression process is reduced.

When the non-compression mode is selected, quantization is performed using the quantization table Q and the inverse quantization table Q ′. By performing such quantization, one quantization table Q prepared in advance without using a non-compression quantization table (all quantization coefficient values are 1) is used.
Thus, a substantially reversible encoding process can be performed.

In this embodiment, the DCT operation is applied as the orthogonal transform. However, the present invention is not limited to this, and another orthogonal transform such as a Hadamard transform may be applied instead. In this case, in the image expansion processing, an inverse Hadamard transform or the like is applied instead of the DCT operation. Further, instead of Huffman coding, another entropy coding such as arithmetic coding may be applied as entropy coding.
In this case, in the image decompression process, arithmetic decoding or the like is applied instead of Huffman decoding.

The quantization table Q applied to the present embodiment
However, the present invention is not limited to this, and another quantization table may be applied. At this time, the value of the threshold value m is determined according to the prepared quantization table and the set compression ratio. In the quantization process in the non-compression mode, 64 DCs are used.
It is not necessary to perform quantization so that half of the T coefficients D _ji are used using the inverse quantization table and the other half are used, and each quantization coefficient Q _ji of the prepared quantization table is used. based on the value may be determined the position of the DCT coefficients D _ji when switching to a quantization table from the reverse quantization table.

Furthermore, in order to enable a wider range of compression ratios to be set, a plurality of thresholds m may be determined for each quantization table while selectively using a plurality of quantization tables. This makes it possible to continuously and finely set the compression ratio.

In the present embodiment, it is determined whether or not the absolute value of the DCT coefficient _Dji becomes equal to or less than the threshold value m for three consecutive times. However, the present invention is not limited to three times, and four or five consecutive times. It may be determined whether or not the value is equal to or less than the threshold value m. Also, among the DCT coefficients D _ji which three consecutive equal to or less than the threshold value m, the first DCT coefficients D _ji became less than the threshold value m instead of the boundary DCT coefficients D _T, 2 second, or third DCT The coefficient _Dji may be the boundary DCT coefficient _DT .

In this embodiment, the image compression device and the image decompression device are provided in the digital camera, but may be provided in the computer instead.

[0083]

As described above, according to the present invention, an image can be recorded and reproduced at various arbitrary compression ratios.

[Brief description of the drawings]

FIG. 1 is a block diagram of an image compression device according to an embodiment.

FIG. 2 is a diagram illustrating a process of a compression process on one pixel block.

FIG. 3 is a diagram showing a DCT coefficient block.

FIG. 4 is a diagram showing a process of quantization.

FIG. 5 is a diagram showing a Huffman encoding process.

FIG. 6 is a block diagram of an image decompression device according to the present embodiment.

FIG. 7 is a diagram showing a decompression process for restoring one pixel block.

FIG. 8 is a diagram showing a process of a compression process when a threshold value is changed.

FIG. 9 is a diagram showing a quantization table and an inverse quantization table.

FIG. 10 is a diagram illustrating quantization in an uncompressed mode.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 10 Image compression apparatus 11 DCT operation part (orthogonal transformation means) 12 Region division | segmentation / filter processing part (region division | segmentation means, filter processing means) 13 Quantization part (quantization means) 14 Huffman encoding part (entropy encoding means) 20 Image decompression device 21 Huffman decoding unit (entropy decoding unit) 22 Inverse quantization unit (inverse quantization unit) 23 IDCT operation unit (inverse orthogonal transformation unit) D _T boundary DCT coefficient (boundary orthogonal transformation coefficient) IM position memory M Recording medium m Threshold value P Pixel block _Pyx pixel data Q Quantization table Q 'Inversion quantization table

Claims (17)

[Claims] 1. An image compression apparatus for compressing digital image data composed of a plurality of blocks of pixel data on a block-by-block basis, wherein said image data is converted into orthogonal transform coefficients by orthogonal transform. An orthogonal transform means, an orthogonal transform coefficient block, along the zigzag scan direction, an effective area composed of the orthogonal transform coefficient having a value greater than a threshold value corresponding to a compression ratio, and an effective area smaller than the threshold value. Area dividing means for dividing into an invalid area composed of the orthogonal transform coefficients as values, and in the block of the orthogonal transform coefficients, the orthogonal transform coefficients in the effective area are left as they are while the orthogonal transform coefficients are in the invalid area. Filter processing means for setting the value of the orthogonal transform coefficient to 0, and an amount for converting the filtered orthogonal transform coefficient into a quantized orthogonal transform coefficient Means a, the quantized orthogonal transform coefficients, the image compression apparatus characterized by comprising an entropy coding means for converting the compressed image data by entropy coding. 2. The apparatus according to claim 1, further comprising a recording unit for recording the compressed image data. 3. The region dividing means sequentially determines along a zigzag scanning direction whether or not an absolute value of the orthogonal transform coefficient is less than or equal to the threshold value continuously for a predetermined number of times in the block of the orthogonal transform coefficient. When the number of consecutive orthogonal transform coefficients is equal to or smaller than the threshold value, one of the consecutive orthogonal transform coefficients is regarded as a boundary orthogonal transform coefficient, and the orthogonal transform is located ahead of the boundary orthogonal transform coefficient along the zigzag scan direction. 2. The image compression apparatus according to claim 1, wherein an area constituted by coefficients is set as the invalid area. 4. The image compression apparatus according to claim 3, wherein the area dividing means determines whether or not the absolute value of the orthogonal transform coefficient is equal to or less than the threshold value three times in a row. 5. The method according to claim 1, wherein, among the three consecutive orthogonal transform coefficients that are equal to or smaller than the threshold value, the area dividing unit first assigns the orthogonal transform coefficient that is equal to or smaller than the threshold value along the zigzag scan direction to the boundary orthogonal. The image compression apparatus according to claim 3, wherein the image compression apparatus uses a conversion coefficient. 6. The image compression apparatus according to claim 1, wherein the entropy encoding unit performs a Huffman encoding process on the quantized orthogonal transform coefficients. 7. The image compression apparatus according to claim 6, further comprising a position storage unit that stores a position of the boundary orthogonal transform coefficient in the block of the orthogonal transform coefficient in a memory. 8. The method according to claim 1, wherein the Huffman encoding unit is configured to determine, based on a position of the boundary orthogonal transform coefficient stored in the position storage unit, the quantum located ahead of the position of the boundary orthogonal transform coefficient along a zigzag scan direction. 8. The image compression apparatus according to claim 7, wherein a code corresponding to ZRL (zero run length) and the remaining run length is output for the generalized orthogonal transform coefficients. 9. An image compression apparatus according to claim 1, wherein said threshold value is determined according to a set compression ratio. 10. The apparatus according to claim 1, wherein said image data comprises image data of a luminance signal and image data of a chrominance signal, and each image data of each signal has one quantization table. An image compression apparatus according to claim 1. 11. The image compression apparatus according to claim 1, wherein said orthogonal transformation means performs a DCT (Discrete Cosine Transformation) operation on said image data. 12. A means for reading out the compressed image data recorded in the recording means according to claim 2, and entropy decoding of the compressed image data to restore the quantized orthogonal transform coefficients. An entropy decoding unit that performs inverse quantization on the restored quantized orthogonal transform coefficient using the quantization table to restore an orthogonal transform coefficient; An image decompression device comprising: an inverse orthogonal transform for restoring image data by performing an inverse orthogonal transform on an orthogonal transform coefficient. 13. An apparatus according to claim 12, wherein said entropy decoding means performs Huffman decoding on said compressed image data. 14. An inverse orthogonal transform unit according to claim 1, wherein said inverse orthogonal transform means applies IDCT (Inverse Discrete Cos
13. The image decompression device according to claim 12, wherein an ine transform is performed. 15. An orthogonal transformation means for transforming digital image data composed of blocks of a plurality of pixel data into orthogonal transformation coefficients by orthogonal transformation, and a quantization used for quantizing the orthogonal transformation coefficients. A quantization table inverting unit that inverts the table to generate an inverted quantization table; and generates a quantized orthogonal transform coefficient by performing quantization along the zigzag scan direction on the block of the orthogonal transform coefficient. The quantization means, for the orthogonal transform coefficient from the first position to a predetermined position along the zigzag scanning direction, performs quantization using the inverse quantization table, and then, in the zigzag scanning direction Along the orthogonal transform coefficient that is ahead of the predetermined position, the quantization is performed using the quantization table. Stage and, the quantized orthogonal transform coefficients, the image compression apparatus characterized by comprising an entropy coding means for converting the compressed image data by entropy coding. 16. The quantization table inverting means generates the inverted quantization table satisfying the following equation by inverting the quantization table represented by a matrix with the diagonal of the matrix as an axis. The image compression apparatus according to claim 15, wherein: Q ′ _ji = Q _7−i7−j Here, Q represents a quantization coefficient which is an element of the quantization table, and Q ′ represents an inverse quantization coefficient which is an element of the inverse quantization table. The subscripts j and i represent the vertical and horizontal positions of the quantization table and the inverse quantization table, respectively. 17. The quantizing means advances the orthogonal transform coefficient block by half the number of orthogonal transform coefficients constituting the block from the orthogonal transform coefficient at the first position along the zigzag scan direction. 16. The image compression apparatus according to claim 15, wherein the orthogonal transform coefficient is quantized to the orthogonal transform coefficient at an odd position by using the inverse quantization table.