JP2003051965A - Image motion detecting circuit - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image motion detecting circuit that can highly accurately detect motion of an image sequence with different static images are connected with each other, and to provide an image motion detecting method. SOLUTION: Inter-frame difference absolute values of an inputted image signal during a predetermined frame period are held by a plurality of registers 105. Then the maximum value is extracted from the inter-frame difference absolute values which are held by the plurality of registers 105, and the extracted maximum value is subtracted from the sum of the inter-frame difference absolute values during the predetermined frame period, thereby detecting the motion of the whole image sequence.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image motion detection circuit and an image motion detection method for detecting the amount of motion of an entire image sequence such as a television signal.

[0002]

2. Description of the Related Art As a conventional technique relating to a motion detection circuit for a television signal, for example, one described in Japanese Patent No. 2585544 is known. FIG. 4 is a block diagram showing the configuration of the motion detection circuit described in the above publication. This motion detection circuit, as shown in FIG.
A frame memory 402, a subtractor 403, a motion information conversion circuit 404, an integration circuit 405, and an output terminal 406 are provided. The television signal input to the input terminal 401 is input to the frame memory 402. Subtraction circuit 40
In 3, the difference between the input and output of the frame memory 402, that is, the difference between frames is calculated. Then, the motion information conversion circuit 4
04 converts the inter-frame difference signal into a signal representing motion. The integration circuit 405 integrates motion information of the same pixel or peripheral pixels over a past finite frame period. By the above operation, the conventional motion detection circuit
The motion information can be output from the output terminal 406.

[0003]

However, in the above-mentioned conventional motion detection circuit, since the absolute value of the inter-frame difference is converted into motion information, the motion information is integrated over a predetermined period in the past. When a certain still image is connected to another still image, a large inter-frame difference value is calculated even when a so-called scene change occurs. As a result, the motion of the entire image sequence is detected as a large motion, which leads to deterioration of motion detection accuracy.

This problem does not become a big problem when the integration cycle is long, but it becomes a big problem as the integration cycle becomes shorter. Further, since the integration cycle corresponds to the time required for motion detection, there is also a demand for shortening the integration cycle from the viewpoint of memory reduction.

Therefore, the present invention provides an image motion detection circuit and an image motion detection method capable of accurately performing motion detection even in an image sequence in which different still images are connected to each other. To aim.

[0006]

An image motion detection circuit according to claim 1 of the present invention comprises first extracting means for extracting an interframe difference absolute value from an input image signal in frame units,
Holding means for holding N (N is a natural number of 4 or more) of the inter-frame difference absolute values, one inter-frame difference absolute value extracted by the first extracting means, and the holding means. Second extracting means for extracting M (M is a natural number not less than 1 and not more than N / 4) inter-frame difference absolute values from the largest of the N inter-frame difference absolute values;
The absolute value of the inter-frame difference of +1 is calculated by the extracted M
The present invention is characterized in that it has an adding means for adding, except for the absolute difference between frames.

The image motion detection circuit according to claim 2 of the present invention further comprises: first extracting means for extracting the absolute value of the inter-frame difference in frame units from the input image signal; and horizontally adjoining in frame units from the input image signal. Second extraction means for extracting a cumulative addition value of the pixel difference values; calculation means for calculating a frame motion amount from the absolute value of the inter-frame difference value and the cumulative addition value of the horizontal adjacent pixel difference values; A natural number greater than or equal to 4) holding means for holding the number of frame movement amounts; one frame movement amount calculated by the calculating means and N frame movement amounts held by the holding means, which is larger. Third extraction means for extracting M (M is a natural number not less than 1 and not more than N / 4) frame motion amounts from the other side;
And an addition unit for adding +1 frame motion amount excluding the extracted M frame motion amounts.

The image motion detection method according to claim 3 of the present invention further comprises a first extraction step of extracting an absolute value of inter-frame difference from the input image signal in frame units, and N (N
Is a natural number greater than or equal to 4) a holding step for holding the interframe difference absolute values, and one interframe difference absolute value extracted in the first extracting step and N frames held in the holding step. A second extraction step of extracting M (M is a natural number not less than 1 and not more than N / 4) inter-frame difference absolute values from the largest of the difference absolute values.
An addition step of adding the N + 1 number of inter-frame difference absolute values excluding the extracted M number of inter-frame difference absolute values.

According to a fourth aspect of the present invention, in the image motion detecting method, a first extracting step of extracting an inter-frame difference absolute value from the input image signal in frame units, and a horizontal adjacency in frame units from the input image signal. A second extraction step of extracting a cumulative addition value of pixel difference values; a calculation step of calculating a frame motion amount from the absolute value of the inter-frame difference value and a cumulative addition value of the horizontal adjacent pixel difference values; A natural number greater than or equal to 4) a holding step of holding the number of frame movements, one of the frame movements calculated in the calculation step and the N number of frame movements held in the holding step, whichever is greater. M (M is 1 or more, N /
Third natural number (4 or less) to extract frame motion amount
And extracting the N + 1 frame motion amounts, excluding the extracted M frame motion amounts,
And an adding step of adding.

[0010]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of an image motion detection circuit of the present invention will be described below with reference to FIGS. (First Embodiment) FIG. 1 is a block diagram showing a configuration example of an image motion detection circuit 100 according to the first embodiment of the present invention. The image motion detection circuit 100 includes a frame memory 101 that delays an input image signal for one frame period, a subtractor 102 that calculates a level difference between frames of pixels existing in the same spatial position of the input image signal, and a subtractor 102. Absolute value conversion circuit 1 for obtaining the absolute value of the level difference calculated for each pixel
03, a cumulative addition circuit 104 that resets the cumulative addition result at the head of the frame and cumulatively adds the absolute value in frame units to obtain a cumulative added value, and N (N is a natural number of 4 or more) in frame units. N registers 105 for holding the cumulative addition value of the absolute difference between frames, and the cumulative addition circuit 1
One cumulative addition value output from 04 and N registers 10
5 and the singularity extraction circuit 106 that extracts M (M is a natural number of 1 or more and N / 4 or less) cumulative addition values from the largest of the N cumulative addition values held by 5 at the beginning of the frame. Prepare In addition, in the following description, the cumulative addition value is
Unless otherwise specified, it means the cumulative addition value of the inter-frame difference absolute values. Also, the image motion detection circuit 1
00 is a cumulative addition value of one M that is output by the cumulative addition circuit 104 and N cumulative addition values held by the N registers 105. Integrating circuit 10 is used as means for adding values excluding values.
7 and the subtractor 108. The integrating circuit 107 adds one cumulative addition value output from the cumulative addition circuit 104 and the N cumulative addition values held by the N registers 105 to obtain the total value. The subtractor 108 is the integration circuit 1
The cumulative addition values for M extracted by the singularity extraction circuit 106 are subtracted from the total value of the cumulative addition values obtained by 07.

Since the minimum number of frames that can effectively recognize a scene change due to the connection of two image sequences is generally 6, N is preferably a natural number of 4 or more. Therefore, in the first embodiment, an example will be described in which the image motion detection circuit 100 includes four registers 105 (registers 105a to 105d) and detects a motion amount in a 6-frame period.

The operation of the image motion detection circuit 100 configured as above will be described in detail. First, an image signal is input. The input image signal is delayed by one frame period in the frame memory 101, and the subtractor 102
Entered in. The subtractor 102 calculates the level difference between pixels existing in the same spatial position from the input image signal and the output of the frame memory 101. The absolute value conversion circuit 103 converts the pixel level difference value between the frames calculated by the subtractor 102 into an absolute value. A cumulative addition circuit 104 cumulatively adds the absolute values in units of frames to calculate a cumulative addition value of inter-frame difference absolute values. After that, the cumulative addition value calculated by the cumulative addition circuit 104 is sequentially shifted and held for each frame by the four registers 105a to 105d connected in series. As a result of the above operation, the four cumulative addition values held in the registers 105a to 105d and the one cumulative addition value output from the cumulative addition circuit 104 are both the singular point extraction circuit 106 and the integration circuit. 107
Entered in.

The singularity extraction circuit 106 extracts M cumulative addition values from the largest of the five input cumulative addition values. Considering a scene change due to the connection of two image sequences over N + 1 frame periods, M
Since it is effective that the value of 1 is a natural number of 1 or more and N / 4 or less, the value of M becomes 1 when detecting the motion amount in the 6-frame period. Therefore, the singular point extraction circuit 106
Extracts the maximum value of the five cumulative addition values. In this case, for example, in an image sequence in which still images are connected to each other, only the cumulative addition value of the absolute values of inter-frame differences at the time of a scene change in which different still images are continuous becomes a very large value. Therefore, the extracted maximum value corresponds to the cumulative addition value of the absolute value of the inter-frame difference at the scene change. On the other hand, the integration circuit 107
Calculates the total value of the five cumulative addition values.

The maximum value extracted by the singularity extraction circuit 106 and the total value calculated by the integration circuit 107 are
It is input to the subtractor 108. The subtractor 108 subtracts the maximum value from the total value of the five cumulative addition values. As a result, the total value of the four cumulative addition values excluding the maximum value is calculated. The image motion detection circuit 100 outputs this total value as the amount of motion of the entire image sequence.

As described above, in the image motion detection circuit 100, when the motion amount in the 6-frame period is obtained, the cumulative addition value of the inter-frame difference absolute values in the 6-frame period is added excluding the maximum value to obtain the image sequence. Since the total amount of motion is detected, when the amount of motion is calculated for an image sequence in which still images are connected to each other, the amount of motion that ignores the scene change portion can be detected. It should be noted that in an image sequence in which similar motions are continuous such as continuous still images or continuous moving images, even if one of the cumulative addition values of the plurality of inter-frame difference absolute values is ignored, the detected image is detected. It does not affect the amount of movement of the entire sequence.

An image motion detecting method using the image motion detecting circuit 100 according to the first embodiment will be described with reference to FIG. First, each data in the register 105, the cumulative addition circuit 104, and the frame memory 101 is initialized to 0 (step S301). Note that the cumulative addition values held in the registers 105a to 105d in FIG. 1 are, in order from the register 105a, frame1, frame2, and frame, respectively.
The following description will be given assuming that the cumulative addition value is 3 and frame4, and the cumulative addition value held in the frame unit by the cumulative addition circuit 104 is frame0.

Next, an image signal is input in frame units,
The frame start of the input image signal is determined (step S3)
02). If the result of determination is that the frame is not at the beginning, the inter-frame difference, which is the difference value between the level of the pixels forming one frame of the input image signal and the pixel level of the image signal delayed by one frame period in the frame memory 101. The value is calculated by the subtractor 102, the difference value is converted into an absolute value by the absolute value conversion circuit 103, and the absolute value of the interframe difference (pixel_
abs_diff) is calculated (step S307). This pixe
l_abs_diff is constantly calculated for pixels for one frame period. Then, in the cumulative addition circuit 104, pixel_abs_
The diff is cumulatively added to calculate the cumulative addition value (frame_abs_diff) of the inter-frame difference absolute value (step S308).
This frame_abs_diff is used when the cumulative addition circuit 104 cumulatively adds pixel_abs_diff over one frame period.

On the other hand, when the result of the judgment in step S302 is the head of the frame, the cumulative addition circuit 104
The frame_abs_diff calculated in step 1 is stored as frame0 in frame units. Furthermore, the cumulative addition values held in frame units are sequentially stored by the registers 105a to 105d.
Shift and hold. In other words, this shift operation causes frame0 to be retained in frame units.
, Frame1 to frame2, frame2 to frame3, frame3 to fr
It will be substituted into ame4 (step S303).

Next, the cumulative addition circuit 104 sets frame0, and the registers 105a to 105d set frame1, frame2, fr
The maximum value (frame_max) is extracted by the singularity extraction circuit 106 from the cumulative addition values of the inter-frame difference absolute values held as ame3 and frame4 (step S304). Then, the total value of the five cumulative addition values (frame0 to frame4) is obtained by the integration circuit 107, and the maximum value extracted by the singularity extraction circuit 106 is subtracted by the subtractor 108 from the total value obtained by the integration circuit 107. Thus, the amount of movement of the entire image is detected (step S305). Finally, frame_ab, which is constantly used for calculation for cumulative addition of absolute values of inter-frame differences.
Reset s_diff to 0.

As described above, according to the image motion detection circuit or the image motion detection method according to the first embodiment, the maximum value is extracted from the absolute value of the inter-frame difference for the predetermined frame period of the input image signal. In addition, the absolute value of the inter-frame difference for a predetermined frame period is added excluding the maximum value to detect the motion of the entire image sequence, so that the peculiar value calculated at the scene change is excluded. You can detect the movement of the entire image sequence,
The motion detection accuracy of the image can be improved.

In the first embodiment, when detecting the motion amount in the 6-frame period, the cumulative addition values of the five inter-frame difference absolute values are added except the maximum value, and the total value is added to the image. Although the case of detecting as the amount of motion of the entire sequence has been described, in the case of a system in which the detection delay can be made longer, the image motion detection circuit 100 uses more cumulative addition values to detect the motion of the entire image sequence. The amount can be detected. If this detection delay can be lengthened, the number of scene changes increases, so it is necessary to remove some values from the largest value to detect the amount of motion, instead of removing only the maximum value. For example, 1
When the motion amount of the entire image sequence is detected by using the cumulative addition value of 0 inter-frame difference absolute values, the one obtained by removing the two cumulative addition values from the larger one is the motion of the entire image sequence. Detect as quantity. In this case, the image motion detection circuit 100 shown in FIG.
05, the singularity extraction circuit 106 extracts two cumulative addition values from the larger of the 10 cumulative addition values of interframe difference absolute values, and the integration circuit 107 extracts 10 interframe difference absolute values. Calculate the sum of the cumulative additions of the values. Then, the subtractor 108 subtracts the total value of the extracted two cumulative addition values from the total value of the ten cumulative addition values to detect the motion amount of the entire image sequence.

(Second Embodiment) FIG. 2 is a block diagram showing a configuration of an image motion detection circuit 200 according to a second embodiment of the present invention. The same components as those of the image motion detection circuit 100 shown in FIG. In the second embodiment, as in the first embodiment, an example will be described in which the image motion detection circuit 200 includes four registers 105 (registers 105a to 105d) and detects a motion amount in a 6-frame period. To do.

The image motion detection circuit 200 according to the second embodiment differs from the image motion detection circuit 100 in that it includes a horizontal adjacent pixel difference calculation circuit 201 and a frame motion amount calculation circuit 202 as shown in FIG. . The horizontal adjacent pixel difference calculation circuit 201 cumulatively adds the horizontal adjacent pixel differences of the input image signal frame by frame to obtain a cumulative addition value (H). The frame motion amount calculation circuit 202 corrects the cumulative addition value (F) of the inter-frame difference absolute values by the cumulative addition value (H) to calculate the frame motion amount.

The operation of the image motion detection circuit 200 according to the second embodiment will be described in detail. First, an image signal is input. The input image signal is delayed by one frame period in the frame memory 101 and input to the subtractor 102. The subtractor 102 calculates the level difference between pixels existing in the same spatial position from the input image signal and the output of the frame memory 101. The absolute value conversion circuit 103 is a subtractor 1
The pixel level difference value between the frames calculated by 02 is converted into an absolute value. The cumulative addition circuit 104 cumulatively adds the absolute values frame by frame to obtain a cumulative addition value (F) of the interframe difference absolute values. Horizontal adjacent pixel difference calculation circuit 201
Calculates a horizontal adjacent pixel difference from the input image signal of each frame and cumulatively adds the difference in frame units to calculate a cumulative addition value (H).

The frame motion amount calculation circuit 202 inputs the cumulative addition value (F) and the cumulative addition value (H) and calculates the motion amount in frame units. This calculation is, for example, F / H
(However, H is other than 0) can be performed by obtaining the ratio of F and H. As a result, when an image with a high spatial frequency (that is, H is large) moves even a little, the value of F becomes very large, while an image with a low spatial frequency (that is, H is small) is displayed. In this case, it can be corrected that the value of F becomes small even with the same movement.

The frame-by-frame motion amount calculated by the frame-motion-amount calculating circuit 202 is sequentially shifted and held frame by frame by the registers 105a to 105d connected in series. As a result of the above operation,
The four frame motion amounts held in the registers 105a to 105d and one frame motion amount output from the cumulative addition circuit 104 are input to the singularity extraction circuit 106 and the integration circuit 107.

The singular point extraction circuit 106 extracts M cumulative addition values from the largest of the five input cumulative addition values. Considering a scene change due to the connection of two image sequences over N + 1 frame periods, M
Since it is effective that the value of 1 is a natural number of 1 or more and N / 4 or less, the value of M becomes 1 when detecting the motion amount in the 6-frame period. In this case, for example, in an image sequence in which still images are connected to each other, only a frame motion amount at a so-called scene change in which different still images are continuous has a very large value. Therefore, the extracted maximum value corresponds to the frame motion amount at the scene change. On the other hand, the integration circuit 107 calculates the total value of the five input frame motion amounts.

The maximum value extracted by the singularity extraction circuit 106 and the total value calculated by the integration circuit 107 are
It is input to the subtractor 108. The subtractor 108 subtracts the maximum value from the total value of the five frame motion amounts. As a result, 4 of the 5 frame motion amounts excluding the maximum value
The total value of the frame motion amounts is calculated. The image motion detection circuit 200 outputs this total value as the amount of motion of the entire image sequence.

Next, an image motion detection method performed by using the image motion detection circuit 200 according to the second embodiment will be described with reference to FIG. This image motion detection method differs from the image motion detection method described in the first embodiment in the following points. First, the processing is different when it is determined in step S302 shown in FIG. 3 that the frame is not the head. That is, the process of step S307 is performed, the horizontal adjacent pixel difference calculation circuit 201 calculates the cumulative addition value of the horizontal adjacent pixel differences of the input image signal, and the frame motion amount calculation circuit 202 calculates the frame_abs_diff calculated in step S308 and the horizontal adjacent pixel. The frame motion amount is obtained from the cumulative addition value of the pixel differences. The difference is that the processes of steps S303 to S306 described in the first embodiment are performed on the frame motion amount.

As described above, according to the image motion detection circuit or the image motion detection method according to the second embodiment, the cumulative sum of the absolute value of the inter-frame difference and the horizontal adjacent pixel difference in frame units of the input image signal. From, by extracting the amount of motion in frame units, out of the frame motion amount for a predetermined frame period, the maximum value is extracted, the frame motion amount for a predetermined frame period is added excluding the maximum value, By detecting the movement of the entire image sequence,
The motion of the entire image sequence can be detected except for the peculiar value calculated at the time of scene change, and the motion detection accuracy of the image can be improved.

In the second embodiment, when detecting the motion amount in the 6-frame period, the motion amount of 5 frames is added excluding the maximum value, and the total value is taken as the motion amount of the entire image sequence. Although the case of detection is described, in a system in which the detection delay can be made longer, the image motion detection circuit 200 can detect a motion amount of the entire image sequence by using a larger amount of frame motion. it can. If this detection delay can be lengthened, the number of scene changes increases, so it is necessary to remove some values from the largest value to detect the amount of motion, instead of removing only the maximum value. For example, when the motion amount of the entire image sequence is detected using 10 frame motion amounts, the value obtained by removing two values from the largest frame motion amount is detected as the motion amount of the entire image sequence. In this case, the image motion detection circuit 200 shown in FIG.
Is provided with nine registers 105, and the singularity extraction circuit 10
In 6, the frame motion amount of 2 frames is extracted from the largest frame motion amount of 10 frame motion amounts, and the integration circuit 107 calculates the total value of the 10 frame motion amounts. Then, the subtractor 108 detects the motion amount of the entire image sequence by subtracting the total value of the extracted two frame motion amounts from the total value of the 10 frame motion amounts.

[0032]

As described above, according to the present invention, the absolute value of the inter-frame difference for the predetermined frame period of the input image signal is added excluding the predetermined number of inter-frame difference absolute values from the larger one. Since the motion of the entire image sequence is detected by doing so, even if the image sequence is such that different still images are connected, it is possible to ignore the connection point and detect the motion of the image sequence. ,
As a result, the effect that the motion detection accuracy of the entire image can be improved is obtained. In particular, when the period required for motion detection of the image sequence needs to be shortened, its effectiveness is very large.

According to the present invention, the input image signal of
The motion amount in frame units is calculated from the absolute value of the inter-frame difference and the cumulative addition value of the horizontal adjacent pixel differences in frame units, and the frame motion amount for a predetermined frame period is calculated from the larger frame motion amount by a predetermined number. Except for the above, since the motion of the entire image sequence is detected by adding, the amount of motion that does not depend on the spatial frequency is calculated for each frame, and an image sequence in which different still images are connected Even in this case, it is possible to detect the motion of the image sequence by ignoring the connection point, and thus it is possible to improve the motion detection accuracy of the entire image sequence. In particular, when the period required for motion detection of the image sequence needs to be shortened, its effectiveness is very large.

[Brief description of drawings]

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

FIG. 2 is a block diagram showing a configuration of an image motion detection circuit according to a second exemplary embodiment of the present invention.

FIG. 3 is a flowchart diagram for explaining an image motion detection method according to the first embodiment of the present invention.

FIG. 4 is a block diagram showing a configuration of a conventional motion detection circuit.

[Explanation of symbols]

100,200 image motion detection circuit 101 frame memory 102, 108 Subtractor 103 Absolute value conversion circuit 104 Cumulative addition circuit 105 registers 106 Singular point extraction circuit 107 integrating circuit 201 Horizontal adjacent pixel difference calculation circuit 202 frame motion amount calculation circuit

Claims (4)

[Claims] 1. A first extracting means for extracting an interframe difference absolute value from an input image signal on a frame-by-frame basis, and a holding means for holding N (N is a natural number of 4 or more) said interframe difference absolute values. , Of the one absolute difference value between frames extracted by the first extracting means and the N absolute difference values between frames held by the holding means, M (M is 1 or more, N /
A natural number less than or equal to 4) second extraction means for extracting interframe absolute difference values; and N + 1 interframe absolute difference values, excluding the extracted M interframe absolute differences. And an adding means for adding the two. 2. A first extracting means for extracting an inter-frame difference absolute value from an input image signal on a frame-by-frame basis, and a second extraction means for extracting a cumulative addition value of horizontal adjacent pixel difference values on a frame-by-frame basis from the input image signal. Means, a calculation means for calculating a frame motion amount from the absolute value of the inter-frame difference and a cumulative addition value of the horizontal adjacent pixel difference values, and holding N (N is a natural number of 4 or more) frame motion amounts. Of the holding means, one frame motion amount calculated by the calculating means, and N frame motion amounts held by the holding means, the larger one is M (M is 1 or more and N / 4 or less). Natural number) third extraction means for extracting frame motion amounts, and N + 1 frame motion amounts for the extracted M
An image motion detection circuit, comprising: an addition unit for adding the frame motion amount for each frame. 3. A first extraction step of extracting an inter-frame difference absolute value from an input image signal in frame units, and a holding step of holding N (N is a natural number of 4 or more) said inter-frame difference absolute values. , Of the one frame-to-frame difference absolute value extracted in the first extraction step and the N-frame-to-frame difference absolute value held in the holding step, M (M is 1 or more,
A second extraction step of extracting (N / 4 or less natural number) interframe difference absolute values; and (N + 1) interframe difference absolute values, the extracted M interframe difference absolute values. And an adding step of adding. 4. A first extraction step of extracting an inter-frame difference absolute value from an input image signal in frame units, and a second extraction step of extracting a cumulative addition value of horizontal adjacent pixel difference values in frame units from an input image signal. A step of calculating a frame motion amount from the absolute value of the inter-frame difference and a cumulative addition value of the horizontal adjacent pixel difference values; and holding N (N is a natural number of 4 or more) frame motion amounts. Of the holding step, the one frame motion amount calculated in the calculating step and the N frame motion amounts held in the holding step, the larger one is M (M is a natural number of 1 or more and N / 4 or less). A third extraction step of extracting a frame motion amount for each of the N frame motion amounts;
An image motion detection method, which comprises an addition step of adding the number of frame motions.