JPH04299688A - Motion compensation variable length code encoding system - Google Patents

Abstract

PURPOSE:To reduce mismatching of vectors and to efficiently encode them by selecting and using optimum one of plural motion compensation variable length code encoding tables with respect to the efficient encoding system of a moving picture. CONSTITUTION:The efficient encoding system of the moving picture is provided with plural motion compensation code encoding tables 21 to 2n, and the motion vector of the just preceding encoded frame is held in a memory 3. A calculating and discriminating part 4 discriminates the property of the frequency in occurrence of the motion vector in accordance with the held motion vector of the just preceding encoded frame, and plural motion compensation encoding tables 21 to 2n are selected in accordance with the discrimination result by switching parts 51 and 52.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a high-efficiency encoding system for moving pictures, and more particularly to a motion-compensated variable-length encoding system that uses a plurality of motion-compensated variable-length encoding tables.

[0002] In a high-efficiency coding method for moving images, information obtained by coding the difference between an original image and a predicted value and a motion vector obtained by motion estimation are each encoded using variable length coding.
Generate the required variable length encoded output.

[0003] Such a motion vector encoding method is as follows:
It is desired to be able to perform more efficient encoding by selecting an optimal table from a plurality of motion compensation encoding tables.

0004

2. Description of the Related Art FIG. 13 shows a conventional encoding circuit, in which 101 is an encoder for encoding the difference between an original image and a predicted value, and 102 is a variable length encoding circuit for an encoded signal. 103 is a variable length coding (VLC) unit for motion vectors, 104 is a frame memory that stores images for one frame, 105 is a variable delay device, 106 is a motion detection unit that performs motion estimation, and 107 is a subtraction unit , 108 is an adder. The variable delay unit 105 and the motion detection unit 106 form a motion compensation unit 109, and the VLC unit 1 for the encoded signal
02, the VLC unit 103 for motion vectors forms a variable length encoder 110.

[0005] The input signal is input in units of blocks each consisting of a plurality of pixels by block-scanning the original image signal, and is applied to a subtraction section 107 and a motion compensation section 109 . In the subtraction unit 107, the difference (prediction error) between the original image signal and the predicted value signal from the variable delay device 105 is calculated.
seek. Encoder 101 encodes this difference signal and inputs it to variable length encoder 110 and adder 108 . Encoding is performed, for example, by vector quantization, discrete cosine transform (DCT), or the like. In the addition section 108,
A reproduced signal is created by decoding this encoded signal and adding a predicted value signal. This signal is stored in the frame memory 104 to generate a signal of a reproduced image one encoded frame before. On the other hand, in the motion compensation unit 109, the motion detection unit 106 matches the reproduced image of the pre-encoded frame from the frame memory 104 with the original image to detect a motion vector. The detected motion vector is transmitted to a variable length encoder 110;
variable delay unit 105. The variable delay unit 105 delays the image signal of the pre-encoded frame from the frame memory 104 according to the motion vector.
A signal of the aforementioned predicted value is generated.

In the variable length encoder 110, a VLC unit 102 for encoded signals performs variable length coding on the encoded difference signal, and a VLC unit 103 for motion vectors converts the detected motion vector into a variable length code. to produce a variable length encoded output. The VLC unit 103 for motion vectors has a motion compensation encoding table,
By referring to this and outputting a code word corresponding to the input motion estimation result, variable length encoding is performed on the motion vector.

FIG. 14 illustrates motion vector matching, where 11 indicates an original image, and 12 indicates a reproduced image (one coded frame before) in the frame memory.
(previous image) is shown. As a general method for matching motion compensation, when the size of the search range for motion compensation is (+x to -x, +y to -y), block (a, b of the original image indicated by 13) For the previous image 12
The difference between the original image block 13 and the image block in the previous image 12 is calculated within the search range indicated by 14 (a+x to a-x, b+y to b-y) determined by the motion compensation table in , and the cumulative error is calculated. When the block with the smallest value is 15, the vector (k, m) for that position (a+k, b+m) is determined as the motion vector 16. The difference in this case is often calculated using an absolute value error, and sometimes using a squared error.

[0008]

[Problem to be Solved by the Invention] In the conventional encoding circuit shown in FIG. 13, input images have different motion properties, such as an image with little motion or an image with large motion. Even in this case, only one type of variable-length encoding table for motion vectors was used. Therefore, for example, in the case of an image with extremely large motion, the motion compensation range of the encoding table is small relative to the motion, so mismatching is likely to occur, resulting in a problem of deterioration of image quality.

Furthermore, for images with small motion, the motion vector encoding range in the encoding table is wider than necessary, and in the encoding table with a wide motion vector encoding range, the motion vector encoding range corresponding to large motion is Since code words are included, there is a problem in that a code word that is longer than necessary is assigned, resulting in a decrease in encoding efficiency and an impediment to improvement in image quality.

[0010] The present invention aims to solve the problems of the prior art, and provides a high-efficiency encoding system for moving images, which includes a plurality of motion compensation variable length encoding tables and which performs one encoding. By using the motion vector information before the frame to select the optimal encoding table according to the size and nature of the motion and encoding the motion vector, vector mismatching can be reduced and efficient motion can be achieved. It is an object of the present invention to provide a motion-compensated variable-length coding method that can perform vector coding.

[0011]

[Means for Solving the Problems] The present invention variable-length encodes a coded signal in which a difference between a reproduced image one coded frame before and a predicted value is encoded, and a motion detecting unit that encodes a coded signal for one coded frame. In a high-efficiency coding method for moving images in which a motion vector detected by matching the previous reproduced image and the original image is variable-length coded using a motion compensation coding table, multiple motion compensation coding tables and , a memory that holds the motion vector of the previous one encoded frame, a calculation/judgment unit that judges the nature of the occurrence frequency of the motion vector from the held motion vector of the previous one encoded frame, and a plurality of The present invention is characterized in that it includes a switching section for selecting a motion compensation encoding table.

0012

[Operation] In a high-efficiency coding method for moving images, the motion detection unit
Then, by matching the reproduced image one encoded frame before with the original image to detect a motion vector, and variable-length encoding this using a motion compensation encoding table,
Encodes moving images. At this time, in the high-efficiency encoding method for moving images, a plurality of motion compensation encoding tables 21
~2n is provided, the motion vector of the previous one encoded frame is held in the memory 3, and a calculation/judgment unit 4 is provided to determine the frequency of occurrence of the motion vector from the motion vector of the previous one encoded frame held in the memory 3. Generate information by determining the nature of the information. Then, the switching units 51 and 52 select a plurality of motion compensation encoding tables 21 according to this information.
-2n is selected.

In this case, the calculation/determination unit 4 calculates the result of determination as to whether the number of motion vectors that become 0 in one frame exceeds a certain threshold value, and the maximum and minimum values of motion vectors in one screen. Information for selecting a plurality of motion compensation encoding tables 21 to 2n is generated depending on the determination result of whether the value exceeds each threshold value.

[0014] Also, in this case, in the calculation/judgment section 4,
Judgment results as to whether the number of motion vectors that take the maximum or minimum value in one frame exceeds the respective thresholds, and whether the number of motion vectors that take the maximum or minimum value in one screen exceeds a certain threshold. Information for selecting a plurality of motion compensation encoding tables 21 to 2n is generated depending on the determination result.

Furthermore, in this case, the calculation/judgment unit 4 calculates the average value of the motion vectors in one frame, and selects a plurality of motion compensation encoding tables 2 according to this average value.
Generate information for selecting 1 to 2n.

Therefore, according to the present invention, it is possible to select an optimal encoding table from among a plurality of motion compensation encoding tables using information on the motion vector of one encoded frame before, so that vector mismatch can be avoided. reduce the
Efficient encoding can be performed.

[0017]

[Embodiment] FIG. 2 shows the configuration of an embodiment of the present invention, in which the same parts as in FIG. be. Further, 202 and 203 are switches, and 204 is a memory that stores motion vectors for one frame.
Reference numeral 205 is a calculation/determination unit that calculates and determines the magnitude, nature, etc. of motion of a motion vector.

The input signal is input block by block by block scanning the original image signal, and is input to the subtracting unit 1.
07 and the motion compensation unit 109. Subtraction section 107
The difference (prediction error) between the original image signal and the predicted value signal from the variable delay unit 105 is obtained, the encoder 101 encodes this difference signal, and the variable length encoder 110 and the adder 108 Add to. The adder 108 decodes this encoded signal and adds it to the predicted value signal to create a reproduced signal, which is stored in the frame memory 104 to generate a reproduced image signal. In the motion compensation unit 109, the motion detection unit 106 matches the reproduced image of the pre-encoded frame from the frame memory 104 with the original image, detects a motion vector, and sends it to the variable length encoder 110 and variable delay. Add to container 105. The variable delay unit 105 delays the image signal of the previous frame from the frame memory 104 according to the motion vector to generate a predicted value signal. The above operation is similar to the conventional encoding circuit shown in FIG.

In the variable length encoder 110, a VLC unit 102 for the encoded signal performs variable length encoding on the encoded difference signal and outputs the encoded signal. Furthermore, the motion vector detection result in the motion detection unit 106 is stored in the memory 204 as a single
Frames are stored. The calculation/determination unit 205 determines the size, nature, etc. of the motion from the motion vector information of the pre-encoded frame stored in the memory 204, and provides the results to the switches 202 and 203.
By switching to select one of the motion compensation encoding tables 2011 to 201n and feeding back information on the determination result to the motion detection unit 106, the search range at the time of motion detection is calculated and determined according to the determination result. The selection of the motion compensation encoding table is made to match the motion range during the motion detection operation. The motion vector for the original image detected by the motion detection unit 106 is
Variable length encoding is performed using the selected motion compensation encoding table.

[0020] As described above, in the motion compensation variable length encoding method of the present invention, the optimal motion compensation encoding table can be selected according to the motion in the original image, thereby reducing mismatches during motion detection and improving image quality. Therefore, efficient encoding can be performed. In addition, the receiving side uses a motion compensation coding table when decoding the transmitted variable length code, but the receiving side also uses information about the motion vector coded one coded frame ago, just like the transmitting side. If the motion compensation encoding table is selected and decoding is performed, the motion vector information of the previous encoded frame has already been sent to the receiving side, so the motion vector information can be calculated without sending any new information. Since decoding can be performed by selecting a compensation encoding table, the method of the present invention can be applied to the conventional method by simply adding a small amount of circuitry. Hereinafter, for ease of understanding, a simple code word will be explained as an example. In this case, the range of the motion vector is set to ±4 to 7.

FIGS. 3 to 6 show examples (a) to (d) of motion compensation encoding tables according to the present invention, respectively. In FIG. 3, the motion is large but concentrated around 0. Figure 4 shows a case where the movement is small but not concentrated around 0, Figure 5 shows a case where the movement is large but not concentrated around 0, and Figure 6 shows a case where the movement is small and not concentrated around 0. This shows the case where they are concentrated in the vicinity.

FIGS. 7 and 8 are flowcharts (1) and (2) showing an example of the operation in the embodiment shown in FIG. , especially examples regarding determination operations are shown.

First, in step 301, motion vector information VX (u, v), VY (u, v) of one encoded frame before is obtained.
Then, threshold values TH1, TH2, and TH3 for determination are prepared. Here, u and v are the numbers of blocks in one frame. In step 302, initial values are set. Note that * in the figure represents X or Y. MINSUM represents the minimum value of a vector in the X or Y direction in one frame, and is set to an arbitrary value larger than the motion vector range (search range) as an initial value. MAXSUM represents the maximum value of a vector in the X or Y direction in one frame, and is set to an arbitrary value smaller than the motion vector range (search range) as an initial value. COUNT· indicates the number of vectors that take a value of 0 for X or Y in one frame, and the initial value is set to 0. COUNTS indicates the number of motion compensations (number of motion vectors) and has an initial value of 0.

Next, regarding the vectors for one screen, the number of motion vectors (COUNTS) is counted for each of the vector VX in the X direction and the vector VY in the Y direction, and the minimum value (MINSUM・) of the vector is determined. Find the maximum value (MAXSUM・) and calculate the number of vector 0 (CO
UNT·) is counted (step 303). In the next steps 304 to 306, the general tendency of the frequency of motion vector occurrence in vectors VX and VY is estimated from the obtained maximum value, minimum value, and number of vector 0s. If there is no motion vector, such as when intra-frame encoding is forced, the maximum and minimum values are not calculated or counted. Note that the threshold value for determining the number of vectors 0 (COUNT・) in step 304 is the predetermined threshold value TH1・, which is the number of motion vectors generated relative to the total number of blocks (u×v) (
Use the value multiplied by the ratio of COUNTS).

In steps 304-306, a threshold is used to estimate the frequency of occurrence type. For example, if the number of vector 0 exceeds a certain threshold in step 304, and in step 306 the minimum value is smaller than the threshold (TH2.) and the maximum value exceeds the threshold (TH3.), then
This is a case where there are some large vectors that are concentrated around 0. In this case, if the motion compensation encoding table (a) shown in FIG. 3 is selected, the encoding efficiency is good, so the process proceeds to step 307.

If the number of vectors 0 exceeds a certain threshold in step 304, and the maximum value and minimum value do not exceed the threshold in step 306, the vectors are concentrated around 0 and there are no large vectors. It is. In this case, Figure 6
If the motion compensation encoding table (d) shown in (d) is selected, the encoding efficiency is good, so the process proceeds to step 309.

If the number of vectors 0 is below a certain threshold in step 304, and the minimum and maximum values exceed the threshold in step 305, it is determined that there are large vectors that are not concentrated around 0. This is the case. In this case, selecting the motion compensation encoding table (c) shown in FIG. 5 provides better encoding efficiency, so the process proceeds to step 308.

If the number of vectors 0 is less than a certain threshold in step 304, and the minimum and maximum values do not exceed the threshold in step 305, it is determined that there are large vectors that are not concentrated around 0. This is the case if you don't. In this case, selecting the motion compensation encoding table (b) shown in FIG. 4 provides better encoding efficiency, so the process proceeds to step 310.

In the operation examples shown in the flowcharts of FIGS. 7 and 8, the information input to the memory 204 includes the following information:
It must be possible to reach the maximum possible value of 201n.”
This condition is necessary. That is, by selecting the motion compensation encoding table detected one encoding frame before,
If a value can only be taken up to 4 instead of ±7, there will be a limit even if a vector with a larger value actually occurs, and from the next frame onwards, a table with a larger vector will be selected. become unable to do so.

FIGS. 9 and 10 are flowcharts showing other operation examples in the embodiment shown in FIG. An example is shown in which it is possible to select a motion compensation encoding table with a larger vector from the next encoded frame when a vector of , table (b) shown in FIG. 4 or FIG.
or (d).

Steps 401-403 are the same as steps 301-303 in the examples of FIGS. 7 and 8. However, in step 402, regarding the X direction and Y direction,
The number of vectors that take the maximum value within the range of motion vectors (C
MAX・) and the number of vectors that take the minimum value (CMIN・
) is set to 0.

In step 404, the number of vectors that have the maximum value within the range of motion vectors (CMAX・),
Similarly, the number of vectors that have the minimum value (CMIN.) is checked. In step 405, the number of vectors (CMAX・) with the maximum value checked in step 404 exceeds the threshold value (TH2), or the number of vectors with the minimum value (CMIN・) checked in step 404 exceeds the threshold value (TH3). ). Furthermore, in steps 406 and 409, it is checked whether the number of vectors 0 (COUNT·) exceeds a certain threshold value. The threshold value for determining the number of vector 0s in this case is the same as that shown in FIG.

If the number of vectors that take the maximum or minimum value within the range of motion vectors exceeds the threshold in step 405, and the vector values are concentrated around 0 in step 406, the process shown in FIG. Motion compensated encoding table (a
) is selected, the encoding efficiency is good, so step 407
Proceed to .

If the number of vectors that take the maximum or minimum value within the range of motion vectors exceeds the threshold in step 405, and if the vector values are not concentrated around 0 in step 406, the process shown in FIG. Motion compensated encoding table (
Selecting c) provides good encoding efficiency, so step 40
Proceed to 8.

If the number of vectors that take the maximum or minimum value within the range of motion vectors in step 405 is less than the threshold value, and if the vector values are concentrated around 0 in step 409, the process shown in FIG. motion compensation encoding table (d
) is selected for better encoding efficiency, so step 410
Proceed to .

If the number of vectors that take the maximum or minimum value within the range of motion vectors in step 405 is less than the threshold value, and if the vector values are not concentrated around 0 in step 409, the process shown in FIG. Motion compensated encoding table (
If b) is selected, the encoding efficiency is good, so step 41
Go to 1.

In the operation examples shown in the flowcharts of FIGS. 7, 8, 9, and 10, the information held in the memory 204 includes "The motion vectors that occur most frequently are 0 and 0. It is necessary to meet the following conditions: That is, the estimated average value of the motion vector is assumed to be 0, and the frequency of occurrence of 0 is counted to determine the highest frequency. In particular, the motion compensation encoding tables shown in FIGS. 3 and 6 are determined to have the highest number of zeros. However, if panning or the like is taken into consideration, the frequency of occurrence in the panning direction will be highest.

FIG. 11 is a flowchart showing still another example of operation in the embodiment shown in FIG. 2, in which the smallest code word is assigned to the vector with the highest frequency of occurrence. In this example, it is assumed that there is no improvement effect when the occurrence frequency is not very concentrated in the vicinity where the occurrence frequency is high, and a case where the occurrence frequency is concentrated in the vicinity is taken up.

FIG. 12 shows an example (e) of the motion compensation encoding table according to the present invention, which is an example of the motion compensation encoding table when assigning the smallest code word to the vector with the highest frequency of occurrence. It shows.

In FIG. 11, step 501 is the same as in the operation example of FIG. 3 or FIGS. 7 and 8. However, in this case no threshold is required. In step 502, the average value AVR of the vector is calculated, and in step 503, the average value AVR of the vector is calculated.
Variable length encoding is performed by selecting from a plurality of motion compensation encoding tables 2011 to 201n so that α in the motion compensation encoding table shown in FIG.

[0041]

[Effects of the Invention] As explained above, according to the present invention, it is possible to select an optimal encoding table from among a plurality of motion compensation encoding tables using information on the motion vector of one encoded frame before. , vector mismatch can be reduced and efficient encoding can be performed. In addition, since the motion compensation encoding table is selected using the existing information of the motion vector of the previous encoded frame,
On the receiving side, it is possible to select a motion compensation coding table and perform decoding using the same method, and there is no need to newly transmit side information for table selection. The method of the present invention can be applied simply by performing the following steps.

[Brief explanation of drawings]

FIG. 1 is a diagram showing the basic configuration of the present invention.

FIG. 2 is a diagram showing an embodiment of the present invention.

[Fig. 3] An example of a motion compensation encoding table in the present invention (
It is a figure showing a).

[Fig. 4] An example of a motion compensation encoding table in the present invention (
It is a figure showing b).

[Fig. 5] An example of a motion compensation encoding table in the present invention (
It is a figure showing c).

FIG. 6 is an example of a motion compensation encoding table in the present invention (
It is a figure showing d).

FIG. 7 is a flowchart (1) showing an example of operation in the embodiment shown in FIG. 2;

FIG. 8 is a flowchart (2) showing an example of operation in the embodiment shown in FIG. 2;

FIG. 9 is a flowchart (1) showing another example of operation in the embodiment shown in FIG. 2;

FIG. 10 is a flowchart (2) showing another example of operation in the embodiment shown in FIG. 2;

FIG. 11 is a flowchart showing still another example of operation in the embodiment shown in FIG. 2;

FIG. 12 is a diagram showing an example (e) of a motion compensation encoding table in the present invention.

FIG. 13 is a diagram showing a conventional encoding circuit.

FIG. 14 is a diagram illustrating motion vector matching.

[Explanation of symbols]

1 Motion detection section 21 to 2n Motion compensation encoding table 3 Memory 4 Calculation/judgment section 51, 52 Switching section

Claims (4)

[Claims] Claim 1: A coded signal in which the difference between the reproduced image of the previous encoded frame and the predicted value is encoded is variable-length encoded, and a motion detection unit (1) encodes the encoded signal with the reproduced image of the previous encoded frame. In a high-efficiency encoding method for moving images in which a motion vector detected by matching with an original image is variable-length encoded using a motion compensation encoding table, a plurality of the motion compensation encoding tables (21 to 2n) are provided. and,
a memory (3) that holds the motion vector of the previous one encoded frame; a calculation/determination unit (4) that determines the nature of the occurrence frequency of the motion vector from the held motion vector of the previous one encoded frame; A motion-compensated variable-length encoding system, comprising a switching unit (51, 52) that selects the plurality of motion-compensated encoding tables (21 to 2n) according to the determination result. 2. The calculation/determination unit (4) determines whether the number of motion vectors that become 0 in one frame exceeds a certain threshold value and the maximum value of motion vectors in one frame. and information for selecting the plurality of motion compensation encoding tables (21 to 2n) according to the determination result of whether the minimum value exceeds the respective threshold value. The motion compensated variable length encoding method described. 3. A determination result by the calculation/determination unit (4) as to whether the number of motion vectors that take the maximum value or the minimum value in one frame exceeds the respective threshold;
generating information for selecting the plurality of motion compensation encoding tables (21 to 2n) according to a determination result as to whether the number of motion vectors that become 0 in one frame exceeds a certain threshold; The motion compensated variable length encoding method according to claim 1. 4. The calculation/judgment unit (4) calculates the average value of motion vectors in one frame, and selects the plurality of motion compensation encoding tables (21 to 21) according to the average value.
2. The motion compensated variable length coding system according to claim 1, wherein information for selecting a motion compensation variable length coding system (2n) is generated.