JP2900998B2 - Block interpolation predictive encoding apparatus, decoding apparatus, encoding method and decoding method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention
In high-efficiency coding for converting an image into a digital signal with a smaller code amount for display, the present invention particularly relates to a method of coding an image in block units.

[0002]

[Prior art]

<Inter-block predictive coding> In image coding, DC
A method of encoding an image in block units using T (discrete cosine transform) or the like is most commonly used. Therefore, instead of encoding each block independently, each image of the encoded block is predicted from the encoded block,
There is a method of encoding the prediction residual. This is, for example, "Extrapolation Prediction-High Efficiency Coding of Images by Discrete Sine Transform" IEICE Transactions (B), J 71-No.6, pp.717-
724 (June 1988). In the case of the above document, since the encoding process is sequentially performed on a block basis, the encoded block exists only on one side of the encoded block. Therefore, the inter-block prediction is performed only from one of the coded blocks, that is, extrapolation prediction.

It is known that DST (Discrete Sine Transform) has higher coding efficiency than DCT for coding such a residual signal of inter-block prediction. In this method, since the correlation between the blocks is used for coding, the coding efficiency is better than coding independently for each block. Further, since the prediction signal is obtained as a continuous signal value from an adjacent block, even if the prediction residual is not completely coded, continuity between the blocks is improved, and block distortion hardly occurs.

<Conventional Coding Apparatus> FIG. 5 shows an example of a conventional interblock predictive coding apparatus. The image signal coming from the image input terminal 1 is output to the prediction subtractor 4 by the block converter 51 by converting the image signal of the raster scan structure into a block unit structure. The prediction subtractor 4 is a block predictor 53
Is subtracted from the image signal of the block to be coded, and the prediction residual signal is supplied to the encoder 52. The encoder 52 performs a DST conversion process, and quantizes the obtained conversion coefficient to obtain a code. The obtained code is output from the code output terminal 7 and supplied to the decoder 55.

A decoder 55 obtains a reproduction coefficient by inverse quantization,
The reproduction prediction residual signal is obtained by the inverse DST and supplied to the prediction adder 17. The prediction adder 17 adds the inter-block prediction signal supplied from the block predictor 53 to the reproduction prediction residual signal, and supplies the obtained reproduction image to the block line memory 54. The block line memory 54 accumulates reproduced image signals corresponding to blocks (block lines) from right to left of the image in block units, and outputs information necessary for prediction to the block predictor 53. The block predictor 53 generates a prediction signal of each pixel of the block to be coded and supplies the prediction signal to the prediction subtractor 4.
The specific method of forming the prediction signal is described in the above-mentioned document.

<Conventional Image Decoding Apparatus> A decoding apparatus corresponding to the inter-block encoding apparatus shown in FIG. 5 will be described.
FIG. 6 shows the configuration. Sign input terminal 2
The code coming from 1 is decoded by a decoder 55,
ST processing is performed, and the prediction residual signal is reproduced. The reproduced prediction residual signal is added by the prediction adder 17 to the inter-block prediction signal supplied from the block predictor 53 to form a reproduced image. The reproduced image thus obtained is supplied to the raster converter 61 and also to the block line memory 54. The raster converter 61 is the inverse process of the block converter 51, and converts a reproduced image in units of blocks into a normal television signal having a raster scan structure, and outputs it from the image output terminal 24. The operations of the block line memory 54 and the block predictor 53 are the same as those of the encoding device. The inter-block prediction signal obtained by the block predictor 53 is provided to the prediction adder 17.

[0007]

In the conventional inter-block prediction, prediction is performed only from one of the encoded blocks. In this case, the change in the image cannot be predicted well, and the improvement in the coding efficiency is not sufficient. Further, since the prediction is cyclic, if a code error occurs in the transmission path, it affects the subsequent images. The present invention has been made by paying attention to the above points, and independently encodes blocks separated from encoded blocks, interpolates and predicts blocks sandwiched between the blocks and the encoded blocks, and encodes residuals. By doing so, it is an object to provide an inter-block prediction encoding device and a decoding device capable of obtaining high encoding efficiency without spreading a code error.

[0008]

According to the present invention, when an image is sequentially coded and decoded in block units, a coded block and a separated block (D) are independently coded to obtain a code. At the same time, an encoding means (6, 18) for locally decoding the code to obtain a decoded image, and a prediction block (B, C) sandwiched between upper and lower or left and right between the encoded block and the independent block. For each pixel, generate an interpolation prediction signal from the coded block and the independent block, and from each pixel in the prediction block, code the prediction residual signal obtained by subtracting the interpolation prediction signal to obtain a code. And an encoding means (4, 6, 18) for locally decoding the code to obtain a decoded image, and a prediction block sandwiched between an encoded block and a previously encoded prediction block. For each pixel in (A) Then, an interpolation prediction signal is generated from the coded blocks and the coded prediction blocks existing in the vertical and horizontal directions of the prediction block (A), and the interpolation prediction signal is generated from each pixel in the prediction block. An inter-block predictive coding apparatus including coding means (4, 6) for coding a prediction residual signal obtained by subtraction and obtaining a code.

Further, the decoding means (22) for decoding the decoded block and the separated block (D) independently within the block to obtain a decoded image, and the upper and lower sides or the left and right sides are determined by the decoded block and the independent blocks. For each pixel in the interposed prediction block (B, C), an interpolated prediction signal is generated from the decoded block and the independent block, and the prediction residual of the prediction block is decoded to obtain the decoded prediction obtained. Decoding means for adding a residual signal to an interpolated prediction signal to obtain a decoded image
(22, 17), and for each pixel in the prediction block (A) sandwiched by the decoded block and the previously decoded prediction block, the top and bottom of the prediction block (A) An interpolation prediction signal is generated from the decoded block and the decoded prediction block existing in the direction and the left and right directions, the prediction residual of the prediction block is decoded, and the obtained decoded prediction residual signal is used as the interpolation prediction signal. And a decoding means (22, 17) for obtaining a decoded image by adding the above. For encoding and decoding in the prediction block, DCT (discrete cosine transform) and DT are performed in a vertical direction and a horizontal direction in a two-dimensional plane according to an interpolation prediction direction.
Use ST (discrete sine transform) properly.

(Operation) In the present invention, a block separated from a coded block is independently coded, and a block sandwiched between the coded block and the coded block is interpolated and predicted by them, and the residual is coded. Therefore, the interpolated block is interpolated and predicted from the upper and lower or left and right encoded blocks and the independent block, and a prediction signal suitable to a certain degree with a change in the image is obtained. In addition, for a block sandwiched between the coded block and the interpolated and predicted block by using the locally decoded image of the interpolated and predicted block, interpolation using all surrounding blocks is performed. Since prediction can be performed, prediction efficiency can be improved. Furthermore, by using DCT and DST in encoding and decoding depending on the prediction direction, the encoding efficiency is improved. On the other hand, although inter-block prediction processing is performed, prediction is performed only on coded blocks and blocks separated therefrom, so that error propagation of decoding processing due to code errors reaches adjacent blocks.

[0011]

BEST MODE FOR CARRYING OUT THE INVENTION

<Inter-Block Prediction Coding Apparatus> One embodiment of the moving picture coding apparatus of the present invention will be described below with reference to the drawings. FIG.
Shows the configuration, and the same components as those in the conventional example of FIG. 5 are denoted by the same reference numerals. The operations of the image input terminal 1, the predictive subtractor 4, the sign output terminal 7, and the adder 17 are basically the same. On the other hand, in FIG. 1, a prediction controller 8, a block memory 9,
11, switches 10, 19, adder 13, multiplier 12,
14, there is a prediction signal generator 20, and switches 3, 5, 1
6 has been added. Further, the operations of the block converter 2, the encoder 6, the block line memory 15, and the decoder 18 are different from those of the conventional example.

An image signal coming from an image input terminal 1 is blocked by a block converter 2 and output to a switch 3. In this case, the order of the blocks is not a simple order as in the conventional example, but is output in the order of D, C, B, and A in units of four, two in the vertical direction shown in step 1 of FIG. In addition, the block converter 2 switches 3, 5, 16,
Output the 19 control signals. The processing for integrating such blocks in units of four is performed when the image format is 4:
This is general processing for a luminance signal in 2: 0 type encoding.

In the switch 3, the independent block of D is connected to the switch 5 according to the control signal, and the prediction blocks of A, B and C are connected to the prediction subtractor 4. The prediction block subtracts the interpolated prediction signal from the switch 19 from the prediction subtractor 4 to form a prediction residual signal, which is provided to the switch 5. The switch 5 is controlled in the same manner as the switch 3, and the independent block supplies the encoder 6 with the signal as it is and the prediction block with the prediction residual signal. The encoder 6 encodes each block, but the processing differs between the independent block and the prediction block. Basically, the signal subjected to the DCT is quantized, and a code whose information is compressed by the variable length coding is supplied to the code output terminal 7 and the decoder 18. Here, in the prediction block, DC
Since T or DST is used and the DC component is small, the variable-length coding may be performed in the same manner as the AC component.

The decoder 18 performs the reverse process of the encoder 6 to obtain a reproduced image and a reproduced prediction residual, and obtains a prediction adder 17.
Give to. In the prediction block, the prediction adder 17 gives an interpolation prediction signal via the switch 16 and adds the signal to the reproduction residual signal to obtain a reproduction image. On the other hand, nothing is added in the independent block. The reproduced image obtained in this way is supplied to the block line memory 15 and the block memories 9 and 11 for the independent block D, and to the block line memory 15 and the block memories 9 and 11 for the predicted blocks B and C, respectively. In the prediction block A, it is provided only to the block line memory 15.

The basic operation of the block line memory 15 is the same as that of the conventional example, but a maximum of 2
It is necessary to hold the signal of the double block. Here, since only the pixels at the end of the block are actually used for prediction, only the pixels held in each memory may be used. The block memories 9 and 11 hold the reproduced image in both memories in the independent block of D, hold the reproduced image in the block memory 9 or 11 at the time of the predicted block processing of B, and store the reproduced image in the block memory 9 or 11 at the time of the predicted block processing of C. Is stored in another memory and another memory storing the reproduced image of the prediction block. The block line memory 15 and the block memories 9 or 11 output an image signal of a pixel at a block end adjacent to a prediction block used for prediction. These are multiplied by the coefficients k and (1-k) by the multipliers 12 and 14, respectively, and added by the adder 13 to become an interpolation prediction signal.
The coefficient k is determined by the prediction controller 8 according to the position of the block and the pixel, and is output.

The prediction controller 8 outputs k and supplies a control signal to the switch 10. The switch 10 is controlled by the control signal so that the block memory 9
And the image signal from both memories of the block memory 11 to the multiplier 12 in order. The specific prediction method is linear prediction as shown in FIG. 4, and k is 1/9 to 8/9 depending on the positional relationship between the predicted pixel and the reference pixel. The interpolation prediction signal thus created is supplied to the switch 19 and the prediction signal generator 20. With the switch 19, the interpolation prediction signal from the adder 13 is used as it is in the prediction block of B and C by the switch 19, and the interpolation prediction signal output from the prediction signal generator 20 in the prediction block of A by the prediction subtractor 4. Give to switch 16.

The prediction signal generator 20 holds two interpolated prediction signals having different prediction directions sequentially given from the adder 13 at the time of the prediction block processing of A.
Generate an interpolation prediction signal for the corresponding pixel. The signal can be generated by averaging two interpolated predicted signals,
An interpolation prediction signal is generated using one of the two interpolation prediction signals according to the position of the pixel.

Steps 1 to 3 in FIG.
Shown in First, in step 1, an independent block of D separated from an already-encoded block is independently encoded in the block to obtain a code, and the code is locally decoded to obtain a decoded image. In step 2, for each pixel in the B and C prediction blocks sandwiched vertically and horizontally by the coded block and the D independent block, the pixels from the decoded image of the coded block and the D independent block are extracted. An interpolated prediction signal is generated, a prediction residual signal obtained by subtracting the interpolated prediction signal from each pixel in the B and C prediction blocks is encoded to obtain a code, and the code is locally decoded and decoded. Get an image.

In step 3, A, which is sandwiched between the upper, lower, left and right by the encoded block and the B and C prediction blocks,
For each pixel in the prediction block of, an interpolation prediction signal in the vertical and horizontal directions is generated from the coded block and the decoded image of the prediction block of B and C, and each pixel in the prediction block of A is generated therefrom. Generate an interpolated prediction signal corresponding to
, A prediction residual signal obtained by subtracting the generated interpolation prediction signal from each pixel in the prediction block is encoded to obtain a code.

<Inter-Block Predictive Decoding Apparatus> An embodiment of a decoding apparatus corresponding to the inter-block predictive encoding apparatus of FIG. 1 of the present invention will be described below with reference to the drawings. FIG. 2 shows the configuration, and the same components as those of the conventional example are denoted by the same reference numerals. Sign input terminal 21, prediction adder 1
7. The operation of the image output terminal 24 is basically the same. On the other hand, FIG. 2 shows a block predictor 5 of FIG.
3. Instead of 3, the prediction controller 8, the block memories 9, 11,
Switches 10, 19, adder 13, multipliers 12, 14,
There is a prediction signal generator 20. Further, switches 3 and 5 are added. The operations of the decoder 22, the raster converter 23, and the block line memory 15 are different from those of the conventional example.

The code input from the code input terminal 21 is subjected to the inverse processing of the encoder 6 in the same manner as the decoder 18 of the encoder, and the reproduced image signal and the reproduced prediction residual signal are converted by the decoder 22. And give it to switch 3. The switch 3 connects the independent block of D to the switch 5 and the prediction blocks of A, B, and C to the prediction adder 17 according to a control signal determined by the processing order of the blocks. The prediction block adds the interpolated prediction signal supplied from the switch 19 to the prediction residual signal in the prediction adder 17 to form a reproduced image, which is supplied to the switch 5. The switch 5 is controlled in the same manner as the switch 3, and the independent block supplies the same signal, and the prediction block supplies the output of the prediction adder 17 to the raster converter 23 and the block line memory 15.

Further, in the case of the independent block, it is given to the block memories 9 and 11, and in the case of the B and C prediction blocks, it is given to either of the block memories 9 and 11, respectively. Prediction controller 8, block memories 9, 11, switches 10, 1
The operations of 9, the multipliers 12, 14, the adder 13, the block line memory 15, and the prediction signal generator 20 are the same as those of the encoding apparatus of FIG. The interpolated prediction signal from the adder 13 obtained in this manner is, as in the coding device, the same signal as it is in the B and C prediction blocks, and the same as the prediction signal by the prediction signal generator 20 in the A prediction block. An interpolation prediction signal generated from two interpolation prediction signals having different directions is provided to the prediction adder 17.

<Equivalent to Claims 2, 4, 6, and 8> In image coding, a method of coding an image in units of blocks using DCT or the like is most commonly used. In the method of predicting each pixel of a coded block from the block of, and coding the prediction residual, as described in the related art, DST has higher coding efficiency than DCT. Are known. This is because the prediction residual better matches the basis function in DST than in DCT in the prediction direction. That is, when the prediction residual is obtained by performing the interpolation prediction in the vertical direction in the prediction block, the prediction residual is a DCT in the vertical direction.
1 shows a tendency similar to that of the first basis function of DST. Therefore, in the encoding of the two-dimensional plane, if the interpolation prediction direction is the vertical direction, DST is used in the vertical direction and DCT is used in the horizontal direction. If the interpolation prediction direction is the horizontal direction, DST is used in the horizontal direction, and DCT is used in the vertical direction.

For example, in the embodiment, since the B prediction block performs vertical interpolation prediction in the vertical direction using the reproduced image in the vertical direction, two-dimensional encoding is performed using DST in the vertical direction and DCT in the horizontal direction. And decryption. Conversely, the prediction block of C performs left and right linear interpolation prediction, so that the vertical direction is D
CT uses DST in the horizontal direction. For the prediction block of A, DCT and DST are used according to the prediction method.

[0025]

According to the present invention, a block separated from a coded block is independently coded, and blocks interposed between the coded block and the coded block are interpolated and predicted by them.
By encoding the residual, interpolation prediction is performed on the interposed blocks, and the prediction efficiency is improved. Furthermore, for a block sandwiched between the encoded block and the interpolated predicted block, in the vertical and horizontal directions, the local decoded image of the interpolated and predicted block is used to perform vertical and horizontal interpolation prediction. Can be performed, and an interpolation prediction signal corresponding to each pixel of the prediction block can be generated using the interpolation prediction signals having different two prediction directions, so that the prediction efficiency can be improved.

As a result, the amount of code generated in the encoder is reduced, and the coding efficiency is improved. When the encoder is controlled so that the transfer rate is constant, the reproduced image quality is improved. Similarly, in encoding and decoding, DC
By using T and DST, the coding efficiency is improved. On the other hand, although inter-block prediction processing is performed, error propagation stops at adjacent blocks, so that there is little problem that a code error in a transmission path affects reproduction image quality.

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration of an embodiment of an inter-block predictive encoding apparatus according to the present invention.

FIG. 2 is a diagram showing a configuration of an embodiment of an inter-block predictive decoding apparatus according to the present invention.

FIG. 3 is a diagram showing a state of processing according to the present invention.

FIG. 4 is a diagram illustrating a state of interpolation prediction according to the present invention.

FIG. 5 is a diagram illustrating a configuration of a conventional inter-block predictive encoding device.

FIG. 6 is a diagram showing a configuration of a conventional inter-block prediction decoding apparatus.

FIG. 7 is a diagram showing a state of processing in a conventional example.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Image input terminal 2,51 Block converter 3,5,10,16,19 Switch 4 Prediction subtractor 6,52 Encoder 7 Code output terminal 8 Prediction controller 9,11 Block memory 12,14 Multiplier 13, Reference Signs List 17 adder 15, 54 block line memory 18, 22, 55 decoder 20 prediction signal generator 21 code input terminal 23, 61 raster converter 24 image output (terminal) 53 block predictor

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁶ , DB name) H04N 7/24-7/68 H04N 1/41-1/419

Claims (8)

(57) [Claims] An inter-block interpolation coding apparatus that divides an image into blocks and sequentially performs coding processing, wherein an independent block separated from a coded block is independently coded within a block to obtain a code. First encoding means for locally decoding a code to obtain a decoded image; and encoding the code for each pixel in a first prediction block vertically or horizontally sandwiched between the encoded block and the independent block. First prediction means for generating a first interpolated prediction signal from a decoded block and a decoded image of the independent block, and subtracting the first interpolated prediction signal from each pixel in the first prediction block Second encoding means for encoding the prediction residual signal obtained as described above to obtain a code, and locally decoding the code to obtain a decoded image; and the encoded block and the first prediction block. For each pixel in the second prediction block sandwiched by the upper, lower, left and right, the coded block and the decoded image of the first prediction block existing in the vertical and horizontal directions of the second prediction block A second prediction means for generating a second interpolated prediction signal from the second prediction block, and encoding a prediction residual signal obtained by subtracting the second interpolated prediction signal from each pixel in the second prediction block Third
Interpolating predictive encoding apparatus, characterized by comprising encoding means of (1). 2. An inter-block interpolation predictive coding apparatus according to claim 1, wherein the vertically interpolated block is a DST (discrete sign) in the vertical direction according to the interpolative prediction direction of the predicted block. Transform), DCT (discrete cosine transform) in the horizontal direction, and encoding and decoding of blocks interpolated and predicted in the horizontal direction are performed using DST in the horizontal direction and DCT in the vertical direction. Inter prediction coding apparatus. 3. An inter-block interpolation predictive decoding apparatus which divides an image into blocks and sequentially performs a decoding process, wherein a code of a decoded block and a code of a separated independent block are decoded independently in the block. A first decoding unit for obtaining an image, the decoded block and the independent block,
For each pixel in a first prediction block sandwiched vertically or horizontally, a first interpolation prediction signal generating a first interpolation prediction signal from the decoded image of the decoded block and the decoded image of the independent block
And a decoding means for decoding the prediction residual signal of the first prediction block to obtain a decoded prediction residual signal, wherein the decoded prediction residual signal and the first
A second decoding unit that obtains a decoded image by adding the interpolated prediction signals of the following: each pixel in the second prediction block sandwiched by the decoded block and the first prediction block A second prediction unit that generates a second interpolation prediction signal from the decoded block existing in the vertical direction and the horizontal direction of the second prediction block and a decoded image of the first prediction block. Decoding the prediction residual signal of the second prediction block to obtain a decoded prediction residual signal;
A third decoding unit for obtaining a decoded image by adding the interpolation prediction signal of (1) to (2). 4. A block interpolated prediction decoding apparatus according to claim 3, wherein the vertically interpolated block is a DST (discrete sign) in the vertical direction according to the interpolated prediction direction of the predicted block. Transform), DCT (Discrete Cosine Transform) is used in the horizontal direction, and blocks interpolated and predicted in the horizontal direction are decoded using DST in the horizontal direction and DCT in the vertical direction. A block interpolation prediction decoding device. 5. An inter-block interpolation coding method for dividing an image into blocks and performing a coding process sequentially, wherein independent blocks separated from coded blocks are independently coded within the blocks to obtain a first code. Along with this, the code is locally decoded to obtain a decoded image, and for each pixel in a first prediction block sandwiched vertically or horizontally by the encoded block and the independent block, the encoded block and First from the decoded image of the independent block
And generating a second code by encoding a prediction residual signal obtained by subtracting the first interpolation prediction signal from each pixel in the first prediction block to obtain a second code, A code is locally decoded to obtain a decoded image, and a second image sandwiched vertically and horizontally by the encoded block and the first prediction block is obtained.
For each pixel in the prediction block, a second interpolated prediction signal is obtained from the encoded block present in the vertical and horizontal directions of the second prediction block and the decoded image of the first prediction block. And generating a third code by encoding a prediction residual signal obtained by subtracting the second interpolation prediction signal from each pixel in the second prediction block to obtain a third code. Interpolation prediction coding method. 6. The inter-block interpolation predictive encoding method according to claim 5, wherein the vertically interpolated and predicted block is a DST (discrete signature) in the vertical direction according to the interpolative prediction direction of the predicted block. Transform), DCT (discrete cosine transform) in the horizontal direction, and encoding and decoding of blocks interpolated and predicted in the horizontal direction are performed using DST in the horizontal direction and DCT in the vertical direction. Inter-block predictive encoding method. 7. An inter-block interpolation predictive decoding method for dividing an image into blocks and performing a decoding process sequentially, wherein a code of a decoded block and a separated independent block are decoded independently within the block to obtain a first block. Is obtained, and for each pixel in a first prediction block sandwiched vertically or horizontally by the decoded block and the independent block, from the decoded image of the decoded block and the independent block. Generating a first interpolation prediction signal, decoding a prediction residual signal of the first prediction block to obtain a decoded prediction residual signal, and calculating the decoded prediction residual signal and the first interpolation prediction signal; The second decoded image is added to obtain a second decoded image. For each pixel in the second predicted block sandwiched by the decoded block and the first predicted block, the upper and lower sides of the second predicted block are obtained. One Generating a second interpolated prediction signal from the decoded block present in the horizontal and horizontal directions and the decoded image of the first prediction block, and decoding a prediction residual signal of the second prediction block. Obtain a decoded prediction residual signal,
A method for inter-block interpolation prediction decoding, characterized in that a third decoded image is obtained by adding the decoded prediction residual signal and the second interpolation prediction signal. 8. A block interpolation predictive decoding method according to claim 7, wherein the vertically interpolated and predicted block is a DST (discrete sign) in the vertical direction according to the interpolative prediction direction of the predicted block. Transform), DCT (Discrete Cosine Transform) is used in the horizontal direction, and blocks interpolated and predicted in the horizontal direction are decoded using DST in the horizontal direction and DCT in the vertical direction. Inter-block interpolation prediction decoding method.