JP2000013797A - Image compression device and image decompression device - Google Patents

Abstract

(57) [Summary] [PROBLEMS] To efficiently compress a high-resolution image and compress and decompress the image so as not to cause image deterioration. A discrete cosine transform (DCT operation) unit, a quantization unit, and a bit separation unit are provided in an image compression apparatus.
7. A Huffman encoding unit 18 is provided. Image compression device 10
The DCT operation unit 15 performs a DCT operation on the image data input into, and the quantization unit 16 quantizes the image data, thereby converting the image data into quantized DCT coefficients. The quantized DCT coefficient is separated into a higher-order quantized DCT coefficient and a lower-order quantized DCT coefficient by a bit separation unit 17, and the Huffman coding unit 18 performs Huffman coding on the higher-order quantized DCT coefficient.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to an image compression apparatus and an image expansion apparatus for compressing and expanding a color still image.

[0002]

2. Description of the Related Art To efficiently transfer an image through a communication circuit by compressing a color still image, JP
An image compression method called an EG (Joint Photographic Expert Group) method has been standardized. In the JPEG-based baseline process, image data is first decomposed into spatial frequency components by discrete cosine transform.
The data represented for each spatial frequency component is quantized, and the quantized data is encoded (compressed). The encoded image data is expanded by decoding, inverse quantization, and inverse discrete cosine transform, thereby restoring the original image.

[0003]

According to the JPEG method, a high-resolution image can be efficiently compressed. However, since some distortion is added in the process of compression and decompression, the compressed image is compressed. Deteriorates. Therefore JP
A lossless encoding method in which distortion is not added to an image has been proposed as the EG method, but this method cannot efficiently compress a high-resolution image.

[0004] The present invention provides an image compression apparatus and an image decompression device for compressing and decompressing an image so as not to add distortion and cause image degradation while adopting the JPEG system for efficiently compressing a high-resolution image. The aim is to get the device.

[0005]

An image compression apparatus according to the present invention comprises: orthogonal transform means for orthogonally transforming image data corresponding to an input still image to obtain an orthogonal transform coefficient represented by a bit string; A bit separation unit that separates the bit string of the coefficient into two bits, an upper bit string and a lower bit string, and obtains an upper orthogonal transform coefficient represented by an upper bit string and a lower orthogonal transform coefficient represented by a lower bit string; Encoding means for obtaining compressed image data by encoding.

It is desirable that the bit separation means separates the quantized orthogonal transform coefficients into bits.

The bit separation means performs a right shift operation on the orthogonal transform coefficient represented by the bit sequence, a lower bit sequence for the bit sequence separated by the right shift operation, and an upper bit sequence for the bit sequence of the orthogonal transform coefficient converted by the right shift operation. Thus, it is desirable to obtain the upper orthogonal transform coefficient and the lower orthogonal transform coefficient.

An image decompression device according to the present invention comprises decoding means for obtaining upper orthogonal transform coefficients by decoding input compressed image data, upper orthogonal transform coefficients to be inputted, and upper orthogonal transform coefficients obtained by the decoding means. It is characterized by comprising bit synthesizing means for obtaining orthogonal transform coefficients by bit synthesizing transform coefficients, and inverse orthogonal transform means for obtaining image data by performing orthogonal transform on the orthogonal transform coefficients.

The inverse orthogonal transform means desirably performs inverse orthogonal transform on the inversely quantized orthogonal transform coefficient.

It is desirable that the bit synthesizing unit obtains the orthogonal transform coefficient by performing a left shift operation on the upper orthogonal transform coefficient represented by the bit string and applying the lower bit string to a bit portion which becomes blank by the left shift operation.

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, an image compression apparatus and an image compression apparatus according to embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram of an image compression apparatus according to an embodiment of the present invention.

A subject image 11 (still image) is
Is imaged on the light receiving surface of the solid-state imaging device 13 via A photoelectric conversion element is provided on a light receiving surface of the solid-state imaging device 13,
A color filter including red (R), green (G), and blue (B) color filter elements is provided on the upper surface of the photoelectric conversion element. Each photoelectric conversion element corresponds to one pixel data, and a subject image is converted into an electric image signal corresponding to a predetermined color by each photoelectric conversion element. The image signal is converted from an analog signal to a digital signal in an A / D converter (not shown).

The digitized image signal is converted into luminance data Y and chrominance data C in a signal processing circuit (not shown).
The data is converted into b and Cr and input to the image memory 14. The image memory 14 stores luminance data Y, color difference data Cb, C
r are stored in independent areas, and each memory area has a storage capacity for one image. The luminance data Y and the color difference data Cb, Cr are input data to the image compression device 10.

The luminance data Y and the color difference data Cb, Cr
Is divided into a plurality of blocks in one screen, and is processed in block units. Each block is 8x
It consists of eight pixel data.

The luminance data Y and the color difference data Cb and Cr read from the image memory 14 are
Is input to The image compression device 10 includes a discrete cosine transform (DCT operation) unit 15, a quantization unit 16, and a bit separation unit 1.
7, and a Huffman encoding unit 18. The luminance data Y and the color difference data Cb, Cr are
0 is separately compressed.

The input luminance data Y and color difference data Cb and Cr are subjected to a DCT operation in a DCT operation section 15. The DCT operation is one of orthogonal transforms, and image data is decomposed for each spatial frequency component and converted into DCT coefficients.

The luminance data Y and the color difference data Cb, Cr
Are quantized by the quantization unit 16 and converted into quantized DCT coefficients. This quantization is linear quantization, and is performed using a quantization table Q including 8 × 8 = 64 quantization coefficients. That is, DC
T coefficients are divided by the corresponding quantization coefficients,
The remainder is rounded. As the quantization tables, a quantization table Q _y for luminance data Y and a quantization table Q _c for color difference data Cb and Cr are provided.

The luminance data Y and the color difference data Cb, Cr
Is represented by a binary bit string in the image compression apparatus 10, and the quantized DCT coefficients of the luminance data Y and the color difference data Cb and Cr are also represented by a binary bit string. The bit string of the quantized DCT coefficient of the luminance data Y and the color difference data Cb, Cr is separated into an upper bit string and a lower bit string by the bit separation unit 17, and is represented by the upper quantized DCT coefficient and the lower bit string represented by the upper bit string. A lower quantized DCT coefficient is obtained. However, here, even one bit is represented as a bit string.

The luminance data Y and the color difference data Cb, Cr
Are encoded (compressed) in the encoding unit 18 and recorded in the compressed image data recording area M2 of the recording medium M as compressed image data. For the encoding, a Huffman encoding method based on the JPEG algorithm is applied. The lower quantized DCT coefficient is recorded in the lower recording area M3 of the recording medium M without being encoded. The quantization table Q used in the quantization unit 16 together with the recording of the upper quantization DCT coefficient and the lower quantization DCT coefficient.
_y and Q _c are the quantization table recording area M1 of the recording medium M
Will be recorded.

A compression process for one pixel block represented by an 8 × 8 matrix will be described with reference to FIGS.

FIG. 2 shows a pixel block P of luminance data Y.
, A DCT coefficient matrix C, a quantization table Q, and a quantized DCT coefficient matrix D are each exemplified by an 8 × 8 matrix. Elements of each matrix are a pixel value P _yx , a DCT coefficient C _ji , a quantization coefficient Q _ji , and a quantized DCT
It is expressed as a coefficient _Dji .

The pixel value P _yx is the value of the luminance data Y. In the pixel value P _yx , the subscript y indicates the position in the vertical direction, and is 0, 1, 2, ... 7 from the top. The subscript x indicates the position in the horizontal direction, and is 0, 1, 2, ... 7 from the left. For example, when y = 1 and x = 1, the pixel value P ₁₁ = 162.

The DCT coefficients C _ji, quantization table Q _ji, subscript j in the quantized DCT coefficients D _ji, like subscript y indicates the vertical position, 0,1,2 from the top, ... 7
It is. The subscript i indicates the position in the horizontal direction similarly to the subscript x, and is 0, 1, 2, ... 7 from the left. For example, for j = 1, i = 1, the value of the DCT coefficients C ₁₁ 36
It is.

First, each pixel value P _yx of the pixel block P is represented by D
It is converted into a DCT coefficient _Cji by CT operation. DCT
DCT coefficient C at position (0,0) in coefficient matrix C
₀₀ is a DC component, and DCT coefficients C _{ji at the} remaining positions
Is the AC component. The DC component represents the average value of the pixel values P _yx of the pixel block P, and the spatial frequency is 0. AC
The components are from the DCT coefficients C ₀₁ to C ₁₀ to the DCT coefficients C ₇₇
Represents how much higher spatial frequency component values are in the pixel block P in the direction of. Thus, the pixel block P is decomposed for each spatial frequency component by the DCT operation.

Next, each DCT coefficient C _ji is a quantization coefficient Q _ji
Is quantized to a quantized DCT coefficient _Dji . That is, each DCT coefficient C _ji corresponds to the corresponding quantization coefficient Q
_Divide by _ji and round the remainder. To substantially image compression processing unstrained, quantization table Q is not a quantization table corresponding to the high-compression image processing, and is configured by a quantization coefficient Q _ji of as small as possible value.

For example, the DCT coefficient C ₀₃ (= 5) is divided by the quantization coefficient Q ₀₃ (= 2), and the remainder is rounded (rounded) to obtain the quantized DCT coefficient D ₀₃ (=
3) is required.

FIG. 3 shows a quantized DCT coefficient matrix D, an upper quantized DCT coefficient matrix DH, and a lower quantized DCT coefficient matrix DL. Each element is a quantization D
_They are represented as CT coefficient D _ji , upper quantized DCT coefficient DH _ji , and lower quantized DCT coefficient DL _ji .

Each quantized DC represented by a bit string
The T coefficient D _ji is separated into an upper bit string and a lower bit string by bit separation, and an upper quantized DCT coefficient DH _ji represented by an upper bit string and a lower quantized DCT coefficient DL _ji represented by a lower bit string are obtained. Here, the number of bits to be bit-separated is 1 bit.

For example, the bit string of the quantized DCT coefficient D ₁₁ (= 36) is divided into a bit string of the upper quantized DCT coefficient DH ₁₁ (= 18) and a lower quantized DCT coefficient D
It is separated into bit strings of L ₁₁ (= 0). Also quantized DC
The bit string of the T coefficient D ₂₁ (= −7) is separated into a bit string of the upper quantized DCT coefficient DH ₂₁ (= −4) and a bit string of the lower quantized DCT coefficient DL ₂₁ (= 1) by bit separation. However, even one bit is represented as a bit string. The bit separation will be described later.

The upper quantized DCT coefficient _DHji is Huffman-coded using a Huffman table. Then A
Upper quantization DCT coefficient DH _ji by C component and DC component
Are encoded separately, the upper quantized DCT coefficients DH _ji of the AC components are encoded after being converted from a two-dimensional matrix is zigzag scanned in a conventional manner known to one row of data. On the other hand, the lower quantized DCT coefficient _DLji is not Huffman-coded.

When the upper quantized DCT coefficient matrix DH is zigzag scanned and rearranged into one column, the frequency of occurrence of zeros increases as compared to when the quantized DCT coefficient matrix D is zigzag scanned. Therefore, the zero run length (the length of consecutive 0s) becomes longer. Further, since the value of the upper quantized DCT coefficients DH _ji nonzero smaller than the value of the quantized DCT coefficients D _ji, the amount of information required when higher quantized DCT coefficient matrix DH _ji is Huffman coding quantization Less than the DCT coefficient _Dji . Therefore, image data can be efficiently compressed by performing bit separation.

The bit separation of one bit will be described in detail with reference to FIGS. Therefore, the quantized DCT coefficient _Dji is represented by a bit string. Usually 16-bit is assigned to one of the DCT coefficients C _ji, 12 bits of the 16 bits for quantized DCT coefficients D _ji is used. In the case of a quantized DCT coefficient _Dji of an AC component having a positive or negative value, the most significant one bit of the 16 bits is used as a sign bit, and the least significant 11 bits are used.
The quantized DCT coefficients D _ji are represented in binary using bits.

FIG. 4 shows the bit separation of the quantized DCT coefficient D ₁₁ (= −36) and the quantized DCT coefficient D ₂₁ (= −7) which are AC components of the quantized DCT coefficient D _ji of FIG. I have.

First, a quantized DCT coefficient D ₁₁ having a positive value
A case where 1-bit separation is performed for (= 36) will be described.

The quantized DCT coefficient D ₁₁ (= 36) is “0000100100100” in the binary system, and the sign bit F is “0” representing the positive. In the 16-bit bit string B0, the binary number “00000100100” is represented using 11 bits from the right end. The sign bit F of “0” is located at the leftmost most significant bit of the bit string B0. When the quantized DCT coefficient _Dji has a positive value, the unused 4-bit J0 is "0000".

By performing the bit separation, the quantum DCT coefficient D ₁₁ (= 36) is right-shifted. The shift operation is an operation for shifting each bit right or left by a designated number of bits. When a one bit right shift operation is performed, each bit is shifted one by one to the right,
The rightmost least significant bit K which is "0" is separated. The left most significant bit left blank by the right shift operation is filled with a sign bit F of “0”.

As described above, the bit string B0 is converted to the binary number “000001” represented by 10 bits by the right shift operation.
0010 ". This bit string B1 is a high-order bit string, and is a binary number “0000100100
The value obtained by expressing “10” in decimal notation is the higher-order quantized DCT coefficient D.
The value of H _11. That is, the upper quantization DCT coefficient DH
The value of ₁₁ is 18.

On the other hand, the least significant bit K, separated by right shift operation is lower bit string, the value of the least significant bit K is the value of the lower quantized DCT coefficients DL _11. Since the binary number “0” is represented as 0 in the decimal system, the lower quantization DC
The value of T coefficient DL ₁₁ is 0.

Thus, the quantized DCT coefficient D₁₁(= 3
The bit string B0 of 6) is obtained by separating the upper bits
Bit sequence B1 and the least significant bit
And the upper quantized DCT coefficient DH₁₁(= 1
8) and the lower quantized DCT coefficient DL ₁₁(= 0) is required
You.

Next, the quantized DCT coefficient D ₂₁ which is a negative value
A case where 1-bit separation is performed for (= −7) will be described.

The quantized DCT coefficient D ₂₁ (= −7) is represented as “11111111001” in the binary system, and the sign bit F is “1” representing negative. 16-bit bit string M
At 0, the binary number “11111111001” is represented using the 11 bits from the right end. The sign bit F of "1" is located at the leftmost most significant bit of the bit string M0. When the quantized DCT coefficient _Dji is a negative value, the unused 4-bit J1 is "1111".

By performing the bit separation, the quantized DCT coefficient D ₂₁ (= −7) is right-shifted. When the right shift operation is performed, the least significant bit K of “1” is separated.
Then, the sign bit F of “1” is filled in the most significant bit left blank by the right shift operation.

As described above, the bit string M0 is converted to the binary number "111111" represented by 10 bits by the right shift operation.
1100 ". This bit string M1 is a high-order bit string, and the binary number “11111111”
00 ”in decimal notation is the higher quantization DCT coefficient D
The value of H _21. That is, the value of the quantized DCT coefficients DH ₂₁ is -4 to the sign bit F is "1".

On the other hand, the most part separated by the right shift operation
The lower bit K is a lower bit string, and the lower bit K is
The value expressed in decimal notation is the lower quantized DCT coefficient DL_{twenty one}of
Value. Is binary "1" represented as 1 in decimal notation
From the lower quantized DCT coefficient DL _{twenty one}Is 1.

Thus, the quantized DCT coefficient D_{twenty one}(= −
The bit string M0 of 7) is obtained by separating the upper bits by bit separation.
Bit sequence B1 and the least significant bit
And the upper quantized DCT coefficient DH_{twenty one}(= −
4) and lower quantized DCT coefficient DL _{twenty one}(= 1) is required
You.

As described above, the quantized DCT coefficient D _ji of the AC component is _converted to the upper quantized DCT coefficient DH by bit separation.
_ji and the lower quantized DCT coefficient _ji .

The quantized DCT coefficient D ₀₀ of the DC component is also
Bit separation is performed in the same manner as the quantized DCT coefficient _Dji of the component. However, the quantized DCT coefficients D of the DC component ₀₀ for always a positive value, the 16-bit bit string 12
Bits are assigned. Therefore, the sign bit F of “0” indicating positive is filled in the most significant bit which becomes blank when the right shift operation is performed.

In the case of 2-bit or 3-bit separation, a 2-bit or 3-bit right shift operation is performed on the quantized DCT coefficient _Dji .

FIG. 5 is a flowchart showing a procedure for performing n-bit bit separation on an 8 × 8 quantized DCT coefficient matrix. A flowchart of the bit separation will be described with reference to FIGS.

In step 201, the subscript j representing the position in the vertical direction is set to 0. In step 202, the subscript i representing the horizontal position is set to 0, and the leftmost quantized DCT coefficient D _{j0 in the} row representing the horizontal direction of the matrix is _subjected to bit separation. The quantized DCT coefficient _D00 is the subject of the first bit separation.

In step 203, the quantized DCT coefficient D
_An n-bit right shift operation is performed to separate n bits from _ji , and the upper quantized DCT coefficient DH
_ji is required.

For example, in the case of 1-bit separation, a higher-order quantized DCT coefficient DH ₂₁ (= −4) is obtained from the quantized DCT coefficient D ₂₁ (= −7). For example, in the case of 3-bit bit separation, the upper quantized DCT coefficients DH ₂₁ (= -1) is determined from the quantized DCT coefficients D ₂₁ (= -7).

In step 204, the lower quantized DCT coefficient _DLji is obtained. An arithmetic expression for _{obtaining the} lower quantization DCT coefficient _DLji is as shown in Expression (1).

DL _ji = D _ji − (DH _ji × 2 ⁿ ) (1)

For example, in the case of 1-bit separation, the lower quantized DCT coefficient D ₂₁ (= −7) is
The value of the coefficient DL ₂₁ is DL _ji = −7 − (− 4 × 2 ¹ ) = 1 because the value of the upper quantization DCT coefficient DH ₂₁ is −4.

For example, in the case of bit separation of 3 bits, the quantized DCT coefficient D ₂₁ (= −7) is
The value of the coefficient DL ₂₁ is DL _ji = −7 − (− 1 × 2 ³ ) = 1 since the value of the upper quantization DCT coefficient DH ₂₁ is −1.

In step 205, 1 is added to the subscript i. Quantized DCT to be bit-separated by this
The coefficient D _ji is moved to the immediately next right quantized DCT coefficient D _ji .

At step 206, it is determined whether or not the subscript i is 8. That is, it is determined whether or not all the quantized DCT coefficients _Dji have been bit-separated for one row in the horizontal direction. If it is determined that the subscript i is 8, the process proceeds to step 207. If it is determined that the subscript i is not 8, the process returns to step 203.

In step 207, 1 is added to the subscript j. That is, the quantized DCT coefficient D that is bit-separated
_ji is moved to the quantized DCT coefficient D _ji in the next lower row.

At step 208, it is determined whether or not the subscript j is 8. That is, it is determined whether or not all the quantized DCT coefficients _Dji have been bit-separated. If it is determined that the subscript j is 8, the bit separation of the quantized DCT coefficient matrix ends. When it is determined that the subscript j is not 8,
Return to step 202.

As described above, if the image is compressed by applying this embodiment, the zero run length becomes longer and the amount of information necessary for encoding becomes smaller, so that the image can be efficiently compressed. Can be.

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

The luminance data Y and the color difference data Cb, Cr
Is read from the recording area M2 of the recording medium M and sent to the image decompression device 30. Image expansion device 3
0 is a Huffman decoding unit 31, a bit synthesis unit 32, an inverse quantization unit 33, an inverse discrete cosine transform (IDCT operation) unit 3
4. The compressed image data of the luminance data Y and the color difference data Cb and Cr are separately decompressed by the image decompression device 30.

The luminance data Y and the color difference data Cb, Cr
Are decoded by the Huffman decoding unit 31 and are converted into higher-order quantized DCT coefficients. This decoding is the reverse operation of the Huffman coding, and the upper quantized DCT coefficients are represented by a binary bit string.

The luminance data Y and the color difference data Cb, Cr
, And the luminance data Y and chrominance data C read from the lower recording area M3 of the recording medium M.
The lower-order quantized DCT coefficients of b, Cr are
, And the quantized DCT coefficients are restored.

The luminance data Y and the color difference data Cb, Cr
Are inversely quantized by the inverse quantization unit 33 using the quantization table Q, and are converted into DCT coefficients. That is, the quantization table Q _y for the luminance data Y and the quantization table Q _c for the chrominance data Cb and Cr are read from the quantization table recording area M1 of the recording medium M, and correspond to each quantized DCT coefficient. Is multiplied.

The luminance data Y and the color difference data Cb, Cr
The DCT coefficient of the IDCT
It is calculated and converted into luminance data Y and color difference data Cb, Cr. The IDCT operation is the reverse operation of the DCT operation. The luminance data Y and the color difference data Cb, Cr obtained by the IDCT operation are input to the image memory 14.

The decompression process for the compressed image data will be described with reference to FIGS. In the following, the luminance data Y is targeted for the compressed image data.

FIG. 7 shows the upper quantization D as in FIG.
CT coefficient matrix DH and lower quantization DCT coefficient matrix DL
And a quantized DCT coefficient matrix D are exemplified by an 8 × 8 matrix. The elements of each matrix are the upper quantization D
_They are represented as CT coefficient DH _ji , lower quantized DCT coefficient DL _ji , and quantized DCT coefficient D _ji .

The compressed image data is subjected to Huffman decoding separately for each of the AC component and the DC component, and the AC component is converted into a matrix that is the inverse of the zigzag scan, thereby obtaining a higher-order quantized DCT coefficient matrix DH.

The upper quantized DCT coefficient DH _ji and the corresponding lower quantized DCT coefficient DL _ji are bit-combined to restore the quantized DCT coefficient D _ji . Here, the number of bits to be combined is 1 bit.

For example, the upper quantized DCT coefficient DH ₁₁ (= 1
When bit 8) is combined with the lower quantized DCT coefficient DL ₁₁ (= 0), the quantized DCT coefficient D ₁₁ (= 36) is restored. When the upper quantized DCT coefficient DH ₂₁ (= −4) and the lower quantized DCT coefficient DL ₂₁ (= 1) are bit-combined, the quantized CT coefficient D ₂₁ (= −7) is restored. The bit composition will be described later.

FIG. 8 shows a quantized DCT coefficient matrix D, a quantization table Q, a DCT coefficient matrix C ′, and a pixel block P ′.
Are exemplified by an 8 × 8 matrix. Elements of each matrix are respectively represented as a quantized DCT coefficient _Dji , a quantized coefficient _Qji , a DCT coefficient _C'ji , and a pixel value _P'yx .

Quantized DCT coefficient D_jiIs the quantization coefficient Q_ji
By the DCT coefficient C '_jiIs inversely quantized. Sand
That is, each quantized DCT coefficient D_jiQuantization corresponding to
Coefficient Q _jiIs multiplied.

For example, the quantized DCT coefficient D ₂₃ (= −1) is multiplied by the quantized coefficient Q ₂₃ (= 2) to obtain the DCT coefficient C ′ ₂₃ (= −2).

The DCT coefficient C ′ _ji is converted into a pixel value P ′ _yx by an IDCT operation. By executing the IDCT operation, the pixel block P ′ including the 8 × 8 image values P ′ _yx is restored.

The bit combination of one bit will be described in detail with reference to FIGS. Therefore, the upper quantized DCT coefficient DH _ji
And the lower quantized DCT coefficient _DLji are represented by a bit string.

In FIG. 9, the upper quantized DCT coefficient D in FIG.
H ₁₁ (= 18) and the lower quantized DCT coefficient DL ₁₁ (=
0), and the bit combination for the upper quantized DCT coefficient DH ₂₁ (= −4) and the lower quantized DCT coefficient DL ₂₁ (= 1).

First, the upper quantized DCT coefficient D of the AC component
H ₁₁ (= 18) and lower quantized DCT coefficient DL ₁₁ (= 0)
A case in which 1-bit bit synthesis is performed on the

Upper quantized DCT coefficient DH ₁₁ (= 18)
Is represented by a 16-bit bit string B1 as shown in FIG. The lower quantized DCT coefficient DL ₁₁ (= 0) is “0” in binary, and is represented by the lower bit string K1.

By performing the bit synthesis, the upper quantized DC
The T coefficient DH ₁₁ (= 18) is subjected to a left shift operation. When the one-bit left shift operation is performed, each bit is shifted one by one to the left, and the code bit F which is “0” at the left end is separated. The rightmost least significant bit blanked out by the left shift operation has a lower bit string K of “0”.
1 is filled. Further, the bit F0 that is the second “0” from the left end is moved to the leftmost most significant bit by a 1-bit left shift operation, and becomes a sign bit F.

By the left shift operation, the bit string B1 becomes 11
The binary number "0000001001" represented using bits
00 ". Then, a value obtained by expressing the binary number “00000100100” in the decimal system becomes the quantized DCT coefficient D ₁₁ . That is, the value of the quantized DCT coefficients D ₁₁ is 36.

Thus, the upper quantized DCT coefficient DH ₁₁
(= 18) and the lower quantized DCT coefficient DL ₁₁ (= 0) are bit-synthesized, so that the quantized DCT coefficient D ₁₁ (=
36) is restored.

Next, the upper quantized DCT coefficient D of the AC component
H ₂₁ (= −4) and lower quantized DCT coefficient DL ₂₁ (= 1)
A case where 1-bit bit synthesis is performed for the following will be described.

Higher order quantized DCT coefficient DH ₂₁ (= −4)
Is represented by a 16-bit bit string M1 as shown in FIG. The lower quantized DCT coefficient DL ₂₁ (= 1) is “1” in binary, and is represented by the lower bit string K1.

By executing the bit synthesis, the upper quantized DC
T coefficient DH ₂₁ (= -4) is left shift operation. When the one-bit left shift operation is performed, each bit is shifted one by one to the left, and the code bit F that is the leftmost “1” is separated. The rightmost least significant bit blanked out by the left shift operation has a lower bit string K of “1”.
1 is filled. Further, the bit F0 that is the second “1” from the left end is moved to the leftmost most significant bit by a one-bit left shift operation, and becomes a sign bit F.

By the left shift operation, the bit string M1 becomes 11
The binary number "111111110" represented using bits
01 ”is converted to a bit string M0. The value representing the binary number "11111111001" in decimal is the value of the quantized DCT coefficients D _21. That is, the value of the quantized DCT coefficients D ₂₁ in consideration of the sign bit F is "1" is -7.

Thus, the upper quantized DCT coefficient DH ₂₁
(= −4) and the lower quantized DCT coefficient DL ₂₁ (= 1) are bit-combined, so that the quantized DCT coefficient D ₂₁ (=
-7) is restored.

As described above, the upper quantization DCT of the AC component
The coefficient DH _ji and the lower quantized DCT coefficient DL _ji are bit-combined to restore the quantized DCT coefficient D _ji .

The upper quantized DCT coefficient DH ₀₀ and the lower quantized DCT coefficient DL _{00 of the} DC component are bit-synthesized in the same manner. However, the higher quantization DCT coefficient DH of the DC component
_{Since 00} is always a positive value, 12 bits are allocated in a 16-bit bit string.

In the case of 2-bit or 3-bit bit synthesis, 2 bits are _added to the upper quantized DCT coefficient _DHji .
A 3-bit left shift operation is performed.

FIG. 10 is a flowchart showing the procedure when n-bit bits are combined with an 8 × 8 upper quantized DCT coefficient matrix and a lower quantized DCT coefficient matrix.
A flowchart of bit combination will be described with reference to FIGS.

In step 301, j representing the vertical position is set to 0. At step 302, index i representing the position in the lateral direction is set to 0, the leftmost upper quantized DCT coefficients DH _j0 and lower quantized DCT coefficients DL _j0 is bit synthesized in a row representing the lateral matrix It is targeted. The upper quantized DCT coefficient DH ₀₀ and the lower quantized DCT coefficient DL ₀₀ are the targets of the first bit combination.

In step 303, an n-bit left shift operation is performed on the upper quantized DCT coefficient _DHji ,
A quantized DCT coefficient _Dji is obtained. An arithmetic expression for _{obtaining the} quantized DCT coefficient _Dji is as shown in Expression (2).

D _ji = DH _ji × 2 ⁿ + DL _ji (2)

For example, in the case of 1-bit bit synthesis, the upper quantized DCT coefficient DH ₂₁ (= −4) and the lower quantized DCT
The value of the quantized DCT coefficient D ₂₁ for the T coefficient DL ₂₁ (= 1) is D _ji = −4 × 2 ¹ + 1 = −7.

For example, in the case of 3-bit bit synthesis, the upper quantized DCT coefficient DH ₂₁ (= −1) and the lower quantized DCT coefficient
For the T coefficient DL ₂₁ (= 1), the value of the quantized DCT coefficient D ₂₁ is D ₂₁ = −1 × 2 ³ + 1 = −7.

In step 304, 1 is added to the subscript i. Thereby, the upper quantization D to be bit-combined
The CT coefficient DH _ji and the lower-order quantized DCT coefficient DL _ji are moved to the immediately next upper-order quantized DCT coefficient DH _ji and the lower-order quantized DCT coefficient DL _ji .

In step 305, it is determined whether or not the subscript i is 8. That is, for one row in the horizontal direction, all upper quantized DCT coefficients DH _ji and lower quantized DCT coefficients
It is determined whether or not the T coefficient _DLji has been bit-combined.
If it is determined that the subscript i is 8, the process proceeds to step 306. If it is determined that the subscript i is not 8, the process returns to step 303.

In step 306, 1 is added to the subscript j. That is, the upper quantization D to be bit-combined
The CT coefficient DH _ji and the lower quantized DCT coefficient DL _{ji correspond} to the upper quantized DCT coefficient DH _ji and the lower quantized DC
_Transferred to T coefficient _DLji .

In step 307, it is determined whether or not the subscript j is 8. That is, all upper quantization DCTs
It is determined whether or not the coefficient DH _ji and the lower quantized DCT coefficient DL _ji have been bit-combined. If it is determined that the subscript j is 8, the bit combination for the 8 × 8 upper quantized DCT coefficient matrix DH and lower quantized DCT coefficient matrix DL ends. If it is determined that the subscript j is not 8, step 302
Return to

As described above, according to the present embodiment, the bit string of the quantized DCT coefficient _Dji is bit-separated into an upper bit string and a lower bit string, and the upper quantization D
Lower quantization D represented by CT coefficient _DHji and lower bit string
By _{obtaining the} CT coefficient _DLji , the image can be efficiently compressed. Also, the upper quantization DCT coefficient DH
By bit synthesizing _ji and lower quantized DCT coefficients DL _ji, it is possible to restore the quantized DCT coefficients D _ji, to restore the image of the substantially unstrained.

To compare the conventional lossless encoding method with the image compression method according to the present invention, a color still image is compressed using each image compression method, and the compression ratio (the ratio of the compressed image data amount to the original image data amount) is calculated. ) Was measured. However,
In the image compression method of the present invention, quantization is not performed to prevent image degradation. In the case of the conventional lossless encoding method,
The compression ratio was 75.6%, whereas the compression ratio of the image compression system according to the present invention was 70.9%. From this measurement result, it is understood that the image compression method according to the present invention can compress the image more efficiently than the conventional lossless encoding method.

In the image compression processing, Hadamard transform may be applied instead of the DCT operation. In this case, the inverse Hadamard transform is used for the image expansion processing instead of the IDCT operation.

[0105]

As described above, according to the present invention, a high-resolution image can be compressed and decompressed with almost no distortion.
And it can be compressed efficiently.

[Brief description of the drawings]

FIG. 1 is a block diagram of an image compression apparatus to which an embodiment of the present invention has been applied.

FIG. 2 shows a pixel block, a DCT coefficient matrix, and a quantized DCT.
It is a figure showing a matrix.

FIG. 3 is a diagram showing a quantized DCT coefficient matrix, an upper quantized DCT coefficient matrix, and a lower quantized DCT coefficient matrix.

FIG. 4 is a diagram illustrating bit separation of two quantized DCT coefficients.

FIG. 5 is a flowchart showing a procedure of bit separation.

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

FIG. 7 shows an upper quantization DCT coefficient matrix and a lower quantization DCT.
FIG. 3 is a diagram illustrating a coefficient matrix and a quantized DCT matrix.

FIG. 8 is a diagram showing a quantized DCT coefficient matrix, a DCT coefficient matrix, and a pixel block.

FIG. 9 shows two upper quantized DCT coefficients and a lower quantized DC.
FIG. 9 is a diagram illustrating bit combination of T coefficients.

FIG. 10 is a flowchart illustrating a procedure of bit combination.

[Explanation of symbols]

 10 Image compression device 30 Image decompression device

────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] January 29, 1999 (1999.1.2
9)

[Procedure amendment 1]

[Document name to be amended] Statement

[Correction target item name] 0036

[Correction method] Change

[Correction contents]

By performing the bit separation, the quantized DCT coefficient D ₁₁ (= 36) is right-shifted. The shift operation is an operation for shifting each bit right or left by a designated number of bits. When a one bit right shift operation is performed, each bit is shifted one by one to the right,
The rightmost least significant bit K which is "0" is separated. The left most significant bit left blank by the right shift operation is filled with a sign bit F of “0”.

[Procedure amendment 2]

[Document name to be amended] Statement

[Correction target item name] 0055

[Correction method] Change

[Correction contents]

For example, in the case of 1-bit separation, the lower-order quantized DCT coefficient D ₂₁ (= −7) is
The value of the T coefficient DL ₂₁ is the upper quantized DCT coefficient DH ₂₁
Is −4, DL ₂₁ = −7 − (− 4 × 2 ¹ ) = 1.

[Procedure amendment 3]

[Document name to be amended] Statement

[Correction target item name] 0056

[Correction method] Change

[Correction contents]

For example, in the case of bit separation of 3 bits, the quantized DCT coefficient D ₂₁ (= −7) is
The value of the T coefficient DL ₂₁ is the upper quantized DCT coefficient DH ₂₁
Is −1, DL ₂₁ = −7 − (− 1 × 2 ³ ) = 1.

[Procedure amendment 4]

[Document name to be amended] Statement

[Correction target item name] 0096

[Correction method] Change

[Correction contents]

[0096] For example, in the case of 1-bit bit synthetic, upper quantized DCT coefficients _DH 21 (= -4) and lower quantized DCT coefficients _DL 21 (= 1) the quantized DCT coefficients _{D 2 1} value for the , D ₂₁ = −4 × 2 ¹ + 1 = −7.

[Procedure amendment 5]

[Document name to be amended] Statement

[Correction target item name] 0101

[Correction method] Change

[Correction contents]

In step 307, it is determined whether or not the subscript j is 8. That is, all upper quantization DCTs
It is determined whether or not the coefficient DH _ji and the lower quantized DCT coefficient DL _ji have been bit-combined. If it is determined that the subscript j is 8, the bit combination for the 8 × 8 upper quantized DCT coefficient matrix DH and lower quantized DCT coefficient matrix DL ends. If it is determined that the subscript j is not 8, step 3
Return to 02.

[Procedure amendment 6]

[Document name to be amended] Statement

[Correction target item name] 0102

[Correction method] Change

[Correction contents]

As described above, according to the present embodiment, the bit string of the quantized DCT coefficient D _ji is bit-separated into an upper bit string and a lower bit string, and the upper quantized DCT coefficient DH _ji and the lower bit string represented by the upper bit string are separated. By obtaining the lower quantized DCT coefficient DL _ji represented, the image can be efficiently compressed. Also, upper quantization DCT
By bit combining the coefficient DH _ji and the lower-order quantized DCT coefficient DL _ji , the quantized DCT coefficient D _ji can be restored, and an almost distortion-free image can be restored.

 ──────────────────────────────────────────────────続 き Continuing on the front page F term (reference) 5C057 AA03 BA14 EA01 EC01 EK04 EL01 EM09 EM11 EM16 GF01 GF02 GM01 5C059 KK03 LA01 MA00 MA23 MA34 MC02 MC14 MC33 MC35 MC36 MC38 ME02 PP01 PP14 SS15 UA02 UA05 BA01BA21 BA21 CA22 DA01 DA02 DB18

Claims (6)

[Claims] 1. An orthogonal transformation means for orthogonally transforming image data corresponding to an input still image to obtain an orthogonal transformation coefficient represented by a bit string, and converting the bit string of the orthogonal transformation coefficient into an upper bit string and a lower bit string. Bit separation means for obtaining an upper orthogonal transform coefficient represented by the upper bit string and a lower orthogonal transform coefficient represented by the lower bit string; and encoding the upper orthogonal transform coefficient to obtain a compressed image. An image compression apparatus, comprising: encoding means for obtaining data; 2. The image compression apparatus according to claim 1, wherein said bit separation means separates the quantized orthogonal transform coefficients into bits. 3. The bit separation means performs a right shift operation on the orthogonal transform coefficient represented by a bit sequence, and converts the bit sequence of the orthogonal transform coefficient converted by the right shift operation into the upper bit sequence and the right shift operation. The image compression apparatus according to claim 1, wherein the upper-order orthogonal transform coefficient and the lower-order orthogonal transform coefficient are obtained by using a bit string to be separated as the lower-order bit string. 4. A decoding means for obtaining the upper orthogonal transform coefficient by decoding the input compressed image data, and the higher order orthogonal coefficient obtained by the input lower orthogonal transform coefficient and the higher order orthogonal coefficient obtained by the decoding means. Bit combining means for calculating the orthogonal transform coefficient by bit combining orthogonal transform coefficients, and inverse orthogonal transform means for obtaining the image data by inverse orthogonally transforming the orthogonal transform coefficient. Image expansion device. 5. The image decompression device according to claim 4, wherein the inverse orthogonal transform unit performs inverse orthogonal transform on the inversely quantized orthogonal transform coefficient. 6. The orthogonal transform coefficient by performing a left shift operation on the upper orthogonal transform coefficient represented by a bit sequence and applying the lower bit sequence to a bit portion that becomes blank by the left shift operation. The image decompression device according to claim 4, wherein