JP2710135B2 - Adaptive coding between frames / intra-frame - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Overview] An inter-frame / intra-frame adaptive coding method for switching between inter-frame coding and intra-frame coding when coding and transmitting coefficient data in a coefficient area after orthogonal transformation. The purpose of this study is to group AC coefficients in the orthogonal transform domain corresponding to several frequency bands, and to select between intra-frame coding and inter-frame coding for each group. In the inter-frame / intra-frame adaptive coding system for switching between the inner coding and coding the coefficients in the coefficient region after the orthogonal transformation of the input image data by one of the two coding systems, the coefficient region after the orthogonal transformation Of the coefficients representing the frequency components from DC to high frequency in Group into a plurality of groups corresponding to the respective areas, and for each of the plurality of groups, all the inter-frame coefficient signals and the intra-frame coefficient signals for the coefficients of the AC components in the own group are input, and the The inter-frame coefficient signal is output when the sum of squares of the signal is smaller than the sum of squares of the coefficient signal in all frames. Conversely, the sum of squares of the coefficient signal in all frames is equal to 2 of the inter-frame coefficient signal.
An inter-frame / intra-frame coefficient signal selecting means for outputting the intra-frame coefficient signal when the sum is smaller than the sum of powers, and the inter-frame / intra-frame coefficient signal in response to a control signal from the inter-frame / intra-frame coefficient signal selecting means. When the output of the selection means is an inter-frame coefficient signal, the output is variable-length coded, and when the output is an intra-frame coefficient signal, the output is fixed-length coded and output by variable-length coding means. I do.

[Industrial applications]

The present invention relates to a digital transmission system for image data, and more particularly, to inter-frame / frame switching for switching between inter-frame coding and intra-frame coding when coding and transmitting coefficient data in a coefficient area after orthogonal transformation. The present invention relates to an inner adaptive coding system.

[Conventional technology]

Recently, digitization of image equipment and transmission methods of image digital signals have been widely developed and studied in many application fields such as videophones. Generally, image data has a very large amount of information. For example, if a current television is directly digitized, a transmission capacity of 100 Mb / s is required, and the image data has 1500 times the information amount of the audio. 1 on TV
The screen switches 30 times per second, and frames are switched in 1/30 seconds. Pixels are generally arranged two-dimensionally on one screen corresponding to one frame. Even if the level of one pixel is represented by, for example, 8 bits, the amount of information of the image data for one frame is very large. More.

However, even when frames are switched in 1/30 second on a television, the picture content of each frame often does not change so much. That is, the image has many parts that are entirely stopped, such as the sky and the background, and few parts that change for each frame. Therefore, one method of compressing and encoding image data is an inter-frame encoding method.

In the inter-frame coding method, a difference in level data is obtained between two temporally consecutive frames, for example, for each corresponding pixel, and the difference is subjected to, for example, Huffman coding and transmitted on a transmission path. Assuming that the level for most pixels does not change, the difference for those pixels is zero,
The difference can be represented by one bit. A code for a pixel having a large level difference requires 8 bits or more. By encoding only these differences, the amount of information to be transmitted is greatly compressed. On the other hand, a method of directly encoding data in one frame is called an intra-frame encoding method.

Regardless of whether inter-frame coding or intra-frame coding is used, the digital representation of an image signal basically requires two operations of sampling and quantization. There are two methods for sampling an image, and the first image is represented by a level value for a discrete point array corresponding to pixels in a frame. The second method is XY corresponding to the frame surface
The level value function (image function) defined on the plane is subjected to orthonormal expansion and its expansion coefficient is used as a sample value.

FIG. 8 is an overall block diagram of an image transmission system using an orthonormal transform which is the second system. In FIG. 8, an input image is blocked by 1 and orthogonally transformed by 2, then quantized by 3, is subjected to, for example, Huffman coding by 4, and is sent to the transmission path 5 from the sending side. Then, on the receiving side, the signal from the transmission path 5 is decoded by 6, inverse orthogonally transformed by 7, deblocked by 8, and output as image data.
Here, one block is formed by dividing data of two-dimensional K × K pixels (for one screen) into one block and performing orthogonal transformation instead of a block of about 4 × 4 to 16 × 16 pixels. This is performed to reduce the conversion time by obtaining the conversion coefficient in units. The two types of orthogonal transform include, for example, Hadamard transform, cosine transform, Karhunen, and Loeve transform. In recent years, a discrete expression of the cosine transform is used, and a discrete cosine transform (DCT) described later is often used.

In the orthogonal transform coding method used in FIG. 8, attention is paid to the correlation of the small area in the screen, and the orthogonal transform is performed using the pixels of the small area as a numerical sequence. The resulting transform coefficients correspond to frequency components and represent each component from low to high frequency. In image signals, low-frequency components are generally large in power and high-frequency components are small in power. Therefore, low-frequency components should be transmitted by performing quantization by assigning a large number of bits to high-frequency components and a small number of bits to high-frequency components. The amount of information is reduced.

FIG. 9 is a diagram showing a transform coefficient area after performing orthogonal transformation on pixel data in one block composed of 4 × 4 pixels. In this transform coefficient area, the upper left area 9 generally represents the DC component of the signal, and the coefficients in the other areas all represent the AC component. The AC component corresponds to each frequency from a low frequency to a high frequency, and the frequency becomes higher toward the lower right.

FIG. 10 is a block diagram of a conventional example of an image data encoding system using orthogonal transform. In the figure, an encoding system performs orthogonal transform such as discrete cosine transform on input image data, for example, as an intra-frame coefficient signal output from the orthogonal transform unit 10 and an output of a subtractor 14 described later. And an inter-frame / intra-frame determination unit 11 for determining which input signal is to be output, a quantizer 12 for quantizing the output of the inter-frame / intra-frame determination unit 11, Variable-length encoding unit 13 that performs variable-length encoding on the output of the transformer 12, and the current intra-frame coefficient signal that is the output of the orthogonal transform unit 10 and the output of the frame memory (FM) 15, that is, one frame before the current frame. A computing unit 14 for taking a difference from a coefficient of a corresponding region in a frame of a frame, a frame memo for storing a value of a previous intra-frame coefficient signal for each coefficient region to generate an inter-frame coefficient signal 15 again before the quantization signal output of the quantizer 12, i.e. the inverse quantizer 16 back to the output format of the inter-frame / intra-frame decision unit 11 consists of an adder 17, and 0 insertion portion 18.

In the case of selecting which of an inter-frame signal and an intra-frame signal is to be encoded in the encoding method using the orthogonal transform as shown in FIG. 10, conventionally, the DC component of the block is generally determined for each block. The sum of the squares of the coefficients corresponding to the removed AC components, that is, the sum of the power, and the interframe coefficient, that is, the sum of the powers, which is the sum of the squares of the prediction error, are compared, and the smaller square sum is selected. . This is because, when the orthogonal transform coefficients are subjected to variable-length coding, an increase in the code amount due to a high-frequency component that is a source of noise poses a problem.

In FIG. 10, an intra-frame coefficient signal representing an AC component obtained by removing a DC component from the coefficient of each block in the current frame, and the difference between these intra-frame coefficients and the corresponding coefficients in the previous frame, that is, the frame The inter-coefficient signal and the inter-frame / intra-frame determination unit 11 are input, and the sum of the powers of the two signals is compared. When the sum of the squares of the inter-frame coefficients is smaller than the sum of the squares of the intra-frame coefficients, the inter-frame coefficient signal is output to the quantizer 12 and a control signal indicating that the inter-frame coefficient signal is to be encoded. Is output to the variable-length encoding unit 13. Conversely, if the sum of squares of the intra-frame coefficient signal is smaller, the intra-frame coefficient signal is output to the quantizer 12 and the control signal for indicating the encoding of the intra-frame coefficient signal is variable-length coded. Department
Output to 13.

The frame memory 15, the inverse quantizer 16, the adder 17, and the zero insertion unit 18 in FIG. 10 use the output of the current quantizer 12 to generate an inter-frame coefficient signal for the next frame. Are stored in the frame memory 15. That is, the quantizer 12
Is an intra-frame coefficient signal, the signal is added to the output of the zero insertion unit 18 by the adder 17 via the inverse quantizer 16.

At this time, the output of the frame memory 15 is
That is, although the intra-frame signal for the previous frame is input, the control signal indicating the output of the intra-frame coefficient signal from the inter-frame / intra-frame determination unit 11 causes the 0 insertion unit 18 to set all its outputs to 0 and adders. Output to 17.
Therefore, the addition result of the adder 17 becomes the current intra-frame coefficient itself, and this is stored in the frame memory 15 in preparation for creation of an inter-frame coefficient signal for the next frame.

On the other hand, when the output of the quantizer 12 is an inter-frame coefficient signal, the 0 insertion unit 18 outputs the output of the frame memory 15, that is, the intra-frame coefficient for the previous frame to the adder 17 as it is. The output of the result adder 17 becomes the current intra-frame coefficient.

[Problems to be solved by the invention]

As described above, the selection between the intra-frame coding and the inter-frame coding in the coding method using the orthogonal transform is performed by calculating the sum of the powers of all the frequency components excluding the DC component among the coefficients of each block in the frame. It was done by comparison. However, in practice, the characteristics of each frequency component differ depending on the frequency value, and the effect on the overall resistance data also differs. Therefore, the coefficients for all AC components are grouped together to select between inter-frame coding and intra-frame coding. Is not always appropriate.

SUMMARY OF THE INVENTION It is an object of the present invention to group AC coefficients in an orthogonal exchange area in accordance with some frequency bands, and to select between intra-frame coding and inter-frame coding for each group.

[Means for solving the problem]

FIG. 1 is a block diagram showing the principle of the present invention. The figure switches between inter-frame coding and intra-frame coding, and calculates the coefficients of the coefficient area after orthogonal exchange of input image data by either of these two coding methods. It is a principle block diagram of an adaptive coding system.

In FIG. 1, among a plurality of coefficients representing frequency components from DC to high frequency in a coefficient domain after orthogonal transformation of input image data, a plurality of AC coefficients excluding a DC component are grouped into a plurality of groups corresponding to a plurality of frequency bands. The inter-frame / intra-frame coefficient signal selecting means (20a, 20b,...) And the variable-length coding means (21a,
21b,...) Are provided.

Inter-frame / intra-frame coefficient signal selection means (20a, 20b,
...) Take the sum of squares of all the inter-frame coefficient signals and the intra-frame coefficient signals for the input coefficients of the AC components in the own group, and the sum of the squares of the inter-frame coefficient signals is the intra-frame coefficient. When the sum of the squares of the signals is smaller than the sum of squares, the inter-frame coefficient signal is conveyed. , And simultaneously outputs a control signal indicating which of the inter-frame coefficient signal and the intra-frame coefficient signal has been selected to the variable-length encoding means 21a, 21b,.

The variable length coding means 21a, 21b,...
According to the control signal from 20a, 20b,...
When the output of the intra-frame coefficient signal selection means 20a, 20b,... Is an inter-frame coefficient signal,
When the output is an intra-frame coefficient signal, the output is fixed-length coded, for example, to 8 bits and output to the transmission path.

Here, the division of the frequency band corresponding to the grouping of the AC coefficients of each block in the frame is arbitrary, and for example, it can be divided into three relatively low frequency bands, intermediate frequency bands, and high frequency bands. .

[Action]

In the present invention, for example, a portion excluding the DC coefficient in the orthogonal transform coefficient region shown in FIG. 9 is divided into a plurality of groups corresponding to a plurality of frequency bands. In FIG. 9, the upper leftmost coefficient represents a DC component, and becomes a coefficient representing an AC component having a higher frequency as proceeding to the lower right. Therefore, the coefficient area is grouped according to the frequency band by dividing the coefficient area by, for example, a polygonal line from upper right to lower left.

And inter-frame / intra-frame coefficient signal selecting means 20a, 20b,... Provided for each coefficient group.
, And the variable-length encoding means 21a, 21b,... Encode either the inter-frame coefficient signal or the intra-frame coefficient signal and output it to the transmission path. As described above, according to the present invention, it is possible to select whether to encode the inter-frame coefficient or the intra-frame coefficient for each frequency band.

〔Example〕

FIG. 2 shows an embodiment of a method for dividing an orthogonal transform coefficient area according to the present invention. In the figure, the AC coefficient in the group corresponding to the 8 × 8 pixel is divided into four frequency bands of AC ₁ , AC ₂ , AC ₃ and AC ₄ from the low frequency side by three broken lines from upper right to lower left. Have been.

FIG. 3 is a block diagram showing the overall configuration of an embodiment of an image data encoding system using an inter-frame / intra-frame adaptive encoding method according to the present invention. In the figure, the coefficient region as an output of the DCT unit 22 for performing discrete cosine transform which is a kind of orthogonal transformation, as shown in FIG. 4, the DC component only (DC) low-frequency band AC ₁ and high frequency band It is assumed that AC ₂ is divided into three groups. Here, the discrete cosine transform (DCT) uses a transform matrix created from a cosine function, which requires a multiplier for the operation, and has a drawback in high-speed operation. It is increasingly used for video transmission, video conferencing, quasi-video transmission of signals, and the like.

In FIG. 3, the portion for coding the coefficients of the DC component, the low frequency band, and the high frequency band group has a similar configuration. For example, the portion for coding the DC component (DC) is the conventional example shown in FIG. / Intra-frame judgment section in
In-frame / inter-frame determination unit (A) 23 corresponding to 11,
Variable length encoders corresponding to 13 Quantizers corresponding to 24 and 12
A frame memory (FM) 26 corresponding to 25, 15; a subtractor 27 corresponding to 14; a dequantizer 28 corresponding to 16; a zero adder 30 corresponding to adders 29 and 18 corresponding to 17; I have.

The low frequency band AC ₁ and part of coding the coefficients in the group of high-frequency band AC ₂ configuration is also substantially the same as the configuration of the portion coding the DC component, between the Figure 10 frame /
Quantizers 32, 42, and 13 corresponding to the intra / interframe determination units (B) 31, 41, and 12 corresponding to the intra-frame determination unit 11.
Subtractors corresponding to the variable-length coding units 33, 43, and 14 corresponding to
Frame memories 35 and 45 corresponding to 34, 44 and 15 and dequantizers 36 and 46 corresponding to 16 and adders 37 and 47 corresponding to 17 and zero insertion units 38 and 48 corresponding to 18 .

The intra-frame / inter-frame determination units (B) 31, 41 for the low-frequency band AC ₁ and the high-frequency band AC _{2 correspond} to the first inter-frame / intra-frame coefficient signal selection means 20a, 20b... As in the conventional example, the sum of squares of the intra-frame coefficient signal and the inter-frame coefficient signal with respect to all the AC coefficients in the group is compared, and the quantizer 32, At the same time as outputting the control signal to the variable-length coding units 33 and 43, the control signal indicates which of the inter-frame signal and the intra-frame signal is to be output.

The variable length coding units 33 and 43 output the outputs of the quantizers 32 and 42 according to the control signals from the intra-frame / inter-frame determination units (B) 31 and 41, respectively, when the output is an inter-frame signal. When the output is an intra-frame signal, the signal is subjected to variable-length coding, for example, 8-bit fixed-length coding, and output on a transmission path. Also, for example, a subtractor 34 for the low-frequency band AC ₁ ,
The operations of the frame memory 35, the inverse quantizer 36, the adder 37, and the zero insertion unit 38 are the same as those in the conventional example shown in FIG.

In this embodiment, the determination of the selection between the inter-frame coefficient and the intra-frame coefficient by the intra-frame / inter-frame determination unit (A) 23 for the DC component in FIG. ) We will use a completely different criterion from 31,41. That is, an 8-bit fixed-length encoding is performed on an intra-frame signal, and an inter-frame signal, that is, a difference between coefficients is subjected to variable-length encoding when the absolute value of the difference is smaller than 4, and If it is larger than the inter-frame signal, the intra-frame signal is subjected to 8-bit fixed-length coding.

In general, when the correlation between the image data of two temporally continuous frames is large, the absolute value of the difference becomes small because the coefficient obtained by orthogonally exchanging the image data has a small difference between the two frames. Therefore, the absolute value can be transmitted as a variable length code having a short code length, and the amount of information is compressed. However, in this case, since the code has a variable length, there is a problem that the necessary code length increases rapidly with an increase in the absolute value of the difference in order to make the data division legible on the receiving side.

Therefore, if the code length of the inter-frame signal, that is, the code length for the absolute value of the difference is longer than the code length of the intra-frame signal, the advantage of using the inter-frame signal is lost. Shall be encoded.

FIG. 5 is an embodiment of an encoding table showing an encoding method for the DC component described above. That is, a fixed-length encoding of 8 bits is performed on an intra-frame signal, and a variable-length encoding of an inter-frame coefficient is performed on an inter-frame signal when the absolute value of the difference is smaller than 4 and the code length is 7 bits or less. But the absolute value of the difference is 5 or more and the corresponding code length is 8
If the number of bits is greater than or equal to the number of bits, the intra-frame signal is encoded.

FIG. 6 is a block diagram showing the configuration of the intra-frame / inter-frame determination section (A) 23 in FIG. In the figure, an intra-frame / inter-frame determination unit (A) is constituted by a read-only memory (ROM) 49 and a selector 50,
49 receives the inter-frame coefficient signal and
The in-frame coefficient signal and the inter-frame coefficient signal are input to the terminal a and the terminal b, respectively, and the output of the selector 50 is output to the quantizer 25.

In the figure, the ROM 49 stores data for outputting an inter-frame coefficient signal, that is, data for outputting a control signal in which the absolute value of the difference is 4 or less and 1 and 5 or more and becomes 0. The control signal is a selector 50 and a variable signal. It is output to the long coding unit 24 and the 0 insertion unit 30, respectively. The selector 50 outputs to the quantizer 25 an intra-frame signal input to the terminal a when the control signal from the ROM 49 is 0 and an inter-frame signal input to the b terminal when the control signal is 1. When the control signal is 0, the 0 insertion unit 30
Is 0, and when the control signal is 1, the output is a DC coefficient value for the previous frame stored in the frame memory 26.

FIG. 7 shows the in-frame / inter-frame determination section (B) 31, 41.
6 is a flowchart of an embodiment of the determination processing of FIG. When the determination processing in FIG starts, first the coefficients of the coefficient region for the group in the frame Somama square sum the sigma ₁ calculated in step 51, 2 the difference coefficients between frames then in step 52 the sum of squares the Σ ₂ is required. The ₁ and sigma ₂ sigma are compared in step 53, the sigma ₂ is the frame signal is used when greater than sigma _1,
In step 54, it is instructed to insert 0 into the 0 insertion units 38, 48, that is, to instruct all the outputs to be 0, and then the process proceeds to step 55. If sigma ₁ is greater than sigma ₂ in step 53, it means that the inter-frame signal is used, the zero inserter unit 38,4
The process proceeds to step 55 without the instruction of inserting 0 into 8 being performed. Then, in step 55, which signal has been selected is notified to the variable-length encoding units 33 and 43, and the process ends.

In the above description, the embodiment using the discrete cosine transform as the orthogonal transform for the image data has been described. However, it goes without saying that the type of the orthogonal transform is not limited to this.

〔The invention's effect〕

As described above in detail, according to the present invention, selection between intra-frame coding and inter-frame coding is performed for each frequency band, and fine determination can be performed for each frequency component.

[Brief description of the drawings]

FIG. 1 is a block diagram showing the principle of the present invention, FIG. 2 is a diagram showing an embodiment of grouping of orthogonal transform coefficient regions after orthogonal transform of image data, and FIG. 3 is an inter-frame / intra-frame adaptive code of the present invention. FIG. 4 is a block diagram showing an overall configuration of an embodiment of an image data encoding system using a coding scheme, FIG. 4 is a diagram showing division of an orthogonal transform coefficient region in the image data encoding system of FIG. 3, and FIG. FIG. 6 is a block diagram showing the configuration of an embodiment of an intra-frame / inter-frame determination unit for a DC component, and FIG. 7 is a block diagram showing the configuration of an inter-frame for an AC low-frequency band and a high-frequency band. FIG. 8 is a flowchart showing an embodiment of a determination process performed by an in-frame determination unit. FIG. 8 is a block diagram showing the overall configuration of an image data transmission system using orthonormal transform. Shows a transform coefficient domain after orthogonal transformation of the data, FIG. 10 is a block diagram showing a conventional example of a configuration of an image data coding system using orthogonal transformation. 22 Discrete cosine transform (DCT) unit 23 Intra-frame / inter-frame determination unit (A) 24,33,43 Variable-length coding unit 26,35,45 Frame memory (FM) , 31,41... Intra-frame / inter-frame determination section (B), 30,38,48... 0 insertion section.

 ──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-1-213085 (JP, A) JP-A-63-232691 (JP, A) JP-A-62-25577 (JP, A) JP-A-62-25 222783 (JP, A) JP-A-63-111791 (JP, A) JP-A-62-247687 (JP, A)

Claims (1)

(57) [Claims] 1. An inter-frame / intra-frame adaptive code for switching between inter-frame encoding and intra-frame encoding and encoding coefficients in a coefficient area after orthogonal transformation of input image data by one of the two encoding methods. In the conversion method, among the coefficients representing the frequency components from DC to high frequency in the coefficient domain after the orthogonal transformation, a plurality of coefficients excluding the coefficient representing the DC component are grouped into a plurality of groups corresponding to a plurality of frequency bands. For each of the plurality of groups, all inter-frame coefficient signals and intra-frame coefficient signals for the coefficients of the AC components in the own group are input, and the sum of squares of all the inter-frame coefficient signals is Outputting the inter-frame coefficient signal when the sum of squares of the coefficient signal is smaller than the sum of squares of the coefficient signal within the entire frame. Ri interframe / intraframe coefficient signal selecting means for outputting a coefficient signal in the frame when small (20a, 20b, · · ·) and, said inter-frame / intra-frame coefficient signal selecting means (20a, 20b,
..), The output of the inter-frame / intra-frame coefficient signal selecting means (20a, 20b,...) Is an inter-frame coefficient signal, and the output is subjected to variable-length coding. And .variable-length coding means (21a, 21b,...) For performing fixed-length coding on the output when the signal is an intra-frame coefficient signal, and outputting the result.