JP2002354483A - Motion vector detecting method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To detect the generation of scene change on the basis of a motion vector. SOLUTION: An input video signal from an input terminal 1 is fed to one input terminal of a difference circuit 3 via a one-frame delay circuit 2. To the other input terminal of the circuit 3, the input video signal is directly fed to compute a differential value between the input video signal and a one-frame delay signal. An absolute value is determined by an absolute value circuit 5 with respect to the differential value, the differential value between frames accumulated in a accumulating circuit 6 is fed to a normalization circuit 7. The differential value between frames is fed to a one-field delay circuit 8 to be delayed by one field, and is fed to one input terminal of a differential circuit 9. To the other input terminal of the circuit 9, the differential value between frames which is not delayed is fed from the circuit 7 to compute the differential value between this differential value between frames and the differential value between frames which is not delayed. A shift of the determined differential value between frames between one field is fed to a comparing circuit 10, and zero representing a possibility of the existence of scene change is outputted.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a motion vector detecting method for detecting the magnitude and direction of a motion of a moving image in a television signal, that is, a method for detecting a motion vector. And a motion vector detecting method.

[0002]

2. Description of the Related Art Motion vectors are used to improve coding efficiency in high-efficiency coding of television signals and to reduce motion discontinuity due to conversion of the number of fields in television system conversion. .

A general motion vector detection method is a method of dividing a television signal into blocks of m pixels × n lines (m and n are integers) and detecting a motion vector for each block. The pattern matching method disclosed in JP-A-55-162683 and JP-A-55-162684 and the repetitive gradient method described in JP-A-60-158786 are well known.

Further, in order to improve the accuracy of detecting a motion vector in the iterative gradient method, Japanese Patent Application Laid-Open No.
As described in Japanese Patent Publication No. 0, there is an apparatus using an initial displacement vector.

In the processing of digital video signals, it is important to detect sudden changes in video signals such as scene changes. For example, in a device that converts the number of fields, such as a television system conversion device, motion is corrected using a motion vector, but when there is a sudden change in a video signal such as a scene change, the motion vector detection is normally performed. Will not be done. Therefore, when a sudden change in the video signal such as a scene change is detected, it is necessary to reset the motion correction.

[0006]

However, in the above-mentioned conventional example, a scene change is detected first, and a motion vector, a motion correction and the like are controlled using the result. Since the time for the field is required, it is necessary to delay the video signal by that amount.

In a general television system converter, the delay time between input and output video signals needs to be suppressed at most to 1 frame (2 fields) or less. That is, as long as it is about one frame or less, the audio signal system does not need to be delayed as in the video signal processing system, but if the difference between the audio signal and the video signal is two or more fields, a video delay is perceived. Therefore, it is necessary to delay the audio signal.

However, in a device having a scene change detection circuit and a motion vector detection circuit separately, as in a conventional television system conversion device, a video signal delay of three or more fields occurs, and an audio signal is also lost. There is a problem that it is necessary to delay and the device becomes complicated and expensive.

An object of the present invention is to solve the above-mentioned problems,
An object of the present invention is to provide a motion vector detecting method in which a scene change is detected with a delay time of one field or less by detecting a scene change in the detection of a motion vector, thereby reducing a motion vector detection error.

[0010]

To achieve the above object, a motion vector detecting method according to the present invention comprises at least one method.
Motion detection between two video signals separated by a frame or one field is performed in units of M pixels × N lines of blocks (M and N are integers) to obtain a motion detection signal, and the motion detection signal and at least one field or more are detected. Calculating a difference value from the delayed motion detection signal; and using the two video signals separated by at least one field or more, calculating the motion vector in block units of O pixels × P lines (O and P are integers). When the motion vector detection operation to be detected is performed by delaying N lines from the motion detection operation, and the two video signals are A and B, respectively, a motion vector Va is detected based on A, and the motion based on B is detected. Detecting the vector Vb;
When the difference value of the motion detection signal exceeds a threshold value α close to zero, and the motion vector is Va + V on the detected block.
forcing the detected motion vector to 0 when b exceeds a threshold β close to zero.

Further, the motion vector detecting method according to the present invention detects a motion between two video signals separated by at least one frame or one field in units of blocks of M pixels × N lines (M and N are integers). Performing a motion detection signal, calculating a difference value between the motion detection signal and the motion detection signal delayed by at least one field, and using two video signals separated by at least one field,
In a motion vector detection method using an initial displacement vector in units of O pixels × P lines (O and P are integers), at least one field whose position is displaced by the magnitude of the selected initial displacement vector value Calculating a difference value C that is distant, and calculating a difference value D that is at least one field apart by a position shifted by the magnitude of the motion vector after detection;
Forcibly setting the detected motion vector to 0 when the difference value of the motion detection signal exceeds a threshold α close to zero and the difference CD between the difference values is smaller than a threshold γ close to zero. And characterized in that:

[0012]

FIG. 1 is a block circuit diagram showing the configuration of a first embodiment of an apparatus for implementing a motion vector detecting method according to the present invention. The input from the input terminal 1 is connected to a one-frame delay circuit 2, a difference circuit 3, and a delay circuit 4. The output of the one-frame delay circuit 2 is connected to a difference circuit 3 and a delay circuit 4, and the output of the difference circuit 3 is connected to an absolute value circuit 5, an accumulation circuit 6, and a normalization circuit 7 in that order. The output of the normalizing circuit 7 is connected to a one-field delay circuit 8 and a difference circuit 9, the output of the one-field delay circuit 8 is connected to a difference circuit 9, and the output of the difference circuit 9 is a comparison circuit 10, a scene change detection. It is connected to the circuit 11. The output of the delay circuit 4 is connected in parallel to a first motion vector detection circuit 12 and a second motion vector detection circuit 13, and the outputs of the detection circuits 12 and 13 are sent to a scene change detection circuit 11 via an addition circuit 14. The output of the detection circuit 11 is connected to the output terminal 16.

When an input video signal to be processed is supplied to an input terminal 1, the input video signal is sent to a one-frame delay circuit 2.
, And is supplied to one input terminal of the difference circuit 3. The other input terminal of the difference circuit 3 is directly supplied with an input video signal from the input terminal 1, and calculates a difference value between the input video signal and the one-frame delay signal. This difference value is supplied to an absolute value circuit 5, and the absolute value of the difference value is obtained.
To supply. In the present embodiment, the accumulation circuit 6 performs an accumulation operation for each block of 64 pixels × 32 lines.

The inter-frame difference values accumulated by the accumulation circuit 6 are supplied to a normalization circuit 7. The normalizing circuit 7 is a circuit for shortening the bit length of the inter-frame difference value output from the accumulating circuit 6 for each block by a limiter circuit, a non-linear circuit, or the like. In this way, the inter-frame difference value having the reduced bit length is supplied to the one-field delay circuit 8, the normalized inter-frame difference value is delayed by one field, and supplied to one input terminal of the difference circuit 9. . The other input terminal of the difference circuit 9 is supplied with a non-delayed inter-frame difference value from the normalization circuit 7, and outputs the inter-frame difference value output from the normalization circuit 7 and the inter-frame difference value delayed by one field. , That is, the amount of shift between one field of the inter-frame difference value.

The amount of shift between the fields of the inter-frame difference value thus obtained is supplied to the comparison circuit 10.
In the comparison circuit 10, when a shift amount of the inter-frame difference value between one field exceeds a threshold α close to zero (a value close to 0 is used as an example of the value of α), a scene change exists. It outputs 0 indicating the possibility of performing. If there is no sudden change such as a scene change in the video signal, the amount of shift on the time axis of the inter-frame difference value is small,
If there is a sudden change, the amount of shift of the inter-frame difference value on the time axis increases. Therefore, in the present embodiment,
The comparison circuit 10 compares the shift amount of the inter-frame difference value between one field with a threshold α close to zero, and when the threshold α is exceeded, judges that there is a possibility that a scene change exists, and sets 0 to 0. Output.

The circuit parts of the one-frame delay circuit 2, the difference circuit 3, the absolute value circuit 5, and the accumulator circuit 6 are general motion detection circuits.
A median filter or LPF can be inserted.

The above frame delay circuit 2 to comparison circuit 10
The serial circuit excluding the delay circuit 4 is a partial scene change detection circuit, and is a method of imitating a scene change when the shift amount of the difference value between one frame on the time axis changes suddenly. However, such a method is disclosed in Japanese Patent Application No. 10-1441.
No. 47, the details are omitted.

The calculation time required for the scene change detection is 3 times in this embodiment because the total calculation is performed for each block having a size of 64 pixels × 32 lines.
2 lines + ε. Here, ε is a delay time including a delay amount due to a circuit element and a short delay amount in one line.

Ideally, the above-described scene change detection is performed over one entire screen, but it takes more than one field to detect the scene change. Therefore, in the present embodiment, basically, a motion vector without malfunction is detected even at the time of a scene change, using the above-mentioned 32 lines of information and the information to be described.

As shown in FIG. 1, an input video signal from an input terminal 1 and a video signal delayed by one frame by a one-frame delay circuit 2 are passed through a delay circuit 4 to a first and second motion detection circuits 12. , 13 to execute the operation of motion vector detection with a delay of 32 lines from the scene change detection operation of the one-frame delay circuit 2 to the comparison circuit 10 described above. The delay circuit 4 is a RAM necessary for detecting a motion vector.
However, in order to clarify that the operation is performed with a delay of 32 lines from the scene change detection, a separate delay circuit is described.

The difference between the first motion vector detection circuit 12 and the second motion vector detection circuit 13 is that the reference field is different. That is, the first motion vector detection circuit 12 detects the motion vector between the input video signal A of the input terminal 1 and the one-frame delay signal B, while the second motion vector Detection circuit 1
In No. 3, the motion vector is detected between the one-frame delay signal B and the input video signal A.
Thus, the first and second motion vector detection circuits 1
By providing 2 and 13 to detect the motion vector in both directions, the accuracy of detecting the motion vector can be improved.

The two motion vectors detected by the first and second motion vector detection circuits 12 and 13 are added by an addition circuit 14. Although the motion vector detection in the first and second motion vector detection circuits 12 and 13 is performed on a block basis, if the coordinates of the detected blocks are the same (but one frame apart), the first motion vector The motion vector detected by the detection circuit 12 is V12, the second
Assuming that the motion vector detected by the motion vector detection circuit 13 is V13, if the motion is a uniform motion such as a panning scene, these motion vectors have substantially the same magnitude and opposite signs. V12 + V13
Approaches zero. On the other hand, when a motion vector is erroneously detected due to a scene change or the like, V12 + V13 has a large value. Therefore, when V12 + V13 is larger than a threshold value β (integer) close to zero, it is determined that the motion vector detection is erroneous and 0 is output.

In the present embodiment, the result of detecting a scene change by the frame delay circuit 2 to the comparison circuit 10 is:
The result of the motion vector evaluation by the delay circuit 4 to the adder circuit 14 is supplied to the scene change detection circuit 11 to take a logical product condition of the two. Therefore, when the shift amount of the inter-frame difference value in one field exceeds the threshold value α, the comparison circuit 10 outputs 0, and the addition circuit 14 outputs V12 +
When 0 is output when V13 is larger than the threshold value β, an output signal for determining that there is a scene change is output from the scene change detection circuit 11 to the output terminal 16. As described above, when the scene change detection signal is output to the output terminal 16, by forcibly setting the motion vector to zero, it is possible to detect a motion vector without erroneous detection due to a scene change.

FIG. 2 is a block circuit diagram showing the configuration of the second embodiment. In this embodiment, the method of evaluating the detection of a motion vector is different from that of the first embodiment described above. However, the configuration of the scene change detection circuit from the frame delay circuit 2 to the comparison circuit 10 is This is the same as the first embodiment, and the output of the comparison circuit 10 is connected to the determination circuit 20.

The output of the input terminal 1 and the output of the one-frame delay circuit 2 are an initial deviation vector selection circuit 21, a deviation vector detection circuit 22, a first inter-frame difference value generation circuit 23, and a second inter-frame difference value, respectively. It is sequentially connected to the generation circuit 24. The output of the initial displacement vector selection circuit 21 is connected to a displacement vector detection circuit 22, an addition circuit 25, and a first inter-frame difference value generation circuit 23.
The output of the displacement vector detection circuit 22 is supplied to a second inter-frame difference value generation circuit 23 and a determination circuit 2 via an addition circuit 25.
0, and the outputs of the first and second inter-frame difference value generation circuits 23 and 24 are output to the determination circuit 20 via the comparison circuit 26.
And the output of the determination circuit 20 is connected to the output terminal 16.

In this embodiment, as a method for reducing erroneous detection of a motion vector at the time of a scene change, a motion vector detecting method using an initial displacement vector described in Japanese Patent Application Laid-Open No. Sho 62-206980 is used. To use. The motion vector detection method itself using the initial displacement vector is described in the above-mentioned publication, and thus the description thereof is omitted, but the point is that the optimal one from the already detected motion vectors around the detected block is used as the initial displacement vector, In this method, the coordinates are shifted by the amount of the vector and the motion vector is detected, and the detection accuracy is improved by detecting the motion vector in two steps.

As shown in FIG. 2, the input video signal from the input terminal 1 and the one-frame delay signal output from the one-frame delay circuit 2 are used to select an initial displacement vector which constitutes a part of the motion vector detection circuit. It is supplied to a circuit 21 and a shift vector detection circuit 22. In this embodiment, the one-frame delay circuit 2 constituting the scene change detection circuit is used. However, a separate delay circuit may be provided in the motion vector detection circuit. In this case, the one-field delay circuit may be used. Good.

The output of the initial displacement vector selecting circuit 21 and the output of the displacement vector detecting circuit 22 are added by an adding circuit 25 to obtain a true motion vector. Further, a first inter-frame difference value generation circuit 23 is provided, and the coordinates of the inter-frame difference value between the input video signal of the input terminal 1 and the one-frame delay signal output from the one-frame delay circuit 2 are set to the initial deviation vector. An inter-frame difference value F1 shifted by the initial shift vector output from the selection circuit 21 is generated. Further, a second inter-frame difference value generation circuit 24 is provided, and the coordinates of the inter-frame difference value between the input video signal of the input terminal 1 and the one-frame delay signal output from the one-frame delay circuit 2 are calculated by the addition circuit 25. An inter-frame difference value F2 shifted by the output true motion vector is generated.

Comparing the outputs of the first and second inter-frame difference value generating circuits 23 and 24, if there is no motion or there is a connected motion, F1> F2, If there is a sudden movement, F1 <
It becomes F2. Therefore, when F1 <F2 is detected, it can be determined that the motion vector is not normally detected.

In this embodiment, this condition is used, and the first and second inter-frame difference value generation circuits 2 are used.
The two signals F1 and F2 output from 3 and 24 are supplied to a comparison circuit 26, and F1-F2 is compared with a threshold γ close to zero. When F1−F2 <γ, the output of the comparison circuit 26 is set to 0 and supplied to the determination circuit 20. The determination circuit 20 is supplied with the output signal of the comparison circuit 10 and outputs the true motion vector output from the addition circuit 25 that adds the output of the initial displacement vector selection circuit 21 and the output of the displacement vector detection circuit 22. Also supply. Accordingly, 0 is supplied from the comparison circuit 10 and 0 is supplied from the comparison circuit 26.
Is supplied, the motion vector output from the output terminal 16 is forcibly set to 0.

[0031]

As described above, according to the motion vector detecting method of the present invention, a scene change is detected in a time of one field or less, and when a possibility that a motion vector is erroneously detected is detected. By forcibly setting the motion vector to 0, it is possible to reduce the error of the motion vector detection.

The general motion vector detection is to detect the amount of motion of both moving objects using signals of one field (or one frame), and it is assumed that a moving object exists in both fields. It is. However, at the time of the scene change, there is no moving object, but the difference between the fields is large, so that the motion vector is erroneously detected. According to the present invention, erroneous detection of a motion vector can be reduced by performing a scene change detection and forcibly setting a motion vector to a motion vector detection method in a scene change when the scene change is detected.

[Brief description of the drawings]

FIG. 1 is a block circuit configuration diagram for implementing a first embodiment.

FIG. 2 is a block circuit configuration diagram for implementing a second embodiment.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Input terminal 2 1 frame delay circuit 3 Difference circuit 4 Delay circuit 5 Absolute value circuit 6 Cumulative circuit 7 Normalization circuit 8 1 field delay circuit 9 Difference circuit 10, 26 Comparison circuit 11 Detection circuit 12 First motion vector detection circuit 13 Second motion vector detection circuit 14, 25 Addition circuit 16 Output terminal 20 Judgment circuit 21 Initial displacement vector selection circuit 22 Displacement vector detection circuit 23 First inter-frame difference value generation circuit 24 Second inter-frame difference value generation Circuit 25 Addition circuit

 ────────────────────────────────────────────────── ─── Continued on the front page F term (reference) 5C059 KK19 NN01 NN10 NN20 NN28 NN43 5C063 AA01 BA04 BA12 5J064 AA01 BB01 BB03 BC08 BC14 BD01 5L096 CA04 DA02 GA08 GA19 HA04 LA10

Claims (2)

[Claims] 1. A motion detection signal is obtained by detecting a motion between two video signals separated by at least one frame or one field in blocks of M pixels × N lines (M and N are integers). Calculating a difference value between the detection signal and the motion detection signal delayed by at least one field; and using two video signals separated by at least one field or more, an O pixel × P line (O and P are integers) When a motion vector detecting operation for detecting the motion vector in block units is performed by delaying N lines from the motion detecting operation, and the two video signals are A and B, respectively,
Detecting a motion vector Va based on A and detecting a motion vector Vb based on B; and determining that the difference value of the motion detection signal exceeds a threshold α close to zero, and that the motion vector When Va + Vb exceeds a threshold β close to zero, the detected motion vector is forcibly set to 0.
A motion vector detecting method. 2. Detection of motion between two video signals separated by at least one frame or one field is performed by M pixels × N
A step of obtaining a motion detection signal by performing a block (M and N are integers) of a line and calculating a difference value between the motion detection signal and the motion detection signal delayed by at least one field; In a motion vector detection method using an initial displacement vector in units of O pixels × P lines (O and P are integers) using two video signals, only the size of the selected initial displacement vector value is used. A step of obtaining a difference value C which is shifted by one or more fields whose position is deviated, and a step of obtaining a difference value D which is shifted by one or more fields whose position is shifted by the magnitude of the motion vector after detection; The difference value exceeds a threshold α close to zero, and the difference CD between the difference values is a threshold γ close to zero.
Forcibly setting the detected motion vector to 0 when the motion vector is smaller than 0.