JP2002084543A - Motion detection circuit and its method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a motion detection circuit capable of raising encoding efficiency, without causing deterioration of the image quality. SOLUTION: A comparator 9 compares a differential absolute value sum input from an image calculator 11 having a threshold, and based on the comparison results, an internal motion vector threshold or a motion vector to be input from the image calculator 11 is output to a motion detection part 3 as a motion vector. The motion detection part 3 generates an address shifted, corresponding to the motion vector to be input from the comparator 9 to output to a memory 2; calculates as absolute value, the difference between a reference signal shifted, corresponding to the motion vector from a normal search start coordinate and a source signal to be input from a filter circuit 1; searches for the motion vector, in which the value of the differential absolute value sum is minimized; outputs a minimal differential absolute value sum; and outputs the motion vector of that time.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a motion detection circuit and a method therefor, and more particularly to a motion detection circuit used for a moving picture compression system.

[0002]

2. Description of the Related Art Conventionally, with regard to a moving image compression method, an MPEG (Moving) targeting a storage medium has been used.
Picture Experts Group) 1, MPEG2 that can be used over a wide range of media including broadcasting, communication, and storage; MPEG4 with advanced functions such as object-based encoding in video; and multimedia information retrieval such as video and audio. Standardization of the target MPEG7 and the like is in progress.

The above-mentioned MPEG2 is based on the international standard ITU-
TH. 262 is a moving picture coding method formulated in 262,
The basis of the video compression algorithm is the same inter-frame motion compensation prediction and discrete cosine transform (DC
T: Discrete Cosine Transfo
rm).

[0004] In the TM5 (test model 5) of the MPEG2, full search is performed in half-pixel units for motion prediction. Since the amount of calculation is enormous in the method of performing the full search, a motion detection circuit as shown in FIG. 5 has been proposed. In FIG. 5, the motion detection circuit is a filter circuit 21.
, Memories (MEM) 22, 25, 27, a motion detection unit (ME: Motion Estimation) 23,
26, and a delay circuit (DLY) 24.

In the above-described motion detection circuit, a motion vector (MV: Mot:
In this method, a motion vector MV2 up to half-pixel accuracy is calculated by the second-stage motion detection unit 26 based on the calculated motion vector MV1.

Further, by using the motion vector of the screen calculated in advance, it is possible to reduce the amount of calculation for the entire search and expand the search range. According to this method, the image quality is hardly degraded, and the calculation amount can be greatly reduced.

[0007]

In the above-mentioned conventional motion detection circuit, as shown in FIG. 6, a moving image exceeding the search range of the first-stage motion detecting section or a moving image having a random motion is encoded. In this case, there is a high possibility that a motion vector is erroneously detected. As a result, since an optimal motion vector is not selected, encoding efficiency cannot be increased, and image quality is greatly degraded.

That is, as shown in FIG. 6, in the conventional method, the position (MB) C of the image for which the search has already been completed and the position of the image to be searched (MB: Macro Block)
D and the position of reference image I1 (MB)
A does not fall within the search range, and an incorrect position (MB) B is selected.

SUMMARY OF THE INVENTION It is an object of the present invention to solve the above-mentioned problems and to provide a motion detection circuit and a method thereof capable of improving the coding efficiency without deteriorating the image quality.

[0010]

A motion detection circuit according to the present invention is a motion detection circuit for obtaining a final motion vector by using a motion vector of a screen calculated in advance.
The motion vector of the screen calculated in advance is used at an optimum position.

In other words, the motion detection circuit of the present invention uses a previously calculated screen motion vector in a method of using a previously calculated screen motion vector at an optimum position to perform a full search and a motion search. Almost the same effects can be obtained. In this case, it is effective to suppress the deterioration of the image quality by minimizing the erroneous detection of the motion vector and giving the code amount to other information.

More specifically, in the motion detection circuit of the present invention, as shown in FIG. 3, the screen B2 shifted by the motion vector MVB2 at the position D of the screen B3 (B3 picture)
By using the motion vector MVB2 calculated in advance at the position C of the (B2 picture), it is possible to refer to the image at the position A of the screen I1 (I1 picture) referred to at the position C.

When there are two or more vector candidates at the position D, it is the sum of the absolute values of the differences between the corresponding pixels of the macroblock (16 pixels × 16 lines) in the reference image and the current image. Sum of absolute differences (AE: A
It is possible to provide a method in which a motion vector having a small value of (bsolve Error) is adopted. Furthermore, when there are two or more vector candidates at the position D, it is also possible to provide a method of setting the motion vector at the position D to 0.

[0014]

Next, an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing a configuration of a motion detection circuit according to one embodiment of the present invention. In FIG. 1, the motion detection circuit according to one embodiment of the present invention includes n
Although it can be applied to a motion detection circuit having a two-stage configuration, a motion detection circuit having a two-stage configuration will be described below for simplicity.

A motion detection circuit according to one embodiment of the present invention includes a filter circuit 1, memories (MEM) 2, 5, 7, and 8,
Motion detection unit (ME: Motion Estimation)
n) 3, 6; a delay circuit (DLY) 4;
P) 9, 10 and picture calculator (PIC CAL) 1
And 1.

The filter circuit 1 has a video source signal (Vi
deo) is input, and the reference signal refi converted into an input format corresponding to the processing unit of the motion detection unit 3 is input.
1 and a source signal src1.

The memory 2 is connected to the filter circuit 1 and
Picture (Intra-code picture: intra-coded frame) and P picture (Predic
5 is written as a reference signal refi1.

The comparator 9 is connected to the picture calculator 11, compares the absolute difference sum (AE: Absolute Error) AE 2 ′ input from the picture calculator 11 with a threshold value TH 1, and, based on the result of the comparison, stores an internal value. Motion vector threshold MVTH or picture calculator 1
The motion vector MV 2 ′ input from No. 1 is output to the motion detecting unit 3 as a motion vector MVO 1.

The motion detector 3 comprises a filter circuit 1 and a memory 2
And an address AD1 connected to the comparator 9 and shifted by the motion vector MVO1 input from the comparator 9.
Is generated and output to the memory 2.

Further, the motion detecting section 3 calculates a search range [the sum of absolute differences, which is the cumulative sum of the absolute values of the differences of the respective pixels in the macroblock (16 pixels × 16 lines)] in the reference image and the current image. Is calculated, and a motion vector (a motion vector in an optimum search direction) in which the obtained sum of the absolute differences is minimized is searched, and the minimum sum of the absolute differences is output as the sum AE1 of absolute differences. The motion vector is output as the motion vector MV1.

Here, the sum of absolute differences by the motion detector 3 is calculated by calculating the difference between the reference signal ref1 shifted from the normal search start coordinate by the motion vector MVO1 and the source signal src1 input from the filter circuit 1. Then, it is obtained by accumulating in units of macro blocks (16 pixels × 16 lines = 256 samples).

The memory 5, the motion detector 6, and the comparator 10 have the same configuration as the memory 2, the motion detector 3, and the comparator 9, and the motion detector 6 has the minimum sum of absolute differences. Is output as the sum of absolute differences AE2, and the motion vector at that time is output as the motion vector MV2.

As for the inputs of the memory 5 and the motion detecting section 6, since the memory 5 is connected to the delay circuit 4, the video source signal (Video) delayed by the delay circuit 4 is input to the memory 5. You.

The picture calculator 11 is connected to the motion detector 6 and writes the motion vector MV2 and the sum of absolute differences AE2 from the motion detector 6 into the memories 7 and 8, respectively. The memory 7 is connected to the picture calculator 11, and writes the motion vector MV2 from the motion detector 6 via the picture calculator 11. The motion vector MV2 recorded in the memory 7
Represents the motion vector M to the comparator 9 via the picture calculator 11.
It is output as V2 '.

The memory 8 is connected to the picture calculator 11 and writes the absolute difference sum AE2 from the motion detector 6 via the picture calculator 11. The sum of absolute differences AE2 recorded in the memory 8 is output to the comparator 9 via the picture calculator 11 as the sum of absolute differences AE2 '.

FIG. 2 is a block diagram showing the configuration of the picture calculator 11 of FIG. In FIG. 2, a picture calculator 11 includes a comparator 11a and an output control circuit (OE: Outp).
outEnable) 11b and 11c.

The comparator 11a receives the sum of absolute differences AE2 from the motion detector 6 and the sum of absolute differences AEA read from the memory 8, compares them, and calculates the sum of absolute differences AE2. If the value is smaller than the value of the value sum AEA, the motion vector MV2 from the motion detector 6 is written to the memory 7 and the absolute difference sum AE2 from the motion detector 6 is written to the memory 8 for the output control circuits 11b and 11c. Control signal.

If the sum of absolute differences AE2 is equal to the value of the sum of absolute differences AEA, the comparator 11a receives the motion vector MV2 from the motion detector 6 and the motion vector MVA read from the memory 7. When the motion vector MV2 is smaller than the value of the motion vector MVA, the motion vector MV2 from the motion detecting unit 6 is stored in the memory 7 for the output control circuits 11b and 11c. Are written so as to write the absolute difference sum AE2.

Further, the motion vector MVA is read out from the memory 7 as necessary and output as a motion vector MV2 '. Similarly, the absolute difference sum AEA is read from the memory 8 as necessary, and is output as the absolute difference sum AE2 '.

The operation of the motion detection circuit according to one embodiment of the present invention will be described with reference to FIGS.
Note that the motion detection units 3 and 6 are described assuming that a macroblock (256 samples) is used as a processing unit.

Each time one sample of the video source signal (Video) is input, the filter circuit 1 inputs the video source signal (Video) and converts the input signal into a reference signal refi1 which is converted to an input format corresponding to the processing unit of the motion detection unit 3. And a source signal src1.

Next, the motion vector M
Although the calculation of V1 and the sum of absolute differences AE1 will be described, the calculation of the motion vector MV2 and the sum of absolute differences AE2 by the motion detection unit 6 is the same as the calculation of the motion vector MV1 and the sum of absolute differences AE1. The description is omitted.

First, an I picture and a P picture are written in the memory 2 as a reference signal refi1.
The comparator 9 is connected to the picture calculator 11 and compares the absolute difference sum AE2 ′ input from the picture calculator 11 with the threshold value TH1, and based on the comparison result, the internal motion vector threshold value MVTH or picture calculation value. The motion vector MV2 'input from the unit 11 is
It is output to the motion detection unit 3 as O1.

The motion detecting section 3 comprises a filter circuit 1 and a memory 2
And an address AD1 connected to the comparator 9 and shifted by the motion vector MVO1 input from the comparator 9.
Is generated and output to the memory 2, and the reference signal r shifted by the motion vector MVO1 from the normal search start coordinates
ef1 and the source signal s input from the filter circuit 1
The absolute value of the difference from rc1 is calculated, and the macroblock (256
A motion vector that minimizes the value of the sum of absolute differences obtained by accumulating in units of samples) is searched for, the minimum sum of absolute differences is output as a sum of absolute differences AE1, and the motion vector at that time is referred to as a motion vector MV1. Output.

The memory 5, the motion detector 6, and the comparator 10 have the same configuration as the memory 2, the motion detector 3, and the comparator 9, and the motion detector 6 has the minimum sum of absolute differences. Is output as the sum of absolute differences AE2, and the motion vector at that time is output as the motion vector MV2.

FIG. 3 is a diagram showing the operation of the motion detection circuit according to one embodiment of the present invention. The operation of the motion detection circuit according to one embodiment of the present invention will be described with reference to FIGS.

As shown in FIG. 3, the motion vector MV2 obtained at the position (MB) C is shifted by the motion vector MVB2 from the position (MB) C in the memory space of the memory 7 by the picture calculator 11. MB) D is written. Similarly, the absolute difference sum AE2 obtained at the position (MB) C is simultaneously written into the memory 8. However, it is assumed that writing is performed at the position (MB) D for the first time.

When there are two or more vector candidates at the position (MB) D, a motion vector having a small value of the sum of absolute differences AE is adopted. Specifically, the position (MB) F
Is calculated by the picture calculator 11
In the memory space of the memory 7, the position (MB) D shifted from the position (MB) F by the motion vector MVB2 '
And the position (MB) D shifted by the motion vector MVB2 from the position (MB) C, a motion vector having a small value of the sum of absolute differences AE is adopted and the memory 7,
Written in 8.

When the sum of absolute difference values AE is equal, the values of the motion vector MV2 and the motion vector MV2 'are compared, and the smaller one is written to the memories 7 and 8. For example, when the motion vector MV2 <the motion vector MV2 ', the motion vector MV2 is already written in the memory 7, so that the writing is not performed. Similarly, the memory 8
Since the sum of absolute differences AE2 has already been written in the register, the writing is not performed.

Further, when the sum of the absolute difference AE is equal to the value of the motion vector MV, a value close to the previously calculated motion vector MV at the adjacent position is used. The motion vector is set to 0 for the position (MB) where writing is not performed.

By reading the difference absolute value sum AE and the motion vector MV written as described above as the difference absolute value sum AE2 'and the motion vector MV2', the motion detecting section 3 shifts by the motion vector MVB2. It is possible to start the vector search from the specified position, and the reference image I1
(MB) A can be searched.

As shown in FIG. 3, by using the motion vector MVB2 previously calculated at the position C of the screen B2 shifted by the motion vector MVB2 at the position D of the screen B3, the screen I1 referred to at the position C is obtained. Position A
Can be referred to.

When there are two or more vector candidates at the position D, it is possible to provide a method of using a motion vector having a small value of the sum of absolute differences AE. Furthermore, when there are two or more vector candidates at the position D, a method of setting the motion vector at the position D to 0 can be provided.

As described above, by using the motion vector of the screen calculated in advance at the optimum position, the motion vector MV at the same position of the screen calculated in advance is used in the screen to be searched. In comparison, since the detection accuracy is higher, it is possible to obtain almost the same effect as the full search in the motion detection circuit that obtains the final motion vector MV using the motion vector MV calculated in advance. Therefore, it is possible to reduce the information amount of the motion vector MV for the entire code information.

Also, since the mode selection is determined by the image data of the video source signal and the threshold values THMV and THAE, it is possible to automatically detect the motion vector MV.

Further, threshold values THMV, TH
Since AE can be set, MPEG1,2,4
And various image formats such as 1080I, 720P, and 480I.

Furthermore, since it can be realized by a simple logical configuration such as a comparator and a selector, it is possible to easily select a more optimal mode by slightly changing the input / output conditions as in the picture calculator 11. Becomes possible.

In one embodiment of the present invention, the motion vector MV calculated in advance and the sum of absolute differences A
Since only E and the threshold value TH are input conditions, n
(N: 1 to infinity) When performing a motion detection search in stages, n
This can be easily handled by adding a comparator to the inputs of the individual motion detectors.

FIG. 4 is a block diagram showing the structure of a picture calculator according to another embodiment of the present invention. In FIG.
A picture calculator 12 according to another embodiment of the present invention includes a comparator (CMP) 12a and an output control circuit (OE) 12b, 1
2c and selection circuits (SEL) 12d and 12e. Note that the picture calculator 12 according to another embodiment of the present invention is the same as the picture calculator 11 according to one embodiment of the present invention.
Since the configuration is almost the same, only the different configuration will be described below.

The picture calculator 12 has a selection circuit 12d,
12e is added, the threshold value THMV and the motion vector MV2 are input to the selection circuit 12d, and the smaller one of the motion vector MVA and the motion vector MV2 is larger than a certain threshold value THMV, If the smaller one of the absolute difference sum AEA and the absolute difference sum AE2 is larger than a certain threshold THAE, the threshold THMV is determined as the motion vector of the position D.
Is written to the memory 7 via the output control circuit 12b.

Similarly, the threshold value TH is applied to the selection circuit 12e.
AE and the sum of absolute differences AE2 are input, and the threshold value THAE is output as the sum of absolute differences AE at the position D by the output control circuit 1.
Write to the memory 8 via 2c.

Therefore, when the motion vector MV is large and the sum of absolute differences AE is large, the possibility of erroneous detection is high. Therefore, by selecting the threshold THMV of the motion vector, erroneous detection can be reduced. Has a new effect. For example, if there are two or more vector candidates at the position (MB) D shown in FIG. 3, the motion vector at the position D is set to 0.

The threshold values THMV and THAE are 108
In order to cope with various conditions such as the 0I and 480I formats, the parameters are input to the selection circuits 12d and 12e as parameters for performing this correction.

The threshold value THMV and the motion vector MV2 are input to the selection circuit 12d, and the position (M
B) When a motion vector has already been written as in D, the threshold value THMV is written to the memory 7 via the output control circuit 12b as the motion vector at the position D.

Similarly, the threshold value TH is applied to the selection circuit 12e.
When AE and the sum of absolute differences AE2 are input and the vector has already been written, the sum of absolute differences A at the position D is calculated.
The threshold value THAE is written to the memory 8 via the output control circuit 12c as E. Therefore, since the motion vector MV is not compared with the sum of absolute differences AE, there is a new effect that the circuit scale can be reduced.

The threshold values THMV and THAE are 10
In order to cope with various conditions such as the 80I and 480I formats, the parameters are input to the selection circuits 12d and 12e as parameters for performing this correction.

[0057]

As described above, according to the present invention, in a motion detection circuit for obtaining a final motion vector using a previously calculated screen motion vector, a previously calculated screen motion vector is calculated. Is used at an optimal position, thereby improving coding efficiency without deteriorating image quality.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a motion detection circuit according to one embodiment of the present invention.

FIG. 2 is a block diagram illustrating a configuration of a picture calculator of FIG.

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

FIG. 4 is a block diagram illustrating a configuration of a picture calculator according to another embodiment of the present invention.

FIG. 5 is a block diagram showing a configuration of a conventional motion detection circuit.

FIG. 6 is a diagram illustrating an operation of a conventional motion detection circuit.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Filter circuit 2, 5, 7, 8 Memory 3, 6 Motion detection part 4 Delay circuit 9, 10, 11a, 12a Comparator 11, 12 Picture calculator 11b, 11c, 12b, 12c Output control circuit 12d, 12e Selection circuit

Claims (16)

[Claims] 1. A motion detection circuit for obtaining a final motion vector by using a previously calculated screen motion vector, wherein the previously calculated screen motion vector is used at an optimum position. A motion detection circuit characterized by having such a configuration. 2. The apparatus according to claim 1, wherein the optimum position is searched based on a sum of absolute differences, which is a cumulative sum of absolute values of differences between corresponding pixels in a search range between the reference image and the current image. The motion detection circuit according to claim 1, wherein: 3. The motion detection circuit according to claim 2, further comprising means for searching for a motion vector that minimizes the sum of absolute differences. 4. An accumulator for independently accumulating the final motion vector and the sum of absolute differences used for searching for the motion vector, wherein a next motion vector is stored based on the contents of the accumulator. The motion detection circuit according to claim 2, wherein the motion detection circuit is configured to obtain 5. A method for comparing the sum of absolute differences used in the search for the motion vector of the screen calculated in advance with a predetermined threshold, and calculating the sum calculated in advance according to the comparison result. And a comparator that outputs one of the motion vector of the screen and a preset motion vector threshold value, and the value of the sum of absolute differences is minimized by shifting from a normal search start coordinate by the output of the comparator. 5. A motion detecting unit for searching for a motion vector.
The motion detection circuit according to any one of the above. 6. The method according to claim 6, wherein when the motion vector is large and the sum of absolute differences is large, a threshold value of a preset motion vector is stored in the storage means instead of the sum of absolute differences. The motion detection circuit according to claim 4 or claim 5, wherein 7. The method according to claim 2, wherein when the value of the sum of absolute differences is equal to the value of the motion vector, a value close to the previously calculated motion vector at the next position is adopted. The motion detection circuit according to claim 6. 8. The motion detection circuit according to claim 4, wherein a motion vector at a position where writing is not performed in said storage means is set to 0. 9. A motion detecting method for obtaining a final motion vector using a previously calculated screen motion vector, wherein the previously calculated screen motion vector is used at an optimum position. A motion detection method characterized by doing so. 10. The search for the optimum position is performed based on a difference absolute value sum that is a cumulative sum of absolute values of differences between corresponding pixels in a search range between a reference image and a current image. The motion detection method according to claim 9, wherein: 11. The apparatus according to claim 1, wherein a motion vector in which the sum of absolute differences is minimized is searched.
0. The motion detection method according to item 0. 12. The storage unit independently stores the final motion vector and the sum of absolute differences used for searching for the motion vector, and calculates the next motion vector based on the content of the storage unit. The motion detection method according to claim 10, wherein the motion is obtained. 13. A method for comparing the sum of absolute differences used for searching for the motion vector of the screen calculated in advance with a preset threshold, and comparing the sum calculated in advance according to the comparison result. Search for a motion vector in which the value of the sum of absolute differences is minimized by shifting from the normal search start coordinate by the output of the comparator that outputs one of the screen motion vector and a preset motion vector threshold value. 13. The motion detection method according to claim 10, wherein the motion detection is performed. 14. When the motion vector is large and the difference absolute value sum is large, a preset threshold value of the motion vector is stored in the storage means instead of the difference absolute value sum. 14. The motion detection method according to claim 12, wherein: 15. The method according to claim 10, wherein when the value of the sum of absolute differences is equal to the value of the motion vector, a value close to the previously calculated motion vector at the next position is adopted. The motion detection circuit according to claim 14. 16. The motion detection circuit according to claim 12, wherein a motion vector at a position where writing is not performed in said storage means is set to 0.