JPH03198488A - Motion vector detector and picture oscillation corrector - Google Patents

Abstract

PURPOSE:To suppress the detection error of a motion vector in a picture as a whole even when one part of an object is not still by controlling the size of the detected motion vector by using divergency showing the degree of dispersion for the motion vector in each detection area. CONSTITUTION:An input picture signal is divided into plural areas by a motion vector detecting means 11 and the motion vector of the picture in each area is calculated. Then, the effectiveness of the motion vector in each area is decided by a motion vector deciding means 12 and the motion vector of the entire picture is determined. A divergency calculating means 13 calculates the divergency showing the degree of dispersion for the motion vector in each detection use area by using the output of the motion vector detecting means 11, and the output of the divergency calculating means 13 is converted according to the size by a converting means 15. Then, a control means 16 controls the size of the motion vector in the entire picture according to the output of the converting means 15. Thus, even when one part of the object is not still, the detection error of the motion vector in the entire picture can be suppressed.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to a motion vector detection device for detecting the amount of motion of an image and an image blur correction device.

2. Description of the Related Art An example of a conventional image motion vector detection device is disclosed in Japanese Patent Application Laid-Open No. 61-269475.

FIG. 1O shows a schematic diagram of this, and as will be described in detail later, l is to the latch, 2 is to the representative point memory,
3 is a latch B, 4 is a correlator, 5 is an address controller, [i is an address switching circuit, 7 is a cumulative adder, 8 is a correlation search circuit, 9 is a correlation validity/invalidity determination circuit, and IO is a determination circuit. .

An image motion vector detection device using the conventional correlation calculation device configured as described above will be described.

First, the motion vector of an image will be explained.

FIG. 9(a) shows an image of a triangular object at a certain time. And (b) shows an image after one field or l frame.

In this way, when an image is translated in parallel due to the movement of the imaging device itself, the vector representing the amount by which the image is translated in parallel, as shown by the arrow in (C), is called a motion vector.

FIG. 11 shows representative points and pixels around them in the representative point matching method, which is the most common method for detecting motion vectors of such images. Motion vector detection is performed by detecting where the image data at the position of a representative point in a certain field has moved among surrounding pixels in the next field.

Next, an image motion vector detection device using a conventional correlation calculation device will be described with reference to FIGS. 10 and 11.

Image data at each representative point on the screen is taken into latch AI by timing pulse LPI, and written to the address corresponding to each representative point in representative point storage memory 2 at the appropriate timing.

Then, in the next field or next frame ζ, the correlation between the image data in the motion vector detection area around the position of each representative point and the representative point of the previous field stored in the representative point memory 2ζ is calculated, and cumulative summation is performed. input into device 7. The cumulative adder 7 cumulatively adds the correlated data at the same coordinate position with respect to the representative point.

When the cumulative additions around all the representative points are completed, the correlation search circuit 8 determines the location having the highest correlation among the cumulative addition values held by the cumulative adder 7ζ. In other words, the position (address) having the highest correlation value with respect to the position of the representative point becomes the motion vector. Furthermore, based on the distribution of correlation values around the representative point (average value, minimum value, maximum value, gradient, etc.), the correlation validity/invalidity determination circuit 9 determines whether the motion vector i obtained by the correlation calculation is valid. or invalid. The above operation is
Each area when the screen is divided into multiple parts (this is done with difficulty).

Then, the motion vector of the entire screen is determined by the determination circuit 10 based on the motion vector obtained from each area of the screen and its validity determination information.

The operations up to this point are performed every field (frame), so
A latch AI is provided to store image data at representative points for the correlation calculation of the next field (frame) while performing the correlation calculation.

Further, the latch B3 is provided to hold the image data of this representative point when correlating the image data of a certain representative point with the image data around it.

In addition, the broken line part in FIG.
A motion vector determining section 12 is a section that calculates the motion vector of the entire screen from the motion vectors of each area of the screen obtained by the motion vector detecting section 11 and their correlation information.

Problems to be Solved by the Invention However, in the above configuration, it is sufficient that the subject remains stationary and only the imaging device moves, and that the motion vectors in each area are all the same (FIG. 12(a)).
If a part of the subject on the screen is not stationary and is moving differently from the movement of the entire screen (Fig. 12(b)),
The motion vectors in each area are not all the same, and if the motion vector of the entire screen is determined by taking the average ring of the motion vectors of each area, the motion vector of the area whose movement is different from the movement of the entire screen will be the cause. This had the problem of causing detection errors.

In view of this, an object of the present invention is to provide a motion vector detection device and an image blur correction device that can suppress errors in detecting motion vectors of the entire screen even when a part of the subject is not stationary. did.

Means for Solving the Problems The present invention as claimed in claim 1 provides a motion vector detection means for dividing an input image signal into a plurality of regions and calculating a motion vector of an image in each region, and a motion vector detection means for dividing an input image signal into a plurality of regions and calculating a motion vector of an image in each region. a motion vector determining means for determining validity and determining a motion vector for the entire screen; and a divergence calculation means for obtaining a divergence indicating the degree of dispersion of motion vectors in each detection area using the output of the motion vector detecting means. , a conversion means for converting the output of the divergence calculation means into a value related to the magnitude of the output, and a control means for controlling the magnitude of the motion vector of the entire screen by the output of the conversion means. This is a featured motion vector detection device.

The present invention as set forth in claim 4 provides an image enlarging device for enlarging an input image, a motion vector detecting means for computing a motion vector of the image, and determining the effectiveness of the motion vector in each region,
a motion vector determining means for determining the motion vector of the entire screen; an integrating means for time-integrating the motion vector of the entire screen; and a divergence indicating the degree of dispersion of the motion vectors of each detection area using the output of the motion vector detecting means. a divergence calculation means for calculating the degree of divergence, a conversion means for converting the output of the divergence calculation means into a value related to the magnitude of the output, and an output of the integration means or motion vector determination means based on the output of the conversion means. control means for controlling, and read-out position control means for calculating an image read-out position of the image enlarging device using the output of the control means or the integrating means, and controlling the read-out position of the enlarged image. This image blur correction device is characterized by moving the image in parallel and suppressing the blur component of the input image.

In the present invention as defined in claim 1, the motion vector detection means divides the input image signal into a plurality of regions, the motion vector of the image in each region is calculated, and the motion vector determination means calculates the motion vector of the image in each region. The effectiveness is determined, a motion vector for the entire screen is determined, a divergence calculation means uses the output of the motion vector detection means to obtain a divergence indicating the degree of dispersion of the motion vectors of each detection area, and a conversion means The output of the divergence calculation means is converted according to the magnitude of the output, and the control means controls the magnitude of the motion vector of the entire screen based on the output of the conversion means.

In this way, the magnitude of the detected motion vector is characterized by using the degree of divergence indicating the degree of dispersion of the motion vector in each detection area. The motion vector detection means calculates the motion vector of the image, the motion vector determination means determines the validity of the motion vector in each area, determines the motion vector of the entire screen, and the integration means calculates the motion vector of the entire both sides. The output of the motion vector detection means is used by the divergence calculation means to obtain the divergence indicating the degree of dispersion of the motion vectors in each detection area, and the output of the divergence calculation means is converted into the output by the conversion means. The control means controls the output of the integrating means or the motion vector determining means using the output of the converting means, and the reading position control means uses the output of the controlling means or the integrating means to enlarge the image. By calculating the image readout position of the device and controlling the readout position of the enlarged image, the image is translated in parallel and the shaking component of the input image is suppressed.

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

FIG. 1 is a block diagram of a motion vector detection device according to an embodiment of the present invention, in which reference numeral 11 denotes a motion vector detection unit as an example of motion vector detection means, and 12 denotes a motion vector determination unit as an example of motion vector determination means. The structure is similar to the conventional structure.

Further, as will be explained in detail later, 13 is a divergence calculation unit as an example of a divergence calculation unit that uses the output of the motion vector detection unit 11 to obtain a divergence indicating the degree of dispersion of motion vectors in each detection area. , 14 is a low-pass filter (LP) that passes the low frequency of the output of the divergence calculation unit 13.
F), 15 is a converter A as an example of a converting means that converts the output of 13 of the divergence calculation means according to the magnitude of the output.
, 16 is a control unit that controls the magnitude of the motion vector of the entire screen based on the output of the converter A15.

The operation of the motion vector detection device configured as above will be explained.

In FIG. 1, first, in the same manner as in the conventional example, a motion vector detecting section 11 detects motion vectors between fields or frames of an input image signal for each region of the screen, and a motion vector determining section 12 detects motion vectors in each region. The validity of the motion vectors is determined, and the motion vector for the entire screen is determined from among the motion vectors detected in the regions determined to be valid.

Next, the divergence calculation unit 13 calculates the divergence of the motion vector based on the output of the motion vector detection unit 11, that is, the motion vector of the area determined to be valid by the motion vector determination unit 12 among the motion vectors of each detection area. Calculate degrees. The calculation method is shown in the following formula.

(1) Divergence (DIV) = MAX (1 (average value of motion vectors in each region) - (motion vector in each region) 1) However, this calculation only applies to motion vectors in regions where the motion vector determination information is OK. Do about. MAX is a function that takes the maximum value of the calculation results in each region, and 11 indicates that it takes the absolute value. Furthermore, a median calculation may be used instead of the MAX calculation. Also, since vectors are two-dimensional,
For the degree of divergence, the average value or the root mean square value of the values obtained for each component, or the sum f+fi is used as the actual degree of divergence.

(2) Divergence (DIV)=CMAX (+Motion vector in a certain area and motion vector in other areas 1) However, this calculation is performed only for the motion vectors in the area where the motion vector determination information is OK. CMAX is a function that performs calculations within 0 for the matching of all areas and takes the maximum value, and 11 indicates that it takes the absolute value. Furthermore, instead of taking the maximum value in the CMAX calculation, the median may be taken. Further, since the vector is two-dimensional, the average value, the root mean square value, or the sum of the values obtained for each component is used as the actual divergence.

(3) Divergence (DIV) = (Σ1 (average of motion vectors in each region ff1) - (motion vector in each region) l)/(motion vector determination information is O
(number of regions where K is the number of regions) However, Σ indicates that the sum of regions for which motion vector determination information is valid is calculated. Also, the vector is 2
Since it is dimensional, the average value, the root mean square value, or the sum of the values obtained for each component is used as the actual divergence.

The larger the value of the divergence DIV obtained by this calculation, the greater the degree to which each motion vector detected from each region points in a different direction. Indicates that it is large. Furthermore, if the degree of divergence is O, the motion vectors all point in the same direction, and the subject can be determined to be stationary.

Next, the output of the divergence calculation unit 13 is passed through a low-pass filter 14 to suppress high frequency components that change quickly over time. The output of the low-pass filter 14 is then converted into a predetermined value by a converter A15. The input of this converter A15 is the divergence (which has been subjected to LPF), and the output is a motion vector coefficient by which the motion vector is multiplied. Figure 2 shows the human output characteristics of the converter A15. In other words, the degree of divergence is a certain level (A in the figure).
) or less, the motion vector coefficient is 1, and if it exceeds this, the value gradually decreases and becomes 0 at point B, and if the degree of divergence is greater than that, the motion vector coefficient becomes 0. Point A・
The characteristics can be adjusted by changing the position of point B and the value of the motion vector coefficient at each point.

As shown in FIG. 3, this conversion characteristic may be realized as a table using a ROM or the like (see (a)), or may be realized using addition/subtraction and clipping (see (b)).

The control unit 16 is composed of a multiplier or a variable gain amplifier, and calculates the magnitude of the motion vector by multiplying the motion vector output from the motion vector determination unit 12 by the motion vector coefficient from the converter A15. Control.

By doing this, when the motion vectors in each region of the screen are different from each other, that is, when a part of the subject on the screen is moving, the larger the divergence, the more the motion vector of the entire screen is output. Reduce the size and suppress false detections. Suppression of false detections here means that the motion vectors to be output are brought closer to 0 as the directions of the motion vectors detected in each region become more disparate. Here, the motion vector output when the degree of divergence is large is a motion vector that is different from the motion vector due to the actual movement of the imaging device, but it is better to , corresponds to controlling the device in the direction of stopping motion vector detection.

As described above, according to this embodiment, when there is movement in a subject in a part of the screen and there is a possibility that an error may occur in motion vector detection, by controlling the size of the output motion vector, Erroneous detection can be suppressed.

FIG. 4 is a block diagram of a motion vector detection device according to a second embodiment of the present invention, in which 1 is a motion vector detection section;
2 is a motion vector determination unit, 13 is a divergence calculation unit, 14 is a low-pass filter (L P I; ), 15 is a converter A, 16
is a control section, and up to this point it is the same as in the embodiment described above. A motion vector integration section 17 is an example of motion vector integration means, and is a means for time-integrating a motion vector and outputting an integrated motion vector. The integral motion vector indicates the amount of translation of the screen from a certain reference time.

The operation of the motion vector detection device configured as above will be explained.

In FIG. 2, similarly to the first embodiment, the motion vector detection unit 11 detects motion vectors between fields or frames 1 of the input image signal for each screen area,
The motion vector determining unit 12 determines the validity of the detected motion vector in each region, and determines the motion vector for the entire screen from among the motion vectors detected in the regions determined to be valid.

Then, the motion vector integration unit 17 integrates the motion vector over time to obtain an integral motion vector. At this time, the integral calculation is given by the following formula.

5-vec t (n)==s vec L <nn
-1)X + vecL(n) However, svect(n): Current field (
(frame) integral motion vector vec t (n-1): Previous field (frame) integral motion vector vecL(n): Current field (frame) differential) motion vector m: ttn attenuation coefficient (≦1) be.

Here, if the value of m is 1, it is a perfect integration, and as the value of m becomes smaller, the amount of integral attenuation increases.

Next, the divergence calculation unit 13 calculates the divergence of the motion vector based on the output of the motion vector detection unit 11, that is, the motion vector of the area determined to be valid by the motion vector determination unit 12 among the motion vectors of each detection area. Calculate degrees. The calculation method is shown in the following formula.

(1) Divergence (DIV) = MAX (1 (average value of motion vectors in each region) - (motion vector in each region) 1) However, this calculation only applies to motion vectors in regions where the motion vector determination information is OK. Do about. MAX is a function that takes the maximum value of the calculation results in each area, and II shows that it takes the absolute value. Furthermore, a median calculation may be used instead of the MAX calculation. Further, since the vector is two-dimensional, the average value, the root mean square value, or the sum of the values obtained for each component is used as the actual divergence.

(2) Divergence (D[V): CMAX (+1 motion vector in a certain region and other motion vectors in other regions) However, this calculation is performed only for motion vectors in regions where the motion vector determination information is OK. CMAX is a number that takes the maximum value when calculations within 0 are performed on all combinations of areas, and indicates that CMAX takes the absolute value. Also, C
Instead of taking the maximum of 1 shift in the MAX operation, the median may be taken. Further, since the vector is two-dimensional, the average value, the root mean square value, or the sum of the values obtained for each component is used as the actual divergence.

(3) Divergence (DIV) = (Σ1 (average value of motion vectors in each region) - (motion vector in each region) N/(number of regions for which motion vector determination information is OK) However, Σ is motion vector determination Indicates that the summation is calculated for the area where the information is valid. Also, the vector is 2
Since it is dimensional, the average value, the root mean square value, or the sum of the values obtained for each component is used as the actual divergence.

Next, the output of the divergence calculation unit 13 is passed through a low-pass filter 14 to suppress components that change quickly over time. The output of the low-pass filter 14 is then converted into a predetermined value by a converter A15. The input of this converter A15 is the divergence (which has been subjected to LPF), and the output is a motion vector coefficient by which the motion vector is multiplied. The output characteristics of converter A150 are also similar to those of the previous embodiment, and are shown in FIG.

Then, the control unit 16 controls the magnitude of the integral motion vector by multiplying the integral motion vector output from the motion vector integrating unit 17 by a motion vector coefficient.

As described below, according to this embodiment, the characteristics of the low-pass filter 14 can be adjusted by directly controlling the integral motion vector obtained by multiplying the motion vector by time (n minutes), that is, the amount of motion of the actual imaging device. It is possible to adjust the temporal control characteristics by only using the method, and it is also possible to suppress the rounding error of the calculation for the motion vector before integration from being multiplied by (n), and to suppress the detection error of the integral motion vector when there is movement in the subject. .

FIG. 5 is a block diagram of a motion vector detection device according to a third embodiment of the present invention, in which 11 is a motion vector detection section;
2 is a motion vector determination section, 13 is a divergence calculation section, 14 is a low-pass filter (LPF), 18 is a converter B, and 17 is a motion vector integration section.The points that differ from the second embodiment are as follows. The point is that the value of the integral attenuation coefficient m in the integral calculation of the motion vector integrating section 17 is controlled by the output of the converter 81B.

The operation of the motion vector detection device configured as above will be explained.

In FIG. 5, similarly to the second embodiment, a motion vector detecting section 11 detects motion vectors between fields or frames of an input image signal for each region of the screen, and a motion vector determining section 12 detects motion vectors in each region. The validity of the vector is determined, and the motion vector for the entire screen is determined from among the motion vectors detected in the area determined to be valid.

Next, the divergence calculation unit 13 calculates the divergence of the motion vector based on the output of the motion vector detection unit 11, that is, the motion vector of the area determined to be valid by the motion vector determination unit 12 among the motion vectors of each detection area. Calculate degrees. The calculation method is shown in the following formula.

(1) Divergence (DIV) = MAX (+ (average value of motion vectors in each region) - (motion vector in each region) 1) However, this calculation only applies to motion vectors in regions where the motion vector determination information is OK. Do about. MAX is a function that takes the maximum value of the calculation results in each region, and 11 indicates that it takes the absolute value. Furthermore, a median calculation may be used instead of the MAX calculation. Further, since the vector is two-dimensional, the average value, the root mean square value, or the sum of the values obtained for each component is used as the actual divergence.

(2) Divergence (DIV)=CMAX (+Motion vector in a certain area and motion vector in other areas 1) However, this calculation is performed only for the motion vectors in the area where the motion vector determination information is OK. CMAX is a function that performs calculations within 0 for all combinations of areas and takes the maximum value, and indicates that CMAX takes the absolute value. Also, C
Instead of taking the maximum value in the MAX operation, the median may be taken. Further, since the vector is two-dimensional, the average value, the root mean square value, or the sum of the values obtained for each component is used as the actual divergence.

(3) Divergence (DIV) = (Σ1 (average value of motion vectors in each region) - (motion vector in each region) l)/(number of regions for which motion vector determination information is OK) However, Σ is the motion vector Indicates that the sum total for the area in which the judgment information is valid is calculated. Also, the vector is 2
Since it is a dimension, the average value, the root mean square value, or the sum W-1a of the values obtained for each component is used as the actual divergence.

Next, the output of the divergence calculation unit 13 is passed through a low-pass filter 14 to suppress components that change quickly over time. The output of the low-pass filter 14 is then converted into a predetermined value by a converter B18. The input of this converter B18 is the divergence (with LPF applied), and the output is the integral attenuation coefficient in the integration of the motion vector. Figure 6 shows the human output characteristics of converter B1B.

The characteristics of converter BIB are similar to those of converter A15,
As the divergence increases, the predetermined divergence (A
point), the value of the integral attenuation coefficient m is set to a characteristic that attenuates from a value close to 1.

The attenuation characteristics can be adjusted by changing the positions of point A and point B and the values mA and mB of the integral attenuation coefficient at each point.

Then, using this integral attenuation coefficient m, the motion vector is time-integrated by the motion vector integrator 17j to obtain an integral motion vector. At this time, the integral calculation is given by the following formula as in the first and second embodiments.

5-vec t (n):=s vc Ct
(n-1) (n): (difference) motion vector of the current field (frame) ■n: integral attenuation coefficient (≦1).

In this way, by controlling the attenuation coefficient in the integral calculation using the degree of divergence, the magnitude of the integral motion vector is controlled and the result is output.

As shown in L above, according to this embodiment, integral motion vector detection can be performed in a subject where a part of the screen moves by controlling the attenuation coefficient in motion vector integral calculation based on the divergence of the motion vector. Erroneous detection can be suppressed.

According to this configuration, the control unit 16 is not required, and the scale of processing can be reduced.

Further, FIG. 7 is a block diagram of an embodiment of the image blur correction device of the present invention, and the broken line portion in the figure is the second to third embodiment of the image blur correction device of the present invention.
This is a motion vector detection device 19 according to an embodiment of the present invention. Also, 2
0 is a reading position control unit, and 21 is an image enlarging device.

This embodiment is an image blur correction device that removes image blur components by controlling the image readout position based on the detected motion vector.

The image signal is input to an image enlarging device 21, and the image is enlarged as shown in FIG. A portion of the original image is enlarged and output, and at this time the value of each pixel of the enlarged image is interpolated by the value of each pixel of the original image. The image reading position is
In the standard state, the positions are set such that A=B and C=r in FIG. Then, using the motion vector output from the motion vector detection device 19, the image reading position is controlled so as to offset the translation of the image. The broken line portion indicates the readout portion when the readout position is controlled based on the motion vector. By doing this, it is possible to output a shaking image in a still state.

As shown in the figure below, according to this embodiment, the control is smoothly suppressed even when a part of the subject is not stationary, and a stable image with suppressed image shaking components is produced without noticeable malfunctions. can be obtained.

In addition, in the first to third embodiments, in the calculation in the divergence calculation unit 13, the average (direct) of the motion vectors of each area was used, but the median may also be used, or the motion vector of the entire screen may be calculated. May be used.

Further, in the first to third embodiments, the low-pass filter 14 may be omitted.

In addition, in the first and second embodiments, the characteristic of the converter A15 (Fig. 2) is a linear characteristic from point to point B, but it may be a curved line or several straight lines. It may also be a polygonal characteristic. Further, although the minimum value of the motion vector coefficient is set to O, it may be a value of 0 or more.

Also, in the third embodiment, the characteristics of the converter I31 El (Fig. 6) are linear characteristics from point to point B, but they may also be curved lines or polygonal lines made up of several straight lines. It can also be a characteristic. In addition, the characteristics after point B are expressed by a line segment (A, mA
) - (B, mB) (dashed line).

In addition, in the first to third embodiments, signal processing is digitalized, and when the output of the motion vector determination section (motion vector, motion vector determination information) is input to the microcomputer, the divergence calculation section 13 The operations of the bandpass filter 14, converter A15, converter B18, motion vector integrator 17, and controller 16 can be easily realized by software calculations by a microcomputer.

Further, in the second embodiment (FIG. 4), the control section 16 is placed on the output side of the motion vector integration section 17, but it may be placed on the input end of the motion vector integration section 17.

Furthermore, in the second embodiment, the converter A15 and the converter B1B may be connected in parallel to the output side of the low-pass filter 14, and each may control the control section 16 and the motion vector integration section 17.

DETAILED DESCRIPTION OF THE INVENTION According to the present invention, as described in detail, it is possible to provide a motion vector detection device that suppresses motion vector detection errors for the entire screen even when a part of the subject is not stationary.

Furthermore, image shaking can be effectively prevented.

[Brief explanation of drawings]

FIG. 1 is a block diagram of a motion vector detection device according to a first embodiment of the present invention, FIG. 2 is a diagram showing the relationship between motion vector coefficients and divergence of the same embodiment, and FIG. 3 is a converter A of the same embodiment. 4 is a block diagram of a motion vector detection device according to a second embodiment of the present invention, FIG. 5 is a block diagram of a motion vector detection device according to a third embodiment of the present invention, and FIG. 6 is a block diagram of a motion vector detection device according to a third embodiment of the present invention. A diagram of the relationship between the integral attenuation coefficient and the divergence of the same embodiment, FIG. 7 is a configuration diagram of the fourth embodiment of the present invention, FIG. 8 is an explanatory diagram of image enlargement processing, and FIG. 9 is a motion vector of the image. Explanatory diagram, 1st
Figure O is a configuration diagram of a conventional motion vector detection device, Figure 11 is an explanatory diagram of the state of a representative point and surrounding pixels in the representative point matching method, and Figure 12 is a diagram showing the state of a representative point when the subject is stationary and when it is moving. FIG. 4 is a diagram showing motion vectors detected in each area on the screen when l...Latch A, 2...Representative point storage memory, 3...
・Latch B, 4...Correlator, 5...Address controller, 6...Address switching circuit, 7...Accumulative adder, 8...Correlation search circuit, 9...Correlation valid Invalidity determination circuit, lO...determination circuit, 11...motion vector detection unit, 12...motion vector determination unit, 13.
...Low pass filter, 14...Low pass filter, 15...Converter A, l 6...Control section, 7...Motion vector integrator, 8...Converter B0

Claims (4)

[Claims] (1) A motion vector detection means that divides an input image signal into multiple regions and calculates the motion vector of the image in each region, and determines the effectiveness of the motion vector in each region and calculates the motion vector of the entire screen. a motion vector determination means for determining a motion vector; a divergence calculation means for obtaining a degree of divergence indicating the degree of dispersion of motion vectors in each detection area using the output of the motion vector detection means; 1. A motion vector detection device comprising: a conversion means for converting into a value related to the size of the motion vector; and a control means for controlling the size of the motion vector of the entire screen based on the output of the conversion means. (2) The motion vector detection device according to claim 1, wherein the control means controls the magnitude of the output of the integration means for time-integrating the motion vector of the entire screen. (3) The motion vector detecting device according to claim 2, wherein the attenuation characteristic of the motion vector in the integration process of the integrating means is controlled by the output of the converting means. (4) An image enlargement device that enlarges an input image, a motion vector detection means that calculates a motion vector of the image, and a motion vector determination means that judges the validity of the motion vector in each area and determines the motion vector of the entire screen. an integrating means for time-integrating the motion vector of the entire screen; and a divergence calculating means for obtaining a divergence indicating the degree of dispersion of the motion vectors of each detection area using the output of the motion vector detecting means.
A conversion means for converting the output of the divergence calculation means into a value related to the magnitude of the output, a control means for controlling the output of the integration means or the motion vector determination means by the output of the conversion means, and the control means Alternatively, it has a readout position control means for calculating the image readout position of the image enlargement device using the output of the integration means, and by controlling the readout position of the enlarged image, the image is translated in parallel and the input image is An image shake correction device characterized by suppressing shake components.