JPH05276506A - Coding and decoding device for moving picture - Google Patents

Abstract

PURPOSE:To obtain high picture quality with less information quantity by selecting in-block prediction coding for coding for a contour of a picture. CONSTITUTION:A block difference signal S2b from a block processing circuit 28 and a discrete cosine transformation signal $3 from a discrete cosine transformation circuit 11 are fed to a coding system changeover discrimination device 3, and when the discrete cosine transformation system is more advantageous, the device 3 generates a coding system switching signal S20 to throw 1st and 2nd selectors 4A, 4B to the position of the discrete cosine transformation coding input terminal (a) thereby allowing a discrete cosine transformation coding processing section 1 to encode the block processing difference signal S2b. On the other hand, when the compression rate for information by the discrete cosine transformation system is disadvantageous, the coding system switching signal S20 is used to throw the 1st and 2nd selectors 4A, 4B to the position of the discrete cosine transformation coding input terminal (b) thereby allowing an in-block prediction coding processing section 2 to execute the coding of the block processing difference signal S2b.

Description

Detailed Description of the Invention

[0001]

[Table of Contents] The present invention will be described in the following order. Field of Industrial Application Conventional Techniques Problems to be Solved by the Invention Means for Solving the Problems Action Example (1) Overall configuration (FIGS. 1 and 2) (2) Discrete cosine transform encoding processing unit ( (FIGS. 3 and 4) (3) Intra-block predictive coding processing unit (FIGS. 5 to 17) (4) Coding method switching decision unit (FIGS. 18 to 28) (5) Variable length coder (FIG. 29, FIG. 29) (FIG. 30) (6) Decoding device (FIG. 31) (7) Drawings necessary for explaining other embodiments (FIGS. 32 to 3)
4) Effects of the invention

[0002]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving picture coding and decoding apparatus, and more particularly to a moving picture signal which can be compressed and transmitted.

[0003]

2. Description of the Related Art Two-dimensional Discrete Cosine Transform (DCT, Di
screte Cosine Transform) coding method and predictive coding method. The Discrete Cosine Transform (DCT) method uses the fact that image signals have two-dimensional correlation, so that the signal power is concentrated on a predetermined frequency component, and the distribution state of the resulting signal power frequency component is a coefficient. It is possible to compress the amount of information by expressing it by encoding. For example, the discrete cosine transform coefficient (DCT coefficient) is distributed so as to concentrate on the low frequency component in the portion where the pattern is flat and the autocorrelation of the moving image signal is high. Therefore, in this case, the information amount can be compressed because the information to be transmitted can be represented only by coding the coefficients distributed so as to concentrate in the low frequency band.

However, according to this discrete cosine transform method, it is attempted to accurately represent a discontinuity of a signal by a DCT coefficient, such as an image signal including a contour (image edge). In this case, since the DCT coefficients are widely dispersed from low frequency components to high frequency components, a very large number of coefficients are required, and there is a problem that coding efficiency drops. As a method for solving this problem, conventionally, a method of roughening the quantization characteristic of the coefficient in order to highly compress and encode the moving image or a method of discarding the coefficient of the high frequency component was used, but the deterioration of the moving image signal It has become noticeable and is still insufficient as a countermeasure. For example, a distortion-like distortion (called corona effect or mosquito noise) occurs around the contour.

On the other hand, the predictive coding system is one in which the quantization characteristic is made coarse by utilizing the property that the luminance discrimination of the eyes is low in the contour portion of the moving image, and a relatively high compression code is used. There is an advantage that can be realized. However, in the case of this predictive coding method, if the quantization characteristic is roughened in the flat part of the moving image, visually noticeable deterioration such as pseudo contours and particle noise is likely to occur. Therefore, there is a problem that predictive coding is not suitable as a high compression means for the flat portion.

Therefore, the discrete cosine transform (D
By compensating for the drawbacks of the CT) method and the predictive coding method, the discrete cosine transform (DCT) method and the intra-block predictive coding method are switched on a block-by-block basis based on the nature of the picture. High compression coding is considered. That is, in the flat block of the image, the discrete cosine transform (D
In contrast to the CT method, intra-block predictive coding (NTC, Non Transform Coding) is applied to the contour portion of the image.
It would be better to use.

[0007]

A problem in the intra-block predictive coding on a block-by-block basis is the block distortion that occurs when coarse quantization is performed. Specifically, as a result of encoding, a phenomenon occurs that each block looks like a mosaic. In intra-block predictive coding, the quantization error caused by coarse quantization appears directly as a change (deterioration) in the brightness level, so that when this change in brightness level appears significantly at the boundary between adjacent blocks, it is visually The result is that the shape of the block stands out like a mosaic.

The present invention has been made in consideration of the above points, and when transmitting compression coded data of a moving image in block units, a moving image code which does not cause block distortion in a restored moving image. An attempt is made to propose an encoding and decoding device.

[0009]

In order to solve such a problem, in the present invention, when one moving image is divided into a plurality of blocks and is coded in block units, discrete cosine transform coding and block coding are performed. Inner predictive coding,
Switch adaptively according to the nature of the design.

First, the Discrete Cosine Transform (DCT) coding method is used in the flat portion of the moving image. At this time, even if coarse quantization is performed by the DCT coding method, smoothness corresponding to the calculation accuracy in the discrete cosine transform can be obtained as the restored moving image. Second, the intra-block predictive coding method is used in the contour portion of the moving image. At this time, in order to reduce the block distortion caused by coarse quantization, the block representative value BASE1 and the quantization width, or the representative value BASE1 and the difference value D between the BASE1 and the representative value BASE2 of another block, Adaptive quantization is performed by transmitting the quantization width Q. The decoding device decodes the image using the switching information, the block representative value BASE1 and the quantization width Q, or the block representative value BASE1, the difference value D and the quantization width Q.

[0011]

According to the present invention, in a moving picture coding apparatus for coding an intra-picture and an inter-picture, a DCT is performed for each unit area (block) for coding an intra-picture and inter-picture signals. Adaptively switch to coding or intra-block predictive coding. Thus, in a moving picture coding apparatus having only conventional Discrete Cosine Transform (DCT) coding, switching information of Discrete Cosine Transform (DCT) coding or intra-block predictive coding (NTC) coding is provided for each coding block. A DCT or NTC can be switched by adding a flag or expanding instruction information indicating an encoding method of a unit area (macro block) and providing an NTC mode. This allows
It is possible to further highly efficiently encode the moving image, and it is possible to decode the moving image with higher image quality based on the highly efficient encoded data.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of a moving picture coding and decoding apparatus according to the present invention will be described in detail below with reference to the drawings.

(1) Overall Configuration In FIG. 1, DV1 indicates an encoding device as a whole,
The digital input moving image is, for example, 8 by a block circuit.
Blocked into a block of × 8 pixels and inputted as a blocked digital input image signal, and in sequence, a Discrete Cosine Transform (DCT) coding processing unit 1, an intra-block prediction coding (NTC) processing unit 2, and a coding method. Processing is performed in the switching determination device 3. That is, the input block digital input image signal DIN is supplied to the predictor 5 and the difference calculation circuit 8, and the difference calculation circuit 8 performs intra-frame coding or inter-frame coding processing to predict the prediction signal of the predictor 5. A differential signal S2 between S1 and the block digital input image signal DIN is obtained, and this is converted into a block differential signal S2b by a block circuit 28, which will be described later, as a discrete cosine transform (DCT) encoding processing unit 1 and an intra-block prediction code. (NTC) processing unit 2.

In this embodiment, the case where the intra-frame coding or inter-frame coding processing is performed will be described. However, instead of this, intra-field coding or inter-field coding processing may be performed. The main point includes the case of performing intra-image (intra-frame or intra-field) encoding or inter-image (inter-frame or inter-field) encoding processing.

In the case of intra-frame coding or inter-frame coding, the block difference circuit 28 determines whether the block difference signal S2b is composed of frame units or field units. , The block differential signal S2b is formed based on the mode. The block mode is determined by the algorithm shown in FIG. 33 when the difference signal S2 is composed of 16 × 16 pixels (referred to as a block or macroblock), and the determination of the block mode is performed based on the instruction shown in FIGS.
Blocking processing is performed on four 8 × 8 pixel sub-blocks to obtain a block difference signal S2b as shown in FIG. Here, the block in frame units means a block in which each sub-block includes both an odd field (hatched portion) and an even field (white portion), as shown in FIG. 34a. On the other hand, the block in the unit of field means a block in which each sub-block is composed of only pixels of odd field or even field, as shown in FIG. 34b. Further, the decision as to whether the block is formed in the frame unit or the field unit in the block circuit 28 is supplied to the variable length encoder 6 described later as a block mode instruction signal S15.

Discrete cosine transform (DCT) encoding processing unit 1 is a discrete cosine transform (DCT).
In the circuit 11, the discrete cosine transform coded signal S3 is obtained by subjecting the block difference signal S2b to the discrete cosine transform, and is converted into the quantized signal S4 by being quantized in the first quantizer 12. Then, the quantized signal S4 is passed through the delay circuit 13 to the first
Discrete Cosine Transform (DCT) of Selector 4A
It is sent to the encoding side input end a.

The selection output S5 of the first selector 4A is converted into the variable length coded signal S6 together with the transmission management information S7 in the variable length coder (VLC) 6, and then the buffer circuit 7 is provided.
The buffer circuit 7 sends out the transmission data DOUT at a transfer rate which is temporarily stored in the buffer circuit 7 and which is compatible with the transfer rate of the transmission system 8 such as a transmission line or a recording device.

The quantized signal S4 of the first quantizer 12 is inversely transformed by the first inverse quantizer 14 and the inverse discrete cosine transform (DCT) circuit 15 to generate the discrete cosine of the second selector 4B. The predicted image of one frame before, which is fed back to the predictor 5 via the transform coding side input terminal a and is thus fed to the buffer circuit 7 in the predictor 5, is reproduced. It is supplied to the difference calculation circuit 8 as the prediction signal S1.

When the predictor 5 generates the prediction signal S1,
Selection is performed by generating management data such as a motion vector, a prediction mode, a calculation method in the difference calculation circuit 8 (that is, intraframe / interframe coding), and supplying this to the variable length encoder 6 as a transmission management signal S7. The variable length coded signal S6 is coded together with the data of the output S5. The Discrete Cosine Transform (DCT) encoding processing unit 1 is, for example, the Institute of Electronics, Information and Communication Engineers, 1987/1 Vol.J.
70-B No. 1 p96-104 The one disclosed in "Intra-frame / inter-frame adaptive extrapolation prediction coding of HDTV signal" or the one described in Japanese Patent Application No. 2-410247 can be applied.

The block differential signal S2b of the block circuit 28 is supplied to the intra-block predictor 21 of the intra-block predictive coding (NTC) processing unit 2, and the intra-block predictor 21 is supplied.
After obtaining the representative value in the block for this block difference signal S2b, the difference signal between the representative value and each pixel value is obtained, and this difference signal is supplied to the second quantizer 22 as the prediction encoded signal S10. It is supplied and converted into a quantized signal S11, which is sent to the scan converter 23.

The scan converter 23 is a quantized signal S1.
After re-converting the array of 1 pixel data so that the data can be further compressed, the converted quantized signal S12 is changed by the differencer 24 via the differencer-side input end c of the switching circuit 27A.
And outputs the differential output S13 to the intra-block predictive coding side input end b of the first selector 4A via the switching circuit 27B. The switching circuits 27A and 27B pass the transformed quantized signal S12 through the differentiator 24 as necessary through the bypass-side input end d, and the intra-block predictive coding (NTC) -side input end b of the first selector 4A. It can be supplied to.

Thus, the compressed image data obtained by intra-block predictive coding of the block difference signal S2b is sent to the buffer circuit 7 via the variable length encoder 6. Here, the delay circuit 13 is used to match the processing time in the discrete cosine transform (DCT) coding processing unit 1 with the processing time in the intra-block predictive coding (NTC) processing unit 2.

The quantized signal S11 of the second quantizer 22 is inversely converted into a prediction difference signal S14 by the second inverse encoder 25 and the intra-inverse block predictor 26, and the second selector 4 is selected.
The predicted image of one frame before represented by the variable length coded signal S6 fed back to the buffer circuit 7 in the predictor 5 is restored through the predictive coding side input terminal b of B. To obtain the prediction signal S1.

In addition to this, the block differential signal S2b of the block circuit 28 and the discrete cosine conversion signal S3 of the discrete cosine conversion circuit 11 are supplied to the coding system switching judging device 3, whereby the pattern in the block is supplied. , The discreet cosine transform is disadvantageous in the compression ratio of the information to generate the coding system switching signal S20, and when the discrete cosine transform system is advantageous, the first and second By switching the selectors 4A and 4B to the input terminal a on the discrete cosine transform coding side, the discrete cosine transform coding processing unit 1 causes the block differential signal S2
Encode b.

On the other hand, when the compression rate of the information by the discrete cosine transform method is unfavorable, the coding method switching judging device 3 uses the coding method switching signal S20 as the first signal.
By switching the second selectors 4A and 4B to the side of the intra-block predictive coding side input terminal b, the intra-block predictive coding processing section 2 executes coding of the block differential signal S2b.

When generating the coding method switching signal S20, the coding method switching determiner 3 generates a coding method switching signal representing the currently selected coding method, which is used as a transmission management signal S7 in a variable length code. Supply to the rectifier 6.

The transmission data DOUT thus supplied from the encoding device DV1 to the transmission system 8 is taken into the buffer circuit 31 of the decoding device DV2 shown in FIG. 2 and inversely encoded by the inverse variable length encoder 32. And the decoded quantized signal S
21, management signal (encoding system switching signal S24, prediction mode signal S25, blocking mode signal S26, etc.)
Are separated. Thus, the decoded quantized signal S21 of the transmission data DOUT is passed through the delay circuit 33 to the inverse quantizer 34.
After being inversely quantized at, the inverse discrete cosine transform circuit 35 performs inverse discrete cosine transform to restore the difference signal S22, and the difference signal S22 is restored via the inverse discrete cosine transform side input terminal a of the switching circuit 40. It is supplied to the addition / restoration circuit 41. In the encoder of FIG. 1, the quantizer 12 and the quantizer 22 can be shared if their characteristics are the same. Similarly, the inverse quantizer 1
4 and the inverse quantizer 22 can be shared if their characteristics are the same.

Further, the decoded quantized signal S21 is transferred to the switching circuit 42.
Inverse differencer 36 receives the inverse differencer through the differencer-side input end c of A and 42B and is given to inverse scan converter 37, or directly reverses through bypass-side input end d of switching circuits 42A and 42B. Sukiyan Converter 3
7 is supplied. Thus, the decoded quantized signal S21 is restored to the original scan order in the inverse scan converter 37, and subsequently inversely transformed in the inverse quantizer 38 and the intra-block predictor 39, and the differential signal S2 thus restored is restored.
3 is supplied to the addition / restoration circuit 41 via the intra-block predictive coding side input terminal b of the switching circuit 40. In addition, in the decoder of FIG. 2, the inverse scan converter 37 and the inverse quantizer 38
Can change their order.

In the switching circuit 40, of the management signals separated by the inverse variable length encoder 32, the coding system switching signal S is selected.
24 is provided, whereby the switching circuit 40 is switched to the input end a or b side in accordance with the coding system of the decoded quantized signal S21 which is currently transmitted. The output of the switching circuit 40 is supplied to the reverse block circuit 44. Of the management signals separated by the inverse variable length encoder 32, the inverse block circuit 44 is supplied with the blocking mode signal S26, whereby the currently transmitted difference signal S22 or S23 is blocked. Reconstructed.

The addition restoration circuit 41 is an inverse block circuit 44.
And the prediction signal S24 obtained by the predictor 43 are added, and the addition output is sent as the restored data DOUTX. The predictor 43 receives the motion vector and the prediction mode signal S25 among the management signals separated by the inverse variable-length encoder 32, and reproduces the image data of one frame before transmitted based on the restored data DOUTX. This is applied to the addition / restoration circuit 41 as a prediction signal S24, whereby the restoration data DOUTX representing one frame of image data is obtained based on the output signal from the inverse block circuit 44 which is currently transmitted. ing. In the decoder of FIG. 2, the inverse quantizer 34 and the inverse quantizer 38 can be shared if their characteristics are the same.

The encoding device DV1 (FIG. 1) and the decoding device DV2 (FIG. 2) having the above configurations have detailed configurations as described below.

(2) Discrete Cosine Transform Coding Processor 1 First, the discrete cosine transform circuit 11 of the discrete cosine transform coding processor 1 of the coding device DV1 (FIG. 1) receives the input block differential signal. When the change in S2b (and hence the change in luminance) is smooth, the value of the discrete cosine conversion signal S3 (that is, DC, as is generally known for the discrete cosine conversion method).
As the T coefficient value), the compression process is executed using the characteristics of the two-dimensional image in which large values tend to concentrate around the direct current (DC) coefficient.

For example, in FIG. 3, one block (8
For the original image K1 of (× 8 pixels), when the information amount of each pixel is schematically shown by the numbers 0 to 100, the original image K1 has a smooth luminance change from the upper left corner to the lower right corner, Each pixel has a brightness level of 30 to 100. When the discrete cosine transform circuit 11 performs the discrete cosine transform (DCT) process on the original image K1, the DCT coefficient in the block becomes almost 0 in the discrete cosine transform signal S3 as shown as a transformed image K2. DCT having a value other than 0
The coefficients lie on the diagonal from the upper left corner to the lower right corner.

Next, the DCT coefficient of this converted image K2 is set to the first
When the quantizer 12 of FIG. 2 quantizes the quantizer with the quantizing width Q = 10 transmitted from the buffer circuit 7 with respect to the buffer remaining amount, as shown in the quantized image K3, the quantized signal S
At 4, most of the quantized DCT coefficients become 0, and only large quantized DCT coefficients remain. This quantized image K3
If the quantized DCT coefficients are sequentially called according to the coefficient calling route K4 (the numbers indicate the calling order), the quantized DCT
A coefficient sequence “45-0-0-4-13-4-0-0 ...” Is obtained, and this is given to the variable length encoder (VLC) 6 via the delay circuit 13 and the first selector 4A. As a result, it is possible to perform an even higher efficiency coding process using a variable-length coding system such as Huffman code (two-dimensional coding in this embodiment).

In the coefficient calling route K4 in the case of FIG. 3, considering that pixels are correlated in the two-dimensional direction, the coefficients are called in a zigzag manner in an obliquely lower right direction from the coefficient calling start point. It was something that was called. On the other hand, in the case of a pattern having a strong coefficient correlation in the horizontal direction, as shown in FIG. 4, the coefficients in the upper row (0 to 7) of the block are sequentially called, and subsequently (8 to 15) and (16 to 23). You can also call them in a zigzag order in the vertical direction.

(3) Intra-block predictive coding processing unit 2 Next, for the intra-block predictor 21 of the intra-block predictive coding processing unit 2, for example, in FIG.
As shown in (A)), when the image data for one block (8 × 8 pixels) having the information amount of the picture including the edge at the lower left corner is given as the block difference signal S2b, the intra-block predictor 21 First, for example, the average value of the flat area is obtained as a representative value "BASE" of the block, and then the difference between the representative value BASE and each pixel value in the block is obtained.

In the case of the original image K11 of FIG. 5, the intra-block predictor 21 obtains BASE = 198 as the representative value BASE, and calculates the difference value between the representative value BASE and each pixel value in the block to calculate the predicted image. Find K12 (Fig. 5 (B)),
This is given to the quantizer 22 as the prediction signal S10. The quantizer 22 quantizes the difference value of the prediction signal S10 with an adaptive quantization width, for example, a quantization width Q = 12, to obtain a quantized coefficient distribution K13 (FIG. 5C). In the case of this embodiment, the fractional part is truncated in the quantization operation (equivalent to a quantizer having a dead zone of 6).

Intra-block predictive coding (NTC) processing unit 2
The adaptive quantization method in (1) selects and applies one or more of the following four adaptive quantization methods for each block.

The first adaptive quantization method is a method using an average value of image signals in a block as a representative value. That is, this method is, as shown in FIG.
For the one-dimensional digital original signal SG corresponding to the block length TBL, the average value M of the amplitudes (X1 to X2) of all the pixels in the block is obtained, and then the average value M and the signal level of each pixel are obtained as shown in FIG. Quantize the difference with L (ie amplitude X). At this time, the quantization width Q is a value output from the buffer circuit 7 based on the remaining amount of data, and the quantization code L
q is

[Equation 1] become. The restored value LX is

[Equation 2] Can be asked as In the first adaptive quantization method, the restoration value distortion increases when the quantization width Q is large, and as a result, as shown in FIG. 8, the restoration value LX for the original signal SG.
Has a disadvantage in that a discontinuous portion UC is formed for each block length, and thus block distortion occurs.

The second adaptive quantization method is the adaptive dynamic range coding method (ADRC, Adaptive Dynamic R).
ange Coding). This method is described in "Study on Quantization Method for Adaptive Dynamic Range Codes" Kondo et al., 1989, 4th Image Coding Symposium (PCS
J) Use the one disclosed in Material (4-3). This adaptive dynamic range coding method (ADRC) is characterized in that the minimum value in a block is used as a representative value, and in this way, the minimum value is often in the peripheral portion of the block. Because. That is, since the normal block is a small area of about (8 × 8) pixels, the possibility that the brightness level is concave is extremely low. Therefore, the minimum value of a certain block is often close to the minimum value of any of the surrounding blocks.

Therefore, as shown in FIG. 9, when the minimum value is in the peripheral portion of the block, continuity is maintained between at least one (in this case, the start side of the block length) peripheral block, and therefore The original signal S between the other peripheral block
Discontinuity UC due to the deviation between G and the restoration value LX
Even if the above occurs, the block distortion can be minimized as a whole. Further, in this adaptive dynamic range coding method, as shown as the one-dimensional coding characteristic in FIG. 10, a new maximum is obtained by using the average value of the signal values included in the highest and lowest gradation levels. After redefining the value MAXX and the minimum value MINX, the quantized code Lq may be obtained, thus making it less susceptible to the influence of noise and isolated points (Japanese Patent Laid-Open No. 2-314910).

The third adaptive quantization method is a method using the edge matching quantization method. First, the case where the digital original signal SG is one-dimensional will be described. The third adaptive quantization method is based on the block length T as shown in FIGS.
For the one-dimensional digital original signal SG of BL, signal values X1 and X2 at both ends of the block (X1 ≦ X2 for simplification)
Error) E for which restoration values LX1 and LX2 of
The quantization width output from the buffer circuit 7 is changed by the following algorithm so that it is output below x.

That is, assuming that the signal level of the pixel in the block is L and the tolerance of the restoration value of the signal values X1 and X2 at both ends of the block is Ex, the signal values X1 and X2 at both ends of the block are shown.
The difference value D of

[Equation 3] And the quantization width Q and the signal value X1 are

[Equation 4] And, and

[Equation 5] Then, the quantization width Q instructed from the buffer circuit 7 is
The intra-block predictor 21 changes the quantization width q to the quantizer 22, and the signal value X1 remains unchanged. However, the quantization width q is for all pairs of the quantization width Q and the difference value D,

[Equation 6] A value larger than the quantization width q that satisfies the above is obtained and written in a read-only memory (ROM) as a table.

In addition,

[Equation 7] And, and

[Equation 8] If so, the quantization width Q and the signal value X1 are left unchanged.
Also

[0045]

[Equation 9] If so, leave the quantization width Q unchanged,

[Equation 10] age,

[Equation 11] And

The quantization code Lq is

[Equation 12] And the restoration value LX is

[Equation 13] Becomes In this method, the error of the restored signal at both ends of the block can be suppressed to Ex or less, so that the continuity between the blocks can be more easily maintained.

In the above description, the third adaptive quantization method is 1
Although the case where the method is applied to the two-dimensional digital original signal SG is described, when the one-dimensional method is extended to the two-dimensional block, the signal values X1 and X2 are respectively represented by the representative value BASE1.
And BASE2, the block signal is quantized and dequantized in the same manner as in the one-dimensional case. Although a quantizer having no dead zone has been described here, a quantizer having a dead zone can also be used.

The fourth adaptive quantization method is a method using the second edge matching quantization method. First, the case where the digital original signal SG is one-dimensional will be described. This 4th
In the adaptive quantization method, the signal values X1 and X2 at both ends of the block (for simplicity, X1 ≦ X2) are directly restored values for the one-dimensional digital original signal SG having the block length TBL as shown in FIG. The decoded value is changed on the decoder side so as to be output as. When the signal level of the pixel in the block is X and the quantization width output from the buffer circuit 7 is Q, the quantization code Lq is

[Equation 14] Becomes

In the encoding device DV1 (FIG. 1), in addition to the quantization code Lq, the signal value X1, the difference value D between the signal values X1 and X2, and the quantization width Q are transmitted to the decoding device DV2. To do. The decoding device DV2 receives the signal values X1 and X2 and the quantization width Q as the quantization parameter, and first, the signal value X2.
The quantized value X2q of is calculated by the following equation.

[Equation 15] After that, if the quantized code Lq is equal to X2q, the reconstructed value LX is

[Equation 16] Otherwise,

[Equation 17] Is restored to the restoration value LX.

The second edge-matching quantization method has a simpler algorithm than the first edge-matching quantization method described above, and RO for changing the quantization width Q is used.
Since the M table is not required, the configuration can be simplified. When the second edge matching quantization method is extended to a two-dimensional block signal, the representative values BASE and BASE 'are used for the signal values X1 and X2, respectively, and the block signal is quantized in the same manner as in the one-dimensional case. And dequantize.
In addition, here, a dead zone is used as a quantizer.
In the above description, a quantizer with a dead zone can be used.

The functions of the quantizer and the inverse quantizer explained in the above-mentioned first to fourth adaptive quantizing methods are the discrete cosine transform except that the block representative value is subtracted or added. It has the same function as that widely used in (DCT) coding. Therefore, the block representative value subtraction or addition processing is performed by the quantizer,
Independent of the inverse quantizer, the intra-block predictive coding (NTC) processing unit 2 and the discrete cosine transform (DCT) processing unit 1 are configured to share the quantizer and the inverse quantizer. May be.

Next, the scan converter 23 (FIG. 1) adaptively calls the coefficient in accordance with the quantized coefficient distribution K13 (FIG. 5C) represented by the quantized signal S11. In the case of this embodiment, the quantization coefficient distribution K13
The quantized coefficient of is the coefficient calling route K14 (FIG. 5C).
Called in zigzag order horizontally along the. Thus, the transformed quantized signal S12 obtained by rearranging the arrangement order of the quantized coefficients is output from the scan converter 23.

Intra-block predictive coding (NTC) processing unit 2
The quantized block signal provided to the Sukyan converter 23 is
Rearranged into dimensions. An EOB (End Of Block) method is applied as a method of transmitting the quantized block signal. This means that the scan-converted signal is viewed in that order, and if, for a certain point, the coefficient of zero value continues to the last coefficient, those zeros are sent only by the code "EOB" (Fig. 5 (E)). Therefore, if the data scan path that can transmit the "EOB" code is selected as early as possible, the information amount can be encoded with a high compression rate. In the case of this embodiment, FIG.
4 types of data scan path DS shown in (A) to (D)
P1 to DSP4 are prepared.

One of the first to fourth data scan paths DSP1 to DSP4 is adaptively selected according to the edge shape in the block. FIG. 15 shows the exact definition of the algorithm in C. Four areas “PIXEL_AREA1” and “PIXEL_AREA2” shown in FIG.
, "PIXEL_AREA3" and "PIXEL AREA4", the absolute value sum of the quantized block signals is calculated, the region having the maximum sum is detected, and the data scan path is selected according to this result.

Finally, through the switching circuit 27A (FIG. 1), the differentiator 24 sequentially forms the rearranged data from the beginning of the data into the difference value between the adjacent data, and the difference distribution K
15 (FIG. 5 (D)) is obtained. As shown in FIG. 16, the differentiator 24 delays the converted quantized signal S12 by one processing time in the delay circuit 24A and subtracts it from the converted quantized signal S12 in the subtraction circuit 24B to form a differential distribution K15. , And outputs the subtraction output as a difference output S13.

The coefficient after the scan conversion may still have a large amount of signal autocorrelation. Yotsutte
There is a possibility that the amount of information can be further compressed by performing the difference processing in the subsequent stage. The differentiator 24 takes the difference between the signal Yi after the scan conversion by the scan converter 23 and the pixel value one pixel before, and outputs the difference as the prediction error signal Ei.

[Equation 18] To get

The prediction error signal Ei can take values from -255 to +255 when the input signal is 8 bits. Therefore, if the signal format is to be sent as it is, 9 bits are required, and an extra code is required for each pixel. However, it is known that the prediction error signal Ei is mostly centered around zero and before and after that. Therefore, not all the signals are represented by 9 bits, but by assigning a code of a short bit length to a large number of appearing signal values, it is possible to input the original 8 bits as well as the average of 9 bits for the entire block. The block signal can be represented by a bit length much shorter than the signal.

As an on / off determination method of the differencer 24, when the difference processing is actually performed and as a result, the number of non-zero signals decreases, it is turned on, that is, the switch circuits 27A and 27B are set to the difference. Switch to the rectifier 24 side. Otherwise, it is off, that is, the switch circuit 27.
Switch A and 27B to the bypass side.

In this way, the coded difference data obtained in the intra-block predictive coding processing unit 2 is coded by a variable length code (VLC) 6 by a variable length code such as Huffman (as described above in the case of this embodiment). Two-dimensionally encoded into the Huffman code sequence K16 (FIG. 5 (E)).
Can be formed and thus highly efficient coded. Thus, the transmission data DOUT transmitted on the basis of the Huffman coded sequence K16 (FIG. 5 (E)) obtained in the encoding device DV1 is the inverse variable length encoder 3 of the decoding device DV2.
2. In the inverse differencer 36, the inverse scan converter 37, and the inverse quantizer 38, the inverse quantization coefficient distribution K18 (see FIG.
After being inversely transformed into (F)), the representative value BASE1 is added in the predictor 43 and the addition / restoration circuit 41 to restore the restored image K19 (FIG. 5 (G)).

When the differentiator 24 (FIG. 1) is unnecessary,
The converted quantized signal S12 is bypassed by switching the switching circuits 27A and 27B to the bypass side output end d side. In this case, as shown in FIG. 17 corresponding to FIG. 5, based on the original image K11 (FIG. 17A), the sequentially predicted image K12 (FIG. 17B) and the quantized distribution K13 (FIG. 17 ( C)), the coefficient calling route K14 (FIG. 17) is obtained.
(C)), the coefficient distribution K1 is obtained by rearranging the quantized values of the quantized distribution K13.
7 (FIG. 17 (D)) is sent to the variable length encoder 6 to obtain a Huffman code string K16 (FIG. 17 (E)). Also in this case, in the decoding device DV2, the inverse quantized coefficient distribution K20 (FIG. 17 (F)) and the restored image K21 (FIG. 17).
(G)) is restored.

(4) Coding method switching deciding device 3 The coding method switching deciding device 3 (FIG. 1) uses a discrete cosine transform (DCT) coding method when coding a moving image signal in block units. Or the intra-block predictive coding (NTC) method. The coding method switching decision unit 3 decides which coding method to select based on the image information in the block according to the spatial domain and / or the discrete cosine transform (DCT) output domain.

The first coding method judging method is a method for judging in the spatial domain, and in the case of a picture whose brightness changes abruptly (specifically, an image including a contour portion and a detail portion),
The dynamic range (DR = maximum value-minimum value) of the image signal in the block takes a large value. For such pictures, the Discrete Cosine Transform (DCT) is disadvantageous in the compression rate of the information and therefore intra-block predictive coding (NTC) should be selected. At this time, the coding method determiner 3 obtains the dynamic range (DR) in each block for each block, and if there is a block whose value exceeds an appropriate threshold value THA selected from the compression rate and the deterioration of the pattern, For this, intra-block predictive coding (N
The TC) processing unit 2 determines that encoding should be performed.

The second coding method judging method is a method of judging in the Discrete Cosine Transform (DCT) output region, and the Discrete Cosine Transform coefficient when the moving image signal is two-dimensional Discrete Cosine Transform is, for example, For the two-dimensional Discrete Cosine Transform with (8 × 8) pixels as a block (macroblock), the coefficient F (0,0) at the 0th row and 0th column corresponding to the upper left corner of the block is the average luminance in the image block. Corresponding to the direct current component, the coefficient represents the high frequency component of the vertical stripes in the image block as it goes from the coefficient F (0,0) to the right and horizontal direction, and the high frequency of the horizontal stripe as it goes downward. Represents an ingredient.

That is, when a discrete cosine transform is performed on a block of a pattern in which the brightness changes abruptly like the contour portion, the converted output is the output of the discrete cosine transform coefficient in the block of (8 × 8) pixels in FIG. As shown as a region, it can be roughly classified into the following three cases. Here, the mark “◯” indicates the position of a pixel with high (or low) luminance, and the mark “x” indicates the position where a large discrete cosine transform coefficient is likely to occur in the block. As shown in FIG. 18 (A), the first pattern has a vertical contour in the block image area K31. At this time, the discrete cosine transform coefficient in the discrete cosine transform output area K32 changes from low to horizontal. DCT coefficients with large energy concentrate so as to spread in the direction. This is called "Case 1".

The second pattern is, as shown in FIG.
In the case where there is a horizontal contour in the block image area K41, at this time, the discrete cosine transform output area K4
In Fig. 2, in the discrete cosine transform coefficient, coefficients having large energy in the vertical direction are concentrated from the low order. This is called "Case 2". As shown in FIG. 18C, the third pattern has a contour in a diagonal direction in the block image area K51A or K51B. At this time, the discrete cosine transform coefficient in the discrete cosine transform output area K52 has a low order. From, the coefficients with large energy are concentrated in the diagonal direction. This is called "Case 3".

Therefore, the coding method switching determiner 3 is shown in FIG.
8 (A), (B) and (C), the sum of absolute values Fa of all discrete cosine transform coefficients excluding the DC component and each discrete cosine transform output area K32 of case 1, case 2 and case 3 , K42 and K52
In each block, the absolute value sums F1, F2, and F3 of the discrete cosine transform coefficients in the area indicated by the "x" mark are obtained, and the maximum one among the absolute value sums F1, F2, and F3 is set as Fmax. A block in which the ratio of the maximum absolute value sum Fmax to the absolute value sum Fa exceeds an appropriate threshold value THB selected from the compression rate and the deterioration of the pattern is determined to be encoded by the intra-block prediction encoding processing unit 2. ..

The third coding method judging method is a method which judges by using both the discrete cosine transform output area and the image block space area. In this case, the coding method switch judging device 3 shows the coding method shown in FIG. The coding method is determined by executing the selection processing procedure RT1. That is, when the coding system switching decision unit 3 enters the coding system selection processing procedure RT1 of FIG. 19, in step SP1, the input image block is discriminated based on the discrete cosine conversion signal S3 of the discrete cosine conversion unit 11. The converted conversion output is inspected and the next step S
At P2, it is determined whether or not there is a conversion coefficient (that is, both a small low-frequency coefficient and a large high-frequency coefficient) representing a pattern whose brightness changes abruptly like a contour portion in the block of the discrete cosine conversion output area. To do.

In the case of a pattern in which the brightness changes abruptly, this determination is made by paying attention to the property that the low frequency component to the high frequency component are widely dispersed and generated in the block of the discrete cosine transform output region. This is used for the switching processing of the coding method, and at this time, the coding method switch 3 executes the algorithm shown in FIGS. 20 and 21.

As shown in FIG. 21A, the sum of squares of the 17 coefficients in the low frequency region excluding the DC component in the upper left corner is "low--
"ac_power" and the sum of squares of all discrete cosine transform coefficients excluding the DC component is "all_ac_po".
wer ”, the low_ac_power is the threshold“ LITT
LE_AC_THRESHOLD "or less and low_ac_po
The ratio of wer and all_ac_power is the threshold value “AC_CONC
If it is equal to or more than "ENTRATE_THRESHOLD", it is determined that the input image block should be encoded by the discrete cosine transform method. At this time, the coding method switching determiner 3 executes the coding using the discrete cosine transform processing unit 1 when a negative result is obtained in step SP2 of FIG.

If not, it is determined that the block is a candidate block for intra-block predictive coding (NTC). At this time, the coding system decision unit 3 obtains a positive result at step SP2 in FIG. , And go to the next step SP3. Here, the threshold "LITTLE_AC
"_THRESHOLD", "AC_CONCENTRATE_THRESHOLD" and the setting of the low frequency region can be set to an appropriate region from the viewpoint of the compression rate and the deterioration of the pattern. In step SP3, the coding method determiner 3 jumps to the contour block detection subroutine RT2 shown in FIG. 20 to execute a process for detecting a block including a contour, and then in step SP4, this block determines a contour image. It is determined whether or not it is a block that includes it.

Contour block detection subroutine R of FIG.
Upon entering T2, the coding system decision unit 3 proceeds to step SP11.
Calculate the representative values "BASE1" and "BASE2" of the block in. Here, the representative values “BASE1” and “BASE2” of the block are average values of a flat area in the block. Now, the pixel values of the block consisting of (8 × 8) pixels are arranged in the order of the numbers shown in FIG.
It is stored in. The typical value of the block is shown in Fig. 22.
It is estimated from the 28 pixels shown in FIG. These pixel values are extracted from x 1 in the order of the arrows (numbers indicate the order) shown in FIG. 22 (C) and stored in the array memory t 1 as shown in FIG. 23 (A).

Then, the representative value of the block is calculated by the DCT / NTC determination algorithm shown in FIGS. 24-28 show the exact definition of the algorithm in C language. First, the difference processing between adjacent samples on the array memory t is executed to
The difference value information of is obtained and the flat area is estimated based on this. The flattest area is 8 continuous array memories t
It is defined that the sum of the absolute values of the difference values of is minimum. Then, the representative value of the block is defined as the average value of the eight array memories t 1 in this flattest area. This value
It is called "BASE1" and the sum of absolute values of the difference values of the array memory t 1 in that area is called "sum_abs_diffl". Here, the encoding system switching determiner 3 is the step SP of FIG.
12. At SP13, it is determined whether or not the block has any one of the following states, and if it has even one state, the block is discrete cosine transform coded.

(A) When peak <PEAK THRESHOLD. Here, peak is the value shown in FIG. 29, and its exact definition is in FIG. PEAK THRESHOLD is a threshold value given by the encoder. (B) When BASE1 is invalid. That is, sum_abs_di
When ffl> FLAT_SAD_THRESHOLD. Where FLAT_
SAD_THRESHOLD is a threshold provided by the encoder.

Next, the coding system switching decision unit 3 moves to step SP14 in FIG. 20 and the representative value "BASE" of another block.
2 "according to the method shown in FIG. Typical value BASE
2 is searched from the area directly opposite (opposite) to the area that has obtained the representative value BASE1 (see FIG. 23 (B)). Typical value BA
The calculation method of SE2 is the same as that of the typical value BASE1. The representative value BASE2 does not necessarily have to exist. The representative value BASE2 is invalid and does not exist if any of the following states is present.

(C) sum _abs _diff2> FLAT _SAD
_THRESHOLD (d) ｜ BASE2-BASE1 ｜ ≤ DIFF _BASE_THRESHOL
D Here, the symbol “||” indicates calculation of an absolute value. Also, D
IFF_BASE_THRESHOLD is a threshold value given by the encoder.

Next, the coding method switching determiner 3 checks the size of the flat area in steps SP15 and SP16 of FIG. In this routine, it is determined whether the flat area of the block is large. FIG. 27
And FIG. 28 shows the exact definition of the algorithm in C language. First, each pixel value (x) and BASE in the block
The absolute value "diff_base1" of the difference signal with 1 is calculated.

Next, diff_base1 is the threshold value "DIF
F_BASE_THRESHOLD ", calculate the number of pixels smaller than
Get "count_base1_pixel". Further, if the representative value BASE2 exists, the same calculation is performed for the representative value BASE2 to obtain "count_base2_pixel". And count_base1_pixel and count_base2
The sum of _pixel is the threshold value "COUNT_FLAT_PEXEL_T".
If it is not larger than "HRESHOLD", an affirmative result is obtained in step SP16, so that the coding system decision unit 3 performs the coding process on this block by the discrete cosine transform system.

On the other hand, when an affirmative result is obtained in step SP16, the coding system decision unit 3 terminates the contour block detection subroutine RT2 in step SP17, and the coding system selection processing procedure R of FIG.
Return to T1. At this time, the coding system decision unit 3 makes the step S
By performing the processes of P4, SP5, SP6, and SP7, the coding by the intra-block prediction coding (NTC) system is performed.

In this way, intra-block predictive coding (NT
When encoding by C) method, if count_base1
If _pixel <count _base2 _pixel, the values of the representative values BASE1 and BASE2 are exchanged. When the block to be encoded (macro block) is an intra-coded macro block, the representative value BASE1 is transmitted as an intra-block prediction value (in discrete cosine transform (DCT) encoding). Corresponds to the DC (direct current) value).

On the other hand, the macroblock to be coded is a non-intracoded macroblock.
roblock), the value of the representative value BASE1 is fixed to zero and this is not transmitted. Instead of this, a macroblock for inter-frame signal coding (Non intracoded m
acroblock), the representative value BASE1 may be sent. Therefore, at this time, if the absolute value of the representative value BASE1 is larger than the threshold value "DIFF_BASE_THRESHOLD", this block is the discrete cosine transform (DC
T) Coded.

Thresholds BASE_DISTANCE_ERR, D used in the contour block detection processing in step SP3
IFF_BASE_THRESHOLD, FLAT_SAD _THRESHOLD, PEA
K_THRESHOLD, BASE_DISTANCE_THRESHOLD, COUNT
_FLAT_PIXEL_THRESHOLD is set to an appropriate value from the compression rate and the deterioration of the pattern.

The method of calculating the representative value of the block is, for example,
The representative value "BASE1" obtained by a series of processes in the contour block detection process in step SP3 is adopted as the representative value "BASE" of the block. Also, the typical value "BASE
2 ”is present, this is the block's typical value“ B
If the representative value "BASE2" does not exist, the representative value "BASE '" of the block is within the block in which the absolute value of the difference between the representative value "BASE" and the sample value within the block is maximum. It is a sample value. Instead of this calculation method, the average value or the minimum value of the sample values in the block may be used as the representative value.

(5) Variable Length Coder Variable Length Coder (VLC) 6
Is the discrete cosine transform (DCT) coding / intra-block predictive coding (NTC) switching information and the representative value BAS of blocks used in intra-block predictive coding (NTC).
E and quantization width (or block representative value BASE, difference value between representative value BASE and another block representative value BASE ′, and quantization width Q), scanning order of samples in block, and difference processing The management signal S7 including the switching information as to whether or not to perform it is variable length coded together with the moving image data to be transmitted as follows.

Block-wise Discrete Cosine Transform (DCT) coding is performed by grouping several coding blocks adjacent to each other as a group.
ck), in which the same coding method is applied and transmitted.

In particular, in the case of this embodiment, since the coding method of the block moving image data in the macro block is switched between the discrete cosine transform (DCT) coding and the intra block predictive coding (NTC), the switching is performed. Information and additional information of intra-block predictive coding (NTC) are added.
The method will be described below.

The following is the coding information in the macroblock layer. The first encoding information is a macroblock type (Macroblock_type), which is a VLC code indicating a macroblock encoding method, and indicates whether the macroblock quantization scale or the mask block is an intraframe encoding mode. Whether the information includes the inter-frame coding mode information, the prediction mode in the case of the inter-frame coding mode, the macroblock motion prediction vector, and the intra-block predictive coding (NTC) of the blocks constituting the macroblock This is information regarding whether or not.

The second coded information is the macroblock quantization scale (Quantize_scale), which is basically the quantum in a block that constitutes a macroblock consisting of a VLC code indicating the value of the macroblock quantization scale. Discrete Cosine Transform (D
Quantization of (CT) coefficients or image signals is performed. In a block that executes intra-block predictive coding, the block quantization scale described later can also be used.

The third coding information is a macroblock motion prediction vector (Motion_vector), which is a VLC code indicating a motion prediction vector value when the macroblock is in the interframe coding mode.

The fourth coded information is a coded block pattern (Coded_block_pattern), which is a discrete cosine transform (DCT) coefficient or an image coefficient to be transmitted among the blocks constituting the macroblock. Is a VLC code indicating the position of the block in which is present. It does not exist if the macroblock is in intraframe coding mode.

The fifth coded information is DCT / NTC switching information, which is a VLC representing the position of a block using it when the block coding includes intra-block predictive coding (NTC). It consists of code. Not present if all blocks are Discrete Cosine Transform (DCT) coding. Further, when the coding method is adaptively switched to the discrete cosine transform or the intra-block predictive coding in MB units and the intra-block predictive coding is selected, if all the blocks included in the MB are intra-block predictive coded. If you decide, you don't need this code.

To describe the method of expressing the code, for example, when the macroblock is composed of four blocks as shown in FIG. 30, when the macroblock is in the intraframe coding mode, the code is 4 bits. And each bit represents the encoding switching information of each block that constitutes the macro block.

In addition, when the macroblock is in the interframe coding mode, the code can be represented by 4 bits in the same manner, or a block having a coefficient judged from the above-mentioned "coding block pattern" exists. It can also be represented by a bit number having a length equal to the number (FIG. 30). If each bit is "0", the discrete cosine transform (D
CT), and the case of "1" indicates intra-block predictive coding (NTC). Note that this information can also be represented by a variable length code such as Huffman.

The following are coding information in the block layer. The first block coding information is additional information of intra-block predictive coding (NTC) and is shown below (a),
The three codes (b) and (c) exist in the case of intra-block predictive coding (NTC).

(A) Data scan path type This code is a code indicating the type of the selected data scan path. For example, when the four types of paths DSP1 to DSP4 as described above with reference to FIG. 14 are prepared, the selected path can be expressed by using a 2-bit code.

(B) Flag for whether or not to perform differential processing This flag is a flag for whether or not to perform differential processing after scan conversion, and is one bit data indicating ON or OFF. However, this code is not necessary when it is determined in advance according to the block encoding mode whether or not the differential processing is performed.

(C) Block Quantization Scale This code is VL related to the block quantization scale value.
In the C code, if the block quantization scale is always set to the same value as the above-mentioned "macro block quantization scale",
Alternatively, if the value obtained by substituting the "macroblock quantization scale" into a certain mathematical expression is determined, this code is unnecessary.

For example, the discrete cosine transform coding and the intra-block predictive coding are adaptively switched for each coding block, and the block-wise discrete cosine transform coding / intra-block predictive coding switching information is transmitted. Further, in the intra-block predictive coding, the representative values BASE and BASE ′ in the two blocks are obtained, and BASE,
It can be omitted in the moving picture coding apparatus that transmits the difference value between BASE and BASE ′ and the quantization width to perform adaptive quantization. If not, the code can represent the value of the block quantization scale as it is with a fixed length of, for example, about 7 bits, or it can represent the difference value from the above-mentioned "macroblock quantization scale" by variable length coding. You can also

The second block coded information is an intra-block representative value, which is a VLC code related to the intra-block representative value. If the block representative value is always set to "0",
This code is unnecessary. Otherwise, the representative value in the block (in the case of discrete cosine transform coding,
DC coefficient, BASE for intra-block predictive coding)
Can be represented by a fixed length of, for example, 8 bits, or the in-block representative values can be differentially processed in the order shown by the arrow in FIG. 31, and the obtained differential value can be variable-length coded.

The third block coded information is the difference value between the two block representative values, and the adaptive cosine transform coding and intra-block predictive coding are adaptively switched in coding block units. Of the discrete cosine transform / intra-block predictive coding are transmitted, and in the intra-block predictive coding, representative values BASE and BASE 'in two blocks are obtained to obtain BASE and BASE.
And BASE ', and a quantization width is transmitted to perform adaptive quantization in a moving picture coding apparatus, representative values BASE and BA of two blocks by intra-block predictive coding.
When transmitting the difference value from SE ′, this difference value is set to, for example, 8
It is expressed as a fixed length bit and transmitted.

The fourth block coded information is a coefficient (image information), which is formed by converting scan-converted one-dimensional data into a VLC code. For example, it is transmitted by two-dimensional Huffman coding. This is to transmit a coefficient having a value other than "0" by forming a variable length code with the value and the relative position set. Since the statistical properties of the Discrete Cosine Transform and the intra-block predictive coding signal are different, a reference table for the two-dimensional Huffman coding is prepared for each, and the discrete cosine transform coding / intra-block predictive coding switching signal is prepared. It is possible to further improve the coding efficiency by properly using them according to the above.

(6) Decoding Device The decoding device DV2 temporarily stores the encoded bit stream input in the buffer circuit 31, as shown in FIG. Next, an inverse variable length encoder (inverse VLC) 32 decodes the DCT / NTC switching signal from the encoded bit stream, and according to the information, DCT or NTC for each block.
Select. The delay circuit 33 is for adjusting the time corresponding to the NTC processing. The first inverse quantizer 34 and the inverse discrete cosine transform (DCT) circuit 35 have a configuration complementary to the above-mentioned discrete cosine encoding processing unit 1.

Similarly, an inverse difference circuit 38 (comprising a processing time delay circuit 38A and an adder circuit 38B as shown in FIG. 32), an inverse scan converter 37, a second inverse quantizer 3
8. The intra-block predictor 39 has a configuration complementary to that of the intra-block predictive coding processing unit 2. The predictor 5 reproduces the driving image by the output of the inverse DCT 35 processed for each block or the intra-inverse block predictor 39.

[0103]

As described above, according to the present invention, the coding of the contour portion of an image is switched to the intra-block predictive coding, so that the amount of information is the same as or smaller than that of the discrete cosine transform coding. Thus, it is possible to perform coding with less interference such as mosquito noise, and as a whole, a high image quality can be obtained with a smaller amount of information as compared with a system using only discrete cosine transform coding.

In the intra-block predictive coding, the block representative value and the quantization width (or the block representative value BA
The block distortion of the decoded image can be reduced by transmitting the SE and the representative value BASE and the difference value between the representative value BASE 'of another block and the quantization width) and performing adaptive quantization. Further, the coding efficiency can be further improved by adaptively switching the coefficient calling order after quantization and further performing coefficient difference processing as necessary.

[Brief description of drawings]

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

FIG. 2 is a block diagram showing an embodiment of a decoder according to the present invention.

FIG. 3 is a schematic diagram showing an encoding order of DCT coefficients.

FIG. 4 is a schematic diagram showing a calling order of coefficients.

FIG. 5 is a schematic diagram showing an intra-block predictive coding procedure.

FIG. 6 is a characteristic curve diagram when an average value is used.

FIG. 7 is a characteristic curve diagram showing quantization and decoding when an average value is used.

FIG. 8 is a characteristic curve diagram showing an example of block distortion when an average value is used.

FIG. 9 is a characteristic curve diagram when ADRC is used.

FIG. 10 is a characteristic curve diagram showing an example of block distortion when ADRC is used.

FIG. 11 is a characteristic curve diagram showing a first edge matching quantization method.

FIG. 12 is a characteristic curve diagram showing a first edge matching quantization method.

FIG. 13 is a characteristic curve diagram showing a second edge matching quantization method.

FIG. 14 is a schematic diagram showing an example of a data scan path used in NTC.

FIG. 15 is an algorithm showing an example of selection of a data scan path used in NTC.

FIG. 16 is a connection diagram showing a configuration of a differencer.

FIG. 17 is a schematic diagram showing another intra-block predictive coding procedure.

FIG. 18 is a schematic diagram showing a relationship between an edge and a DCT coefficient.

FIG. 19 is a flow chart showing a coding method selection processing procedure.

FIG. 20 is a flow chart showing a contour block detection subroutine.

FIG. 21 is an algorithm showing an example of calculating the degree of concentration of DCT coefficients.

FIG. 22 is a schematic diagram showing positions of pixels in a block used for estimating a representative value of a block.

FIG. 23 is a schematic diagram showing in-memory data after processing.

FIG. 24 is an algorithm showing a definition in C language of a DCT / NTC determination method.

FIG. 25 is an algorithm showing a definition in C language of a DCT / NTC determination method as a drawing following FIG. 24;

FIG. 26 is an algorithm showing a definition in C language of a DCT / NTC determination method as a drawing following FIG. 25;

FIG. 27 is an algorithm showing a definition in C language of a DCT / NTC determination method as a drawing subsequent to FIG. 26;

28 is an algorithm showing a definition in C language of a DCT / NTC determination method as a drawing following FIG. 27. FIG.

FIG. 29 is a signal waveform diagram showing an example of a block image signal including a contour.

FIG. 30 is a schematic diagram showing an example of transmission of DCT / NTC switching information.

FIG. 31 is a schematic diagram showing an example of transmission of a representative value within a block.

FIG. 32 is a connection diagram showing an inverse differencer.

FIG. 33 is a list showing an algorithm for determining a blocking mode.

FIG. 34 is a diagram showing a macroblock having a frame / field structure.

[Explanation of symbols]

1 Discrete Cosine Transform (DCT) Coding Processor 2 Intra-block Predictive Coding (NTC) Processor 3 Coding Scheme Switching Judgment Device 4A, 4B First and Second Selector 5 Predictor 6 Variable Length Coding Device (VLC) 7 buffer circuit 11 discreet cosine transform (DCT) circuit 12 first quantizer 13 delay circuit 14 first inverse quantizer 15 inverse discrete cosine transform unit 21 intra-block predictor 22 second quantum 23 Sukiyan converter 24 Differentiator 25 Second inverse quantizer 26 Inverse block predictor 31 Buffer circuit 32 Inverse variable length encoder (VLC) 33 Delay circuit 34 First inverse quantizer 35 Inverse disk Rieto cosine converter 36 Inverse difference circuit 37 Inverse scan converter 38 Second inverse quantizer 39 Inverse block predictor 43 Prediction DV1 Encoding device DV2 Decoding device.

Claims (13)

[Claims] 1. A moving picture coding apparatus for compressing and transmitting a moving picture signal by performing intra-picture and inter-picture coding processing, wherein the coding method is a discrete cosine for each coding block. At the same time as adaptively switching to the transform coding system or intra-block predictive coding system, block-wise discrete cosine transform coding / intra-block predictive coding switching information is transmitted together with the coded moving image data, and In intra-prediction coding, in a moving picture coding apparatus that adaptively quantizes by transmitting a block prediction value and quantization width information, the discrete cosine transform coding or intra-block prediction coding is composed of frames. Switch block or a block made up of fields, and switch the switching signal. Be transmitted with sine transform coding or the intra-block predictive coded video data moving picture coding apparatus according to claim. 2. A moving picture coding apparatus for compressing and transmitting a moving picture signal by performing intra-picture and inter-picture coding processing, wherein the coding method is discrete cosine transform coding in coding block units. Block or intra-block predictive coding system, and at the same time, block-wise discrete cosine transform coding / intra-block prediction coding switching information is transmitted together with the coded moving image data, and intra-block prediction is performed. At the time of coding, in a moving picture coding apparatus that adaptively quantizes by transmitting a block prediction value and dynamic range information, the discrete cosine transform coding or intra-block prediction coding is a block composed of frames. Or the block made up of fields, and the switching signal is switched. Video encoding apparatus characterized by transmitting with discrete cosine transform coding or the intra-block predictive coded video data. 3. The moving picture coding apparatus according to claim 1, wherein the coded information is further subjected to differential pulse code modulation (DPCM). 4. A moving picture decoding apparatus for decoding a bit stream of coded moving picture data, wherein a block decoded by an inverse discrete cosine transform and a block decoded by an inverse block predictive coding are provided. In the moving picture decoding device which selects by the transmitted discrete cosine transform coding / block predictive coding switching information and decodes the moving picture from these selected blocks, the transmitted discrete cosine The transform coding / block predictive coding switching information is such that, when coding, whether discrete cosine transform coding or intra-block predictive coding was performed by a block composed of frames,
A moving picture decoding apparatus including a signal indicating whether a block composed of a field has been performed, and decoding a moving picture by the above-mentioned instruction signal. 5. A moving picture coding apparatus for compressing and transmitting a moving picture signal by performing intra-picture and inter-picture coding processing, wherein the coding method is discrete cosine transform coding in coding block units. Block or intra-block predictive coding system, and at the same time, block-wise discrete cosine transform coding / intra-block prediction coding switching information is transmitted together with the coded moving image data, and intra-block prediction is performed. At the time of encoding, in a moving picture coding apparatus for adaptively quantizing by transmitting a block representative value and quantization width information, the discrete cosine transform coding or intra-block predictive coding is a block composed of frames. Or a block composed of fields is used to switch the switching signal. Video encoding apparatus characterized by transmitting together with the moving picture data predicted coding the in-transform coding or block. 6. A moving picture coding apparatus for compressing and transmitting a moving picture signal by performing intra-picture and inter-picture coding processing, wherein the coding method is discrete cosine transform coding in coding block units. Block or intra-block predictive coding system, and at the same time, block-wise discrete cosine transform coding / intra-block prediction coding switching information is transmitted together with the coded moving image data, and intra-block prediction is performed. Adaptive by obtaining the first and second representative values in the block at the time of encoding and transmitting the first representative value, the difference value between the first representative value and the second representative value, and the quantization width information. In a moving picture coding apparatus for quantizing, the discrete cosine transform coding or intra-block predictive coding is performed by a block composed of frames or a file. In switched or performed in configured block, moving picture encoding apparatus characterized by transmitting with its switching signal discrete cosine transform coding or the intra-block predictive coded video data. 7. Quantized information is converted into one-dimensional information by adaptive scanning according to the distribution state in the block, and further, if necessary, the one-dimensional information is subjected to a difference processing between adjacent samples. 7. The moving image code according to claim 5, wherein the switching information indicating whether to perform the scanning order and the difference processing is transmitted together with the encoded moving image data converted into the difference value information. Device. 8. An image coding apparatus for compressing and transmitting a moving image signal by performing intra-picture and inter-picture coding processing, wherein the coding method is discrete cosine transform coding in coding block units. System or intra-block predictive coding system is adaptively switched, and at the same time block-wise discrete cosine transform coding / intra-block predictive coding switching information is transmitted together with the coded moving image data, and the intra-block predictive code is also transmitted. At the time of coding, adaptive quantization is performed by transmitting the representative value of the block and quantization width information, and the quantization information is one-dimensionalized by adaptively scanning the quantization information according to the distribution state in the block. The one-dimensionalized information is converted into difference value information by performing a difference processing between adjacent samples, and the difference value information is shared with the encoded moving image data. , The scanning order and the switching information indicating whether or not to perform the difference processing, and the representative value of the block used in the discrete cosine transform coding / intra-block prediction coding and the quantization width of the sample in the block, and the scanning order. In the moving picture coding apparatus for variable length coding the switching information indicating whether or not the differential processing has been performed, the discrete cosine transform coding or intra-block predictive coding is performed by a block composed of frames. A moving picture coding apparatus, characterized in that it is switched between carrying out or a block composed of fields, and the switching signal is transmitted together with moving picture data which is discretized cosine transform coded or intra-block predictive coded. . 9. An image coding apparatus for compressing and transmitting a moving picture signal by performing intra-picture and inter-picture coding processing, wherein the coding method is discrete cosine transform coding in coding block units. System or intra-block predictive coding system is adaptively switched, and at the same time block-wise discrete cosine transform coding / intra-block predictive coding switching information is transmitted together with the coded moving image data, and the intra-block predictive code is also transmitted. At the time of encoding, the first and second representative values in the block are obtained, and the adaptive quantum is transmitted by transmitting the first representative value, the difference value between the first representative value and the second representative value, and the quantization width information. The quantized information is made into one-dimensional by adaptive scanning according to the distribution state in the block, and further the one-dimensionalized information is subjected to a difference processing between adjacent samples, if necessary The difference information is converted into difference value information, and the switching information indicating the scanning order and whether or not the difference processing is performed is transmitted together with the encoded moving image data, and the discrete cosine transform encoding / intra-block prediction encoding switching is performed. Information, a difference value between the first representative value of the block used in intra-block predictive coding, and the second representative value that is the representative value of the first block and another block, and the intra-block In a moving picture coding apparatus that performs variable length coding on a quantization width of a sample, a scanning order, and switching information as to whether or not to perform difference processing, the discrete cosine transform coding or intra-block prediction coding is performed. , A block composed of frames or a block composed of fields is switched, and the switching signal is converted to discrete cosine transform coding or block. Video encoding apparatus characterized by transmitting together with the inner predictive coded video data. 10. A moving picture decoding apparatus for decoding a bit stream of coded moving picture data, comprising a block converted by the Discrete Cosine Transform and a block coded by intra-block predictive coding. , Select by the transmitted discrete cosine transform / intra-block predictive coding switching signal, and in the case of intra-block predictive coding, further receive the representative value of the block and the quantization width information, and select from these selected blocks. In a moving picture decoding apparatus for decoding a moving picture, the transmitted discrete cosine transform coding / block predictive coding switching information is converted into discrete cosine transform coding or intra-block predictive coding at the time of coding. Was it done with a block composed of frames,
A moving picture decoding apparatus including a signal indicating whether a block composed of a field has been performed, and decoding a moving picture by the above-mentioned instruction signal. 11. A moving picture decoding apparatus for decoding a bit stream of coded moving picture data, comprising a block converted by the Discrete Cosine Transform and a block coded by intra-block predictive coding. , Selected by the transmitted discrete cosine transform / intra-block predictive coding switching signal. In the case of intra-block predictive coding, the first representative value, first representative value and second representative value in the block are further selected. In the moving picture decoding apparatus which receives the difference value of the values and the quantization width information and decodes the moving picture from the selected block, the above-mentioned transmitted discrete cosine transform coding / block predictive coding switching When encoding, the information is discretized cosine transform coding or intra-block predictive coding and is composed of frames. Or it was carried out using the clock,
A moving picture decoding apparatus including a signal indicating whether a block composed of a field has been performed, and decoding a moving picture by the above-mentioned instruction signal. 12. A moving picture decoding apparatus for decoding a bit stream of coded moving picture data, comprising a block converted by the Discrete Cosine Transform and a block coded by intra-block predictive coding. , Select by the transmitted discrete cosine transform / intra-block predictive coding switching signal, and in the case of intra-block predictive coding, the representative value of the block, the quantization width information, and the intra-block quantization information are one-dimensionalized. In the moving picture decoding apparatus, which receives the information indicating whether or not the one-dimensionalized information has been subjected to the difference processing for the adjacent samples, and decodes the moving picture from the selected block. , The transmitted discrete cosine transform coding / block predictive coding switching information is transmitted at the time of coding. , Or were performed by the block composed of the discrete cosine transform coding or the intra-block predictive coding in a frame,
A moving picture decoding apparatus including a signal indicating whether a block composed of a field has been performed, and decoding a moving picture by the above-mentioned instruction signal. 13. A moving picture decoding apparatus for decoding a bit stream of coded moving picture data, comprising: a block converted by the Discrete Cosine Transform and a block coded by intra-block predictive coding. , In the case of intra-block predictive coding, it is further selected by the transmitted discrete cosine transform / intra-block predictive coding switching signal. Further, in the case of intra-block predictive coding, the first representative value, the first representative value and the second representative value in the block are selected. The difference between the values, the quantization width information, the scan order used when the quantization information in the block is one-dimensional, and the 1
In the moving picture decoding apparatus for receiving the information indicating whether the dimensionalized information has been subjected to the difference processing between adjacent samples and decoding the moving picture from the selected block, the transmitted discrete cosine The transform coding / block predictive coding switching information is such that, when coding, whether discrete cosine transform coding or intra-block predictive coding was performed by a block composed of frames,
A moving picture decoding apparatus including a signal indicating whether a block composed of a field has been performed, and decoding a moving picture by the above-mentioned instruction signal.