JPH10234041A - Image encoding method, image decoding method, image encoder and image decoder - Google Patents

Abstract

PROBLEM TO BE SOLVED: To encode an image signal with excellent quality by deciding an adjacent block to predict a low-order coefficient value of an encoding object block based on the difference between corresponding low-order coefficient values of adjacent blocks in the case of predicting a low-order coefficient value from a low-degree coefficient among orthogonal transformation coefficients. SOLUTION: An encoding object image 1 is input to a rectangular block division section 2, where the image 1 is divided into uniform rectangular blocks. A block unit encoding object image 3 is input to an orthogonal transformation section 4, in which the image 3 is encoded by DCT or the like, and the pixel value is converted into the frequency domain. A low-order coefficient 6 among conversion coefficients obtained by the DCT is quantized by a quantization section 5 and given to an adaptive low-order coefficient predict section 9 along with a low-order coefficient value 8 of an adjacent block stored in a low-order coefficient value memory 7, in which adaptive low-order prediction is conducted. After a block being an object of low-order coefficient prediction is decided, a difference between a difference between a low-order coefficient of a block and a low-order coefficient of the coding object block is obtained and outputted as a low-order coefficient difference 10.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image encoding method and an image decoding method in which an image to be encoded is divided into rectangular blocks, and each block is encoded using orthogonal transform.

[0002]

2. Description of the Related Art ITU-T recommendation H.264. H.263 is a communication video coding standard entitled "Low Bit Rate Communication Video Coding Scheme".

[0003] H. In 263, the frame is divided into rectangular coding units (macroblocks: 16 × 16 pixels), and temporal redundancy of the moving image signal is suppressed using motion compensation prediction. The motion compensation prediction is performed for each macroblock, and selectively switches between intra-frame coding (intra) and inter-frame coding (inter) for each macroblock.

When a certain macroblock is coded as intra, coding is performed by discrete cosine transform (DCT), which is one of typical orthogonal transforms, for each orthogonal transform unit (block: 8 × 8 pixels). . At this time, as shown in FIG. 6, the input image has a luminance (Y) and a color difference (Cb, Cb) in advance.
r) and the input signal is 4: 2: 0,
The luminance signal (Y) has four times the information amount of the color difference signals (Cb, Cr), and there are four orthogonal transformation units (blocks) in one macroblock in the case of a luminance signal.

Similarly, when a certain macroblock is coded as an inter, temporal redundancy of a moving picture signal is suppressed by motion compensation prediction for each macroblock, and a motion compensation prediction difference is suppressed by DCT.

Further, in an environment where the quality of the transmission path is poor, image quality degradation occurs due to bit errors or dropping of buckets, and all macroblocks are intra-framed at regular time intervals in order to prevent the degradation from propagating to subsequent frames. Encoding is recommended.

When an encoding target block is intra-coded using DCT, the DC coefficient value which is the lowest coefficient of the block is equal to the average value of the pixels in the block. Can be taken. But,
H. In 263, the DC coefficient value of each block is described in coded data without any compression using an 8-bit fixed-length code.

On the other hand, IS, which is a video encoding standard for storage,
In O / IEC 11172 and ISO / IEC 10918/1, which is a still image coding standard, a technique called DC prediction (Intra DC Prediction) is used to remove this redundancy. In this method, the difference between the DC coefficient value of the block coded immediately before coding the block to be coded and the DC coefficient value of the block to be coded is variable-length coded.

FIG. 7 shows ISO / IEC 11172, I
It shows how to perform DC prediction of a luminance signal in SO / IEC 10918/1 or the like. The zigzag order of the encoding indicated by the arrow indicates that in the case of a luminance signal, four orthogonal transform units (blocks) are included in one macroblock (encoding unit), and in each macroblock,
This is because encoding is performed for each block in the order of upper left → upper right → lower left → lower right.

[0010]

When encoding in intra without using inter-frame prediction, H.264 is used. In H.263 and the like, prediction of low-order coefficients including DC coefficients is not performed, and the coding is described in the encoded data without sufficiently removing the redundancy of the original signal.

[0011] Also, ISO / IEC 11172, IS
In O / IEC 10918/1, DC prediction is used to remove spatial redundancy across blocks. However, since DC prediction is performed only from the block coded immediately before, sufficient coding efficiency can be obtained. Not raised.

An object of the present invention is to provide an image encoding method and apparatus capable of encoding an image signal with high quality, and an image decoding method and apparatus corresponding thereto.

[0013]

According to the image encoding method of the present invention, when encoding the low-order coefficient value of the orthogonal transform of the encoding target block, the encoding of the already encoded block adjacent to the encoding target block is performed. Among the orthogonal transform coefficients, when predicting the low-order coefficient value from a low-order coefficient corresponding to the low-order coefficient and encoding the prediction difference value with a variable-length code, a corresponding low-order coefficient between adjacent blocks is used. An adjacent block for predicting a low-order coefficient value of the current block is determined based on the difference between the numerical values.

Here, "predict the low-order coefficient value from the low-order coefficient corresponding to the low-order coefficient" means "the DC coefficient is predicted from the DC coefficient of the adjacent block, and the AC first coefficient is the AC coefficient of the adjacent block. The second coefficient is predicted from the first coefficient, and the second AC coefficient is predicted from the second AC coefficient of the adjacent block. "

The image decoding method according to the present invention decodes a predicted low-order difference value from a variable-length code in encoded data, and obtains a low-order coefficient value between already-decoded blocks adjacent to a decoding target block. It is determined from which of the lower-order coefficients of the block the coding target block is predicted from the difference of the blocks, and the decoded predicted lower-order difference value and the lower-order coefficient value of the determined block are added. The low order coefficient of the decoding target block is obtained.

According to the present invention, which block is predicted from the correlation between the corresponding low-order coefficient values of the blocks which are already coded and which are adjacent to the block to be coded, is adaptively used by both the encoder / decoder. The image signal can be coded with high quality.

According to the embodiment of the present invention, of the adjacent blocks which have been coded, the left of the block to be coded,
Using the four blocks adjacent to the upper left, upper, and upper right, the lower order coefficient value of the block adjacent to the left of the current block, the lower order coefficient value of the block adjacent to the upper left of the current block, The combination in which the correlation between the low-order coefficient values of the blocks adjacent above and the corresponding low-order coefficient values of the three blocks is greatest is determined, and the low-order coefficient value of the current block is set to the left of the current block. When the correlation between the low-order coefficient values of the blocks adjacent to the upper left is maximized, the prediction is performed from the lower-order coefficient values of the blocks adjacent above the current block, and the upper left and upper neighbors of the current block are predicted. When the correlation of the low-order coefficient values of the block to be maximized is maximum, the prediction is performed based on the low-order coefficient values of the block adjacent to the left of the current block and the low order of the blocks adjacent to the top and left of the current block. Next When the correlation of the numerical values becomes maximum, the prediction is performed based on the low-order coefficient value of the block adjacent to the upper right of the current block, and the low-order coefficient value of the selected adjacent block and the low-order coefficient value of the current block. The difference between the coefficient values is described in coded data using a variable length code.

According to the embodiment of the present invention, a predicted low-order difference value is decoded from a variable-length code in coded data, a low-order coefficient value of a block adjacent to the left of the block to be decoded, and an upper left corner of the block to be decoded. And the combination of the low-order coefficient values of the blocks adjacent to the block to be decoded and the low-order coefficient values of the blocks adjacent to the block to be decoded, which correspond to the low-order coefficient values of the three blocks, are determined. When the low-order coefficient value of the block has the maximum correlation between the low-order coefficients of the blocks adjacent to the left and upper left of the current block, the prediction is made from the low-order coefficient values of the blocks adjacent to the current block, When the correlation between the low-order coefficient values of the upper left and upper adjacent blocks of the current block to be decoded is maximized, the lower-order coefficient value of the block adjacent to the left of the current block is predicted and the current block of the current block is decoded. When the correlation between the low-order coefficient values of the blocks adjacent to the top and the left is maximum, it is determined that the prediction is made from the low-order coefficient values of the block adjacent to the upper right of the current block to be decoded. The low-order coefficient value of the block to be decoded is obtained by adding the low-order coefficient value and the decoded predicted low-order difference value.

The image encoding apparatus according to the present invention encodes a low-order coefficient value of an orthogonal transform of a block to be encoded,
Among the orthogonal transform coefficients of the already coded blocks adjacent to the current block, the low-order coefficient value is predicted from the low-order coefficient corresponding to the low-order coefficient, and the prediction difference value is represented by a variable-length code. When encoding, an adaptive low-order prediction unit that determines an adjacent block for predicting a low-order coefficient value of the current block from a difference between corresponding low-order coefficient values between the adjacent blocks,
A variable length encoding unit that encodes the low-order coefficient difference;

An image decoding apparatus according to the present invention includes a variable length decoding unit for decoding a predicted low-order difference value from a variable length code in coded data, and a low-length decoding unit which is adjacent to a decoding target block and has already been decoded. From the difference between the next-order coefficient values, it is determined from which low-order coefficient value the block to be decoded is predicted, and the above-described decoded predicted low-order difference value and the low-order coefficient value of the determined block are added. And an adaptive low-order coefficient prediction unit for obtaining a low-order coefficient value of the decoding target block.

[0021]

Next, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing an image encoder according to an embodiment of the present invention.

In this embodiment, for the sake of convenience, the case where the entire screen is coded in a frame is considered. Also, the case where DCT which is one of typical orthogonal transforms is used as the orthogonal transform is described.

In the encoder, first, the image 1 to be encoded is
It is input to the rectangular block dividing unit 2 and divided into uniform rectangular blocks. The image 3 to be coded in the block unit is input to the orthogonal transform unit 4, coded by DCT or the like, and the pixel value is converted to the frequency domain.

Here, the DCT used is called a two-dimensional DCT, which is given by equation (1), and its inverse transform is given by equation (2). Here, N is the length of one side of the block.

[0026]

(Equation 1) At this time, F (0,0) is the coefficient of the DC coefficient component,
F (u, v) other than (0, 0) is called an AC coefficient component, and the larger u, v, the higher the frequency corresponds to the transform coefficient. When both u and v of F (u, v) are close to 0, the coefficient is called a low-order coefficient in the present invention. Also, in the present invention, the coefficient of a certain two blocks

[0027]

[Outside 1] Is called a “corresponding coefficient”. Where i,
j is

[0028]

[Outside 2] Any integer that satisfies

The lower-order coefficient 6 (F _current (i, j)) of the transform coefficients obtained by DCT is quantized by the quantization unit 5 and then stored in the lower-order coefficient value memory 7 in the adjacent block. Input to the adaptive low-order coefficient prediction unit 9 together with the low-order coefficient value 8,
Adaptive low-order coefficient prediction is performed.

As shown in FIG.
Is a block Bl adjacent to the left of the current block.
_a , A block Bl adjacent to the upper left_b , The adjoining bro
Cook Bl_c , The block Bl adjacent to the upper right_d Each of
Low order coefficient F_a (I, j), F _b (I, j), F_c 
(Ij), F_d (I, j) are stored.

FIG. 2 shows the processing of the adaptive low-order coefficient prediction unit 9.
It is a flowchart. First, the left block Bl_a Low
Order coefficient value F_a (I, j), upper left block Bl_b Low order
Coefficient value F_b (I, j), upper block Bl_c Low order coefficient of
Value F_c (I, j), upper right block Bl_d Low order coefficient value of
F_d (I, j) is input from the low-order coefficient value memory 7.
Next, | F_a (I, j) -F_b (I, j) |, F_b 
(I, j) -F_c (I, j) |, | F_c (I, j) -F
_a (I, j) | (step 20), which value is the minimum
The target block for low-order coefficient prediction adaptively
Switch. For example | F_a (I, j) -F_b (I, j) |
Is minimum, it is said that the correlation is high in the vertical direction of the screen.
That is, the block Bl above the block to be encoded_c 
Low order coefficient F _c From (i, j), the low-order coefficient (F
_current(i, j)) is more efficient (step
21 and 22). Similarly, | (F_b (I, j) -F_c 
When (i, j) | is minimum, the correlation is in the horizontal direction of the screen.
Is high, and the block to the left of the current block is
Rock Bl_a Low order coefficient F_a (I, j) to F
_currentIt is efficient to predict (i, j) (step
24, 25). Also, | F_c (I, j) -F_a (I,
j) When | is minimum, the correlation is high in the oblique direction of the screen.
The upper right block of the block to be coded.
Cook Bl_d Low order coefficient F_d(I, j) to F
_currentIt is efficient to predict (i, j) (step
24, 27).

After the block to be subjected to the low-order coefficient prediction is determined in this way, the difference between the low-order coefficient of the block and the low-order coefficient of the block to be coded is determined (steps 23, 26, 2).
8), and output as low-order coefficient difference 10 (step 2)
9).

The calculated low-order coefficient difference 10 is input to the variable length coding unit 11, compressed and coded, and sent to the multiplexing unit 12. Examples of encoding of low-order coefficient differences include ISO /
Those used for IEC 11172 and the like can be used as they are.

The AC coefficient 13 among the transform coefficients obtained by the DCT is input to the quantizer 14 and, after being quantized, input to the variable-length encoder 15 to perform two-dimensional variable-length encoding. After that, it is input to the multiplexing unit 12.

The multiplexing unit 12 converts the low-order prediction difference value 10 of the low-order coefficient component into a variable-length code,
3 is combined with variable-length coded data and output as block coded data 16.

FIG. 3 is a block diagram of a decoder corresponding to the encoder of FIG.

The block coded data output from the coder
The data 16 is a separating unit 30, which is an encoded low-order coefficient difference.
And quantized and coded AC coefficients.
The coded low-order coefficient difference is converted to a low-order coefficient
It is decoded into the number difference 10. The adaptive low-order coefficient prediction unit 32
From the lower order coefficient value memory 33 and the adjacent block difference
Block low order coefficient value 8 (block Bl_a Low order coefficient value F_a 
(I, j), block Bl _b Low order coefficient value F_b (I,
j), block Bl_c Low order coefficient value F_c (I, j),
Rock Bl_d Low order coefficient value F_d (I, j))
The low-order coefficient 5 is obtained by the processing shown in FIG. First, | F
_a (I, j) -F_b (I, j) |, | F_b (I, j)-
F_c (I, j) |, F_c (I, j) -F_a (I, j) |
(Step 40). ｜ F_a (I, j) -F_b 
When (i, j) | is minimum, the block Bl_c Low-ranking person
Number F_c It is determined that low-order coefficient prediction has been performed at (i, j).
And F_c(I, j) and the low-order coefficient F of the block to be decoded
_current(i, j) is added (steps 41 to 43).
｜ F_b (I, j) -F_c When (i, j) | is minimum,
Block Bl_a Low order coefficient F_a (I, j) is the lower order coefficient
It is determined that measurement has been performed, and F_c (I, j) and F
_current(i, j) is added (steps 44 to 46),
｜ F_a (I, j) -F_c When (i, j) | is minimum,
Block Bl_d Low order coefficient F_d (I, j) is the lower order coefficient
It is determined that measurement has been performed, and F_d (I, j) and F
_current(i, j) is added (steps 44, 47, 4)
8). Thereafter, the calculated low-order coefficient value 6 is output (step
49). On the other hand, the quantized and coded AC coefficients
Is decoded by the variable length decoding unit 34, and the inverse
It is converted to an AC coefficient 13. Low-order coefficient 6 is inverse quantized
Inverse orthogonal transform with the AC coefficient 13 quantized by the unit 36
Input to the section 37 to generate and output the block unit decoded image 3.
Is forced.

In this embodiment, the entire screen is blocked.
In the case of encoding in a closed form, ISO / IE
C. 11172 and H.C. P- for video coding like 263
Intra-coded blocks and images, such as Picture
In the case where blocks coded by
This method is also used only for intra-coded blocks.
It is possible to use by applying. Specifically, adjacent
Block (Bl_a , Bl _b , Bl_c , Bl_d )
If at least one is coded in intra,
Thus, it can be realized by performing adaptive low-order coefficient prediction.

[0039]

As described above, according to the present invention, when the current block is coded in intra, the correlation between the low-order coefficient values of the blocks adjacent to the current block is calculated.
By adaptively switching which block to predict from in both the encoder and the decoder, it is possible to sufficiently remove the spatial redundancy of the input image, and to encode the image signal with high quality. .

[Brief description of the drawings]

FIG. 1 is a configuration diagram of an image encoder according to an embodiment of the present invention.

FIG. 2 is an adaptive low-order coefficient prediction unit 6 in the embodiment of FIG.
6 is a flowchart showing the processing of FIG.

FIG. 3 is a configuration diagram of an image decoder corresponding to the image encoder of FIG. 1;

FIG. 4 is a flowchart illustrating a process of an adaptive low-order coefficient prediction unit 32 in the image decoder of FIG. 3;

FIG. 5 is an explanatory diagram of an encoding target block and adjacent blocks.

FIG. 6 is a diagram showing definitions of an input image signal and blocks and macroblocks.

FIG. 7 is a diagram showing a low-order coefficient prediction method in a conventional image coding method.

[Description of Code] 1 Image to be encoded 2 Rectangular block division unit 3 Image to be encoded in block units 4 Orthogonal transformation unit 5 Quantization unit 6 Low-order coefficient 7 Low-order coefficient value memory 8 Neighboring block low-order coefficient value 9 Adaptive low Order coefficient prediction unit 10 Low order coefficient difference 11,15 Variable length coding unit 12 Multiplexing unit 13 AC coefficient 14 Quantization unit 16 Block coded data 20-29 Step 30 Separation unit 31,34 Variable length decoding unit 32 Adaptive low Higher order coefficient prediction unit 33 Low order coefficient value memory 35, 36 Inverse quantization unit 37 Inverse orthogonal transformation unit 40 to 49 steps

Claims (6)

[Claims] 1. An image encoding method for dividing an encoding target image into rectangular blocks and encoding the encoding target image using orthogonal transform for each block, comprising: In encoding a numerical value, among the orthogonal transform coefficients of an already-encoded block adjacent to the current block, the low-order coefficient value is predicted from the low-order coefficient corresponding to the low-order coefficient. An image characterized in that, when encoding a difference value with a variable-length code, an adjacent block for predicting a low-order coefficient value of a current block is determined from a difference between corresponding low-order coefficient values between the adjacent blocks. Encoding method. 2. The encoded adjacent block,
Using the four blocks adjacent to the left, upper left, upper, and upper right of the current block, the lower order coefficient value of the block adjacent to the left of the current block, the lower order value of the block adjacent to the upper left of the current block, The combination of the coefficient value, the lower order coefficient value of the block adjacent to the current block to be encoded, and the corresponding lower order coefficient value of the three blocks, which has the largest correlation with each other, is obtained. When the correlation between the low-order coefficient values of the blocks adjacent to the left and upper left of the current block is maximized, the coefficient value is predicted from the low-order coefficient values of the blocks adjacent above the current block, When the correlation between the lower-order coefficient values of the upper left and upper adjacent blocks of the current block is maximum, the prediction is performed based on the lower-order coefficient values of the block adjacent to the left of the current block. Block When the correlation between the low-order coefficient values of the blocks adjacent to the top and left of is maximum, the prediction is performed based on the low-order coefficient values of the block adjacent to the upper right of the current block. 2. The image coding method according to claim 1, wherein a difference between the next-order coefficient value and the low-order coefficient value of the current block is described in coded data by a variable length code. 3. An image decoding method for decoding encoded data encoded by the image encoding method according to claim 1, wherein a predicted low-order difference value is decoded from a variable length code in the encoded data. From the difference between the low-order coefficient values between the blocks that have already been decoded and are adjacent to the current block, it is determined from which low-order coefficient value the block to be decoded is predicted. An image decoding method, comprising: adding a low-order coefficient value of a block determined as a difference value to obtain a low-order coefficient value of the current block. 4. An image decoding method for decoding encoded data encoded by the image decoding method according to claim 3, wherein a predicted low-order difference value is decoded from a variable length code in the encoded data, The three low-order coefficient values of the low-order coefficient value of the block adjacent to the left of the block, the low-order coefficient value of the block adjacent to the upper left of the block to be decoded, and the low-order coefficient value of the block adjacent to the top of the block to be decoded. When a combination in which the mutual correlation of the next coefficient values is the largest is obtained, and the low order coefficient value of the current block to be decoded is such that the correlation of the low order coefficients of the blocks adjacent to the left and upper left of the current block to be decoded is maximum Is predicted from the low-order coefficient value of a block adjacent above the current block, and when the correlation between the low-order coefficient value of the upper left block and the low-order coefficient value of the block adjacent above the current block becomes maximum, the decoding is performed. Predicted from the low-order coefficient value of the block adjacent to the left of the target block, and when the correlation between the low-order coefficient values of the block adjacent to the upper left of the current block and the left is maximum, the upper right of the current block to be decoded is displayed. Determining that the prediction is performed based on the low-order coefficient value of the adjacent block; adding the determined low-order coefficient value of the adjacent block to the decoded low-order difference value; An image decoding method characterized by obtaining a numerical value. 5. An image encoding apparatus which divides an encoding target image into rectangular blocks, and encodes the encoding target image using orthogonal transform for each block. In encoding a numerical value, among the orthogonal transform coefficients of an already-encoded block adjacent to the current block, the low-order coefficient value is predicted from the low-order coefficient corresponding to the low-order coefficient. When encoding the difference value with a variable-length code, an adaptive low-order prediction unit that determines an adjacent block for predicting a low-order coefficient value of the current block from a difference between corresponding low-order coefficient values between the adjacent blocks. An image encoding apparatus, comprising: a variable-length encoding unit that encodes the low-order coefficient difference. 6. An image decoding apparatus for decoding encoded data encoded by the image encoding apparatus according to claim 5, wherein a variable length decoding unit decodes a predicted low-order difference value from a variable length code in the encoded data. The decoding unit, adjacent to the decoding target block, from the difference between the low-order coefficient values between the blocks that have already been decoded, to determine from which block the lower-order coefficient value of the encoding target block is predicted, An image decoding apparatus comprising: an adaptive low-order coefficient prediction unit that adds a decoded predicted low-order difference value and a low-order coefficient value of a determined block to obtain a low-order coefficient value of the current block. .