JPH07264598A - Compensating method for motion, motion vector detecting circuit and motion compensating circuit - Google Patents

Abstract

PURPOSE:To attain efficient motion compensation reduced at its predictive error even when an object having different motion is included in a block by compensating the motion of one block by two motion vectors. CONSTITUTION:Block matching between a current frame 11 and a frame 12 at preceding time is executed to detect a motion vector. Then an outline detecting part 22 judges whether a block in the frame 12 includes an outline or not, and at the time of judging the inclusion of an outline, whether an outline is included or not also in a block of the frame 11 is judged. At the time of judging that both the blocks of the frames 11, 12 include outlines, a motion vector is recalculated based upon the two extracted outlines through a matching error calculating part 30 and a motion vector 43 is outputted. The vector 43 is inputted to a compensating circuit as a motion vector 53 and motion is compensated by using the motion vectors of blocks neighbored each other in a direction divided by the outline.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a high-efficiency coding method for moving images.

[0002]

2. Description of the Related Art Conventionally, when it is determined that a block has a contour, the encoding is performed by adding a flag indicating the contour shape.

[0003]

In the conventional moving picture coding method using fixed size blocks, motion compensation is performed using only one motion vector for one block. When the motion compensation is performed with one motion vector, the motion compensation follows only the motion of one object. Therefore, when one block includes an object that performs another motion, there is a portion where the motion compensation does not work well. . In this case, the prediction error increases and the coding efficiency decreases.

In order to avoid this, there is a method of detecting a contour and performing motion compensation with a separate motion vector using the contour as a boundary [Miyamoto: "Motion Compensation Interframe Prediction Using Adaptive Block Subdivision", 1993]. School Spring University, D-
294]. The amount of information of the motion vector can be reduced by using the motion vector of the block adjacent in the direction delimited by the contour line as the motion vector for motion compensation. In addition, by using a motion vector that matches the motion of each object, the prediction error can be reduced and the amount of information involved can be reduced. However, it is necessary to encode the contour information as well, and it cannot be said that the effect of reducing the information amount as a whole is high depending on the contour selection method. Further, since the contour information is added, it is necessary to change the data structure.

An object of the present invention is to solve the above-mentioned problems, and even when an object having different motions is included in a block,
An object of the present invention is to provide a motion compensation method with high coding efficiency, which performs motion compensation with the same amount of information and data structure as the conventional method, and a motion vector detection circuit and motion compensation circuit used for this method.

[0006]

A motion compensation method according to the present invention comprises a step of performing block matching between a current frame and a previous frame to detect a motion vector, and an image of a previous frame referred to by the detected motion vector. The step of determining whether or not the block includes the contour, and when the block of the image of the previous frame referred to by the detected motion vector does not include the contour, motion compensation is performed using the vector calculated by matching for the entire block. And a step of determining whether the block of the current frame includes a contour if the block of the image of the previous frame referred to by the detected motion vector includes a contour, and a step of determining whether the block of the current frame does not include the contour. It is judged that there is a contour in the block of the frame at the same position as the block of the previous frame, and Dividing the area in line,
If the block of the current frame includes a contour, the step of determining a motion vector between two contour lines, the step of dividing the block of the current frame into regions by the contour line, and the step of determining the motion vector of each divided region And a step of performing motion compensation in each area using the determined motion vector.

The motion vector detection circuit of the present invention comprises a first frame memory in which an image of the current frame is stored, and a second frame memory in which an image of the previous frame is stored.
The image of the current frame stored in the first frame memory is read out and divided into blocks that are units for detecting a motion vector, and the current frame block signal is used, and the offset value of the address of the read image is used as a block offset address signal. The first small area dividing means for outputting and the image of the previous frame accumulated in the second frame memory are read out, divided into blocks which are motion vector detection units, and read out again as a previous frame block signal. Second small area dividing means for outputting the offset value of the address as a block offset address signal, an address memory for storing the motion vector, and a current frame block outputted from the first small area dividing means when activated. Input a signal and whether the current frame block has a contour If it is determined that the contour line is not included, the address stored in the address memory is output as the motion vector of the block, and if it is determined that the contour line is included, the position of the contour line in the current frame block signal is determined. Are output sequentially from the second frame sub-dividing means and the first frame detecting means for outputting the current frame block signal and the contour line shift amount in the current frame block signal while the contour lines are extracted. The previous frame block signal is input, and it is determined whether or not the contour line is included in the block signal. When it is determined that the contour line is not included, the motion vector stored in the address memory is used as the motion vector of the block. When it is determined that the contour line is included, the first contour detection unit is activated, and the first contour line detection unit causes the current frame to be displayed. When it is determined that the block includes the contour line, the second frame detecting unit that outputs the previous frame block signal in the form of extracting the contour line and the current frame that is respectively output from the first and second small area dividing units. A block signal and a previous frame block signal are input, a first difference value between the two signals is calculated, and a contour line is extracted from the current frame block signal output from the first and second contour detection means. And a previous frame block signal are input, a matching error calculating means for calculating a second difference value of both signals and a minimum value memory are built in, and at the time of block matching, the first difference value is obtained from the matching error calculating means. Input and the first difference value is
If it is smaller than the value stored in the minimum value memory,
The minimum value memory is updated with a first difference value, an address which is a value obtained by subtracting the current frame block offset address signal from the previous frame block offset address signal is recorded in the address memory, and the matching error is detected when a contour line is detected. The second difference value is input from the calculating means, and the second difference value is input.
Is smaller than the value stored in the minimum value memory, the contour line shift amount output from the first contour detecting means at that time is added to the motion vector in the address memory obtained by block matching. However, it has a comparison means for storing it in the address memory.

The motion compensating circuit of the present invention stores the current frame reproduction signal by adding a difference signal to the previous frame block signal, a frame memory in which the image of the previous frame is stored, a vector memory in which the motion vector of each block is stored. A reproduction signal generating means for generating and outputting, and an image of the previous frame stored in the frame memory for each area divided by the contour line is read as an offset of the value of the address signal, input, and the motion vector of the motion vector It is divided into blocks in motion compensation units, small area dividing means for outputting the previous frame block signal, and which of the adjacent blocks to use in the area divided by the contour line is determined from the position of the contour line, The motion vector stored in the vector memory corresponding to each area is stored in the small area dividing unit. The vector determining means to be applied and the previous frame block signal are input, and it is determined whether or not a contour line is included in the input block. If the contour line is not included, the input is from the small area dividing means. Contour detection for outputting the same block signal as the previous frame block signal to the reproduction signal processing means and outputting a signal indicating a contour position to the small area dividing means and the vector determining means when a contour line is included. And means.

[0009]

1 is a flow chart showing a motion compensation method of the present invention, and FIG. 2 is a view showing a motion vector selection method.

In the motion vector detection, first, block matching is performed between the current time frame and the previous time frame to detect the motion vector (step 1). Next, it is determined whether the block of the frame at the previous time point referred to by the detected motion vector is a block including a contour (step 2). If the contour line is not extracted, motion compensation is performed using the vector calculated by matching in the entire block (step 3). If the previous block contains contours,
It is judged whether or not the block at the current time includes the contour (step 4). If it is determined that the block contains the motion vector, the motion vector is recalculated using the two extracted contour lines, and this is set as the motion vector of the block (step 6).

In the motion compensation, when the block at the previous time includes a contour, for each part delimited by the contour line, adjacent blocks in the direction demarcated by the contour line, for example, left and right Performs motion compensation using the motion vectors of the blocks adjacent to the left and right, and the blocks adjacent to the top and bottom if they are above and below. For example, when there is a contour line in the current time block as shown in FIG. 2A, the region 1 uses the motion vector of the block d adjacent to the left side, and the region 2 uses the motion vector of the block e adjacent to the right side.
Further, as shown in FIG. 2B, when one area divided by the contour has three adjacent blocks, the area 1 has a motion vector of the block a, an average value of the motion vectors of the blocks a, b, c, etc. You can choose freely. The same applies to the selection of the motion vector in the area 2. At this time, even if only the block at the previous time includes the contour and the block at the current time does not include the contour, it is determined that the contour exists at the same position as the previous time block in the current block, and the area is defined by the contour line. Is divided (step 5).

When the motion vector is determined, the block of the frame at the current time is divided into areas by the contour line (step 7). Next, the motion vector of each area is determined (step 8). Finally, motion compensation is performed on each area (step 9).

According to the motion compensation method described above,
A motion vector is detected by the motion vector detection circuit, and motion compensation is performed by the motion compensation circuit.

Therefore, according to the present invention, it is possible to perform efficient motion compensation with a small prediction error even when a plurality of motions are included in a block. Since it is not necessary to encode the contour information, the coded data is the same as that of the conventional moving image coding method using a fixed-size block.

[0015]

Embodiments of the present invention will now be described with reference to the drawings.

FIG. 3 is a block diagram of a motion vector detecting circuit according to an embodiment of the present invention.

The image 11 of the current frame is stored in the frame memory 13 which is the first frame memory, and the previous time frame image 12 which is referred to when detecting the motion vector is stored in the frame memory 14 which is the second frame memory. To be

The current frame image 16 stored in the frame memory 13 is read out by using the value of the address signal 15 consisting of two components (x, y) as an offset value, and the first small area dividing means is a small area dividing unit. It is input to the device 19. The small area divider unit 19 divides the image 16 of the current frame into blocks that are units of motion vector detection, and outputs a current frame block signal 24. Specifically, first, the image signal at the position corresponding to the upper left of the current frame block signal 24 to be output is read by outputting the stored offset value as the address signal 15 to the frame memory 13. Hereinafter, the number of pixels in the block is read while changing the value of the address signal 15 so as to scan all the coordinates in the block to be output, and the signals read out at the same time are output as the current frame block signal 24. Finally, the offset value (x, y)
Is output as a block offset address signal 27 to a comparator 33 which is a comparing means. The operation of the small area divider 20 which is the second small area dividing means is also the same.

The image 18 of the previous frame is a small area divider 20.
Is divided into blocks and is output as the previous frame block signal 26. Current frame block signal 24
And the previous frame block signal 26 is input to the matching error calculating section 30 which is a matching error calculating means.
In the matching error calculation unit 30, the two small area dividers 1
The difference square value of the pixel values output from 9 and 20 is calculated,
The sum of them is output as the difference value 31 to the comparator 32. However, the difference value calculation is not limited to the difference square, and the difference absolute value or the like can be used. Comparator 3
2 has a built-in minimum value memory 23, compares the input difference value 31 with the value stored in the memory 33, and if the former is smaller than the latter, updates the value in the memory 33 with the difference value 31. An address 34, which is a value obtained by subtracting the current frame block offset address signal 27 from the previous frame block offset address signal 29 each consisting of two components (x, y), is recorded in the address memory 35. After the above operation is completed within the motion vector search range in the previous frame image, the motion vector of the block is stored in the address memory 35.

Next, the small area divider 20 reads the previous frame image from the frame memory 14 using the value of the address signal 17 as an offset value by the same operation as described above, and the offset value at this time is the frame in the previous operation. The offset value at which the current frame image is read from the memory 13 is shifted by the motion vector detected in block units. Therefore, the small area divider 20 uses, as the read address signal 17 of the previous frame image, the one obtained by adding the address signal 37 which is the motion vector of the block to the offset value from which the current frame image signal is read. Furthermore, the second operation is performed by the same operation as described above.
The previous frame block signal 39 is input to the contour detecting section 22 which is the contour detecting means of the above. In the contour detection unit 22, first,
It is determined whether or not a contour line is included in the input block. If it is determined that the block is not included, the motion vector detecting operation is terminated, and the vector stored in the address memory 35 is output as the motion vector 43 of the block. When it is determined that the contour line is included, the control signal 42
And 41 output the current frame block signal 38 from the small area divider 19 to the contour detecting section 21 which is the first contour detecting means, and the contour detecting section 21 determines whether or not a contour line exists in the current frame block. To perform. When it is determined that the contour line is not included, the motion vector detecting operation is ended, and the vector stored in the address memory 35 is output as the motion vector 43 of the block. When it is determined that the contour line is included, the matching error calculation unit 30 compares the previous frame block signal 25 and the current frame block signal 23 with the contour line extracted.
Then, the motion vector 36 of the block itself and the contour shift amount 28 ((0, 0) at the initial position) in the current frame block signal 23 are output to the comparator 32. Contour detector 21
Represents the current frame block signal 23 to the matching error calculation unit 30 while sequentially shifting the contour line position in the current frame block signal 38, and the shifted amount (contour line shift amount 28) in the form of (x, y). Is output to the comparator 32, and the difference value between the blocks at that time and the minimum value memory 33, as in the case of the block matching performed before the contour detection.
Is compared with the minimum value stored up to now, and if the value is smaller than the minimum value, the shift amount 28 at that time is added to the motion vector 36 obtained by the first block matching in the comparator 33 to obtain the motion vector. It outputs to the address memory 35 as 34. When the contour line is shifted in the entire block and the above-mentioned operation is completed, the motion vector of the contour line is stored in the address memory 35, and it is output as the motion vector 43 of the block.

FIG. 4 is a block diagram of a motion compensation circuit according to an embodiment of the present invention.

The image 44 of the previous frame is the frame memory 4
5, and the motion vector 53 of each block (motion vector 43 obtained by the motion vector detection circuit) is stored in the vector memory 54.

The image 47 of the previous frame stored in the frame memory 45 is read out by using the value of the address signal 46 consisting of two components (x, y) as an offset value, and the small area divider 48 as a small area dividing means. Entered in. The value of the address signal 46 at this time is the motion vector 56 as the offset address value of the block for which motion compensation is currently performed.
The value of is added. The small area divider 48 divides the image 47 of the previous frame into blocks that are also units of motion vector detection, that is, units of motion compensation, and outputs the previous frame block signal 50 to the contour detection unit 51 that is contour detection means.

The contour detecting section 51 first determines whether or not a contour line is included in the input block. If it is determined that they are not included, the corresponding small control signal 49 controls the small area divider 48, and the same block signal input from the small area divider 48 to the contour detection unit 51 is used as the previous frame block signal 59. The signal is input to the reproduction signal generation unit 61 which is reproduction signal generation means. When it is determined that the contour line is included, the contour detection unit 51 outputs the signal 49 indicating the contour position.
And 52 are output to the small area divider 48 and the vector determining unit 57 which is vector determining means, respectively. The vector determining unit 57 determines from which position of the contour line the motion vector of each adjacent block in the region divided by the contour line should be used, and the motion vector 58 corresponding to each region is divided into small regions. Output to the container 48. In the small area divider 48, the value obtained by adding the motion vector of each area to the offset address value of the block for which motion compensation is performed is given to the frame memory 45 as the address signal 46 for each area divided by the contour line. , The previous frame block signal 59 corresponding to each area is read from the frame memory 45 and output to the reproduction signal generation unit 61.

The reproduced signal generator 61 adds the difference signal 60 given as encoded data to the previous frame block signal 59 to generate and output the current frame reproduced signal 62.

[0026]

As described above, according to the present invention, by performing motion compensation using two motion vectors for one block, even when an object having different motion is included in the block, With the same amount of information and data structure as the conventional one, it is possible to perform efficient motion compensation with less prediction error.

[Brief description of drawings]

FIG. 1 is a flowchart showing a motion compensation method of the present invention.

FIG. 2 is a diagram showing a motion vector selection method.

FIG. 3 is a circuit diagram of a motion vector detection circuit according to an embodiment of the present invention.

FIG. 4 is a circuit diagram of a motion compensation circuit according to an embodiment of the present invention.

[Explanation of symbols]

 1 to 9 Step 11 Current frame image signal 12 Previous frame image signal 13, 14 Frame memory 15 Current frame image read address signal 16 Current frame image 17 Previous frame image read address signal 18 Previous frame image 19, 20 Small area Dividers 21 and 22 Contour detection unit 23 Current frame block signal after contour extraction 24 Current frame block signal 25 Previous frame block signal after contour extraction 26 Previous frame block signal 27 Current frame block offset address signal 28 Current frame block contour line shift Amount 29 Previous frame block offset address signal 30 Matching error calculator 31 Difference value 32 Comparator 33 Memory 34 Address data 35 Address memory 36, 37 Address data 38 Current frame Frame block signal 39 Previous frame block signal 40, 41, 42 Control signal 43 Motion vector 44 Previous frame image signal 45 Frame memory 46 Address signal for reading previous frame image 47 Image of previous frame 48 Small area divider 49 Contour line position 50 Previous frame Block signal 51 Contour detection unit 52 Contour line position 53 Motion vector 54 Vector memory 55, 56 Motion vector 57 Vector determination unit 58 Motion vector 59 Previous frame block signal 60 Difference signal 61 Reproduction signal generation unit 62 Current frame reproduction signal

Claims (3)

[Claims] 1. A motion compensation method that divides an image into blocks and uses two motion vectors for each block to generate a predicted image, wherein block matching is performed between a current frame and a previous frame. And detecting whether the block of the previous frame image referenced by the detected motion vector contains a contour, and the block of the previous frame image referenced by the detected motion vector contains a contour. If not, the step of performing motion compensation using the vector calculated by matching in the entire block, and if the block of the image of the previous frame referred to by the detected motion vector includes the contour, the block of the current frame includes the contour. Determining if and if the block in the current frame does not contain a contour, The step of dividing the region by the contour line by determining that the block of the current frame has the contour at the same position as the block of the previous frame, and the motion vector between the two contour lines when the block of the current frame includes the contour , The step of dividing the block of the current frame into regions by contour lines, the step of determining the motion vector of each divided region, and the step of performing motion compensation in each region using the determined motion vector A motion compensation method including and. 2. A first frame memory in which an image of the current frame is stored, a second frame memory in which an image of the previous frame is stored, and an image of the current frame stored in the first frame memory. First small area dividing means for reading and dividing into blocks which are motion vector detection units, and outputting as a current frame block signal and an offset value of the read image address as a block offset address signal, and a second frame A second frame in which the image of the previous frame stored in the memory is read, divided into blocks that are units for detecting a motion vector, and output as a block signal of the previous frame and an offset value of the address of the read image as a block offset address signal. Small area dividing means, an address memory for storing motion vectors, and Then, the current frame block signal output from the first small area dividing means is input, it is determined whether or not a contour line exists in the current frame block, and when it is determined that the contour line is not included, the address memory If it is determined that the contour line is included by outputting the address stored in the block as the motion vector of that block, the contour line position in the current frame block signal is sequentially shifted, and the contour line is extracted in the current frame block. The signal, the first contour detecting means for outputting the contour line shift amount in the current frame block signal, and the previous frame block signal outputted from the second small area dividing means are inputted, and the contour line is added to the block signal. If it is determined that the contour line is not included, it is determined that the motion vector stored in the address memory is included in the block. When it is determined that the contour line is included, the first contour detecting unit is activated, and when the first contour detecting unit determines that the current frame block includes the contour line, the contour line is detected. Second contour detecting means for outputting the previous frame block signal in the form of extracting the current frame block signal and the previous frame block signal respectively output from the first and second small area dividing means, Computes a first difference value of the signals and
And matching error calculation means for inputting the current frame block signal and the previous frame block signal from which the contours have been extracted, which are respectively output from the second contour detection means, and which calculates a second difference value between the two signals. A minimum value memory is built in, and at the time of block matching, the first difference value is input from the matching error calculating means,
Is smaller than the value stored in the minimum value memory, the minimum value memory is updated with a first difference value, and the current frame block offset address signal is subtracted from the previous frame block offset address signal. Is recorded in the address memory, a second difference value is input from the matching error calculating means when the contour line is detected, and if the second difference value is smaller than the value stored in the minimum value memory, A motion vector detecting circuit having a comparing means for adding the contour shift amount outputted from the first contour detecting means at that time and the motion vector in the address memory obtained by the block matching and storing the result in the address memory. 3. A frame memory for storing an image of a previous frame, a vector memory for storing a motion vector of each block, and a reproduction for generating and outputting a current frame reproduction signal by adding a difference signal to the previous frame block signal. Signal generating means, for each area divided by the contour line, the image of the previous frame stored in the frame memory is read out by inputting the value of the address signal as an offset and input, and is divided into blocks by the motion compensation unit of the motion vector. The small area dividing means that divides and outputs the previous frame block signal, and which of the blocks adjacent to each other in the area divided by the contour line should be determined from the position of the contour line and correspond to each area , Vector determination for outputting the motion vector stored in the vector memory to the small area dividing means And a previous frame block signal are input, it is determined whether or not a contour line is included in the input block, and if the contour line is not included, the same block as that input from the small area dividing means. A signal is output to the reproduction signal processing means as a previous frame block signal, and when a contour line is included, it has a contour detection means for outputting a signal indicating a contour position to the small area dividing means and the vector determining means. Motion compensation circuit.