JPH0646316A - Shaking controller for video camera - Google Patents

Abstract

PURPOSE:To suppress the influence by a passing object by verifying the intrusion of the passing object in a block based on the increase degree, the passing degree, and the change degree of a moving vector at every block dividing image fields and performing the shaking correction only for a no-intrusion block. CONSTITUTION:A video signal at every image block via a camera circuit 3 and an A/D converter 4 is supplied to a frame memory 5 and a shaking detection circuit part 8. This circuit part 8 calculates a correlation value by a representative matching method, detects the shaking based on the correlation value and corrects the contents of a memory 5. At this time, a state that a passing object does not intrude in a block whose correlation value is reliable from the values of the increase degree, the passing degree and the changing degree of a moving vector at every block is verified and whether the block in which the passing object is intruded and the block in which it is not intruded are surely discriminated. Then, the shaking correction is performed for the only block in which the passing object is not intruded, a correction is not performed for an intruded block in which an error becomes large and the influence by the passing object is suppressed.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a camera shake control device for a video camera, which calculates a motion vector of a video signal in a video field due to camera shake and executes camera shake correction processing based on the magnitude of the motion vector. .

[0002]

2. Description of the Related Art A video camera having a camera shake correction function is conventionally known, for example, National Techni
cal Report Vol. 37 Jun. 199
1 P48-54 "Pure electronic image shake correction technology".

This technique detects a motion vector for extracting the motion of the image from the image signal for each of the four blocks in the video field by the representative point matching method, and specifies the motion vector for each block. At this time, the magnitude of the absolute value of the difference between the average of the motion vectors and each motion vector is obtained as the divergence, and the average motion vector obtained according to this divergence is multiplied by a coefficient, and this is used as the motion detection motion. The video extraction field of the video field that follows as a vector is specified, and camera shake correction is executed.

[0004]

However, in the case of the conventional method used to determine the divergence and determine whether or not to perform the camera shake correction, a portion having a small divergence, that is, each block in the video field is passed. If there is no object, accurate camera shake correction can be expected, but if a passing object enters a specific block, the divergence becomes large to some extent and there is an error in the average motion vector due to the motion vector of the block where the passing object enters. However, there was a problem in that camera shake correction could not be performed accurately.

Therefore, the present invention devises a more reliable method of detecting a passing object entering, in place of the divergence of the prior art, and applies it to the remaining blocks of each block in the video field where the passing object entered. It is an object to remove the influence as much as possible and perform an accurate camera shake correction.

[0006]

According to a first aspect of the present invention, a video field is divided into a plurality of blocks, and a motion vector for each block is calculated by a correlation value calculation method. Means for performing necessary camera shake correction processing, means for calculating an increase degree serving as an index of a period in which the directions of the calculated motion vectors in the corresponding blocks of each video field are the same, and And a means for determining whether to execute the camera shake correction processing of each block.

A second invention of the present invention is to divide a video field into a plurality of blocks and calculate a motion vector for each block by calculating a correlation value, and a camera shake required according to the calculated motion vector. Means for performing correction processing,
Means for calculating the degree of passage, which is an index of a period in which the calculated correlation value in the corresponding block of each video field is a reliable value, and execution of the camera shake correction process for each block is determined by the value of the degree of passage. And means.

Further, a third invention of the present invention is to divide a video field into a plurality of blocks and calculate a motion vector for each block by calculating a correlation value, and a means required according to the calculated motion vector. Means for performing blur correction processing, means for calculating a degree of change representing the absolute value of the difference in the magnitudes of the calculated motion vectors in the corresponding blocks of each video field, and means for calculating the degree of change of each block according to the value of the degree of change. And means for determining execution of the blur correction process.

A fourth invention of the present invention is an integration of the above-mentioned three inventions, a means for dividing an image field into a plurality of blocks, and calculating a motion vector for each block by a correlation value calculation; Means for performing a necessary camera shake correction process according to the motion vector thus calculated, means for detecting an increase degree serving as an index of a period in which the directions of the calculated motion vectors in the corresponding blocks of each video field are the same, Means for calculating the degree of passage, which is an index of the period during which the calculated correlation value in the corresponding block of the video field is a reliable value, and the absolute value of the difference in the magnitude of the calculated motion vector in the corresponding block of each video field. Means for calculating the degree of change that is normalized, and each calculated value of the increase degree calculating means, the passage degree calculating means, and the change degree calculating means as parameters. Means for determining the execution of the shake correction based on and in accordance with the rules calculate the reliability of camera shake of the corresponding block to become more.

[0010]

According to the first to third inventions, each block in the video field into which a passing object has entered can be specified, and this block can be excluded from the video field to be subjected to camera shake control. It is possible to reduce the error received by the entry of the.

According to the fourth aspect of the invention, the reliability of the block in which the passing object has entered is increased by each of the methods of the first to third aspects of the invention, and more accurate camera shake correction processing can be realized. Becomes

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of a camera shake control device of the present invention will be described in detail below with reference to the drawings.

First, the basic structure of a video camera to which the present invention is applied will be briefly described with reference to FIG. In the figure, 1 is a lens, 2 is an image sensor, 3 is a camera circuit section,
4 is an ADC (Analog-Digital-Conv)
erter), 5 is a frame memory, 6 is an electronic zoom circuit section, and 7 is a DAC (Digital-Analog-).
Converter), 8 is a camera shake detection circuit section, 9 is a camera shake control circuit section, and 10 is a memory control circuit section.

In such a structure, the light from the subject passes through the lens 1, is received by the solid-state image sensor (here, CCD) 2, is converted into an electric signal, and is output as a video signal through the camera circuit section 3. This signal is sent to the ADC 4 Is converted into an AD signal and stored in the frame memory 5, and is further compared by the camera shake detection circuit unit 8 with the next video signal (the video captured through the lens 1 after 1/60 seconds) and between the video fields. The movement of the image is detected, and the amount of camera shake is calculated based on this by using the representative point matching method.

In the camera shake control circuit section 9, the position of the image in the video field is moved based on the amount of camera shake obtained as described above. However, since it cannot be moved as it is, correction is made using the principle of electronic zoom. To do.

FIG. 2 is a diagram for explaining the principle of such electronic zoom, in which a frame 12 for the zoom ratio is determined for the size of the image in one video field 11. This frame 12
The position of is allowed to move freely within the original video field 11. Then, the frame 12 is enlarged by an interpolation method so as to have a size corresponding to the original video field 11.

In this way, the image at an arbitrary place in the video field 11 is electrically zoomed by the electronic zoom circuit section 6 to leave a margin space (correction range 13) by the difference between the original video field 11 and the clipping frame 12. As shown in FIG. 3, when a camera shake of the video camera 14 occurs, an image of a target person is located in the lower left part of the video field 11 and an image of the target person is located in the upper right part of the next video field 15, as shown in FIG. The image blurs as if there were an image.

Therefore, by moving the clipping frame 12 as in both video fields 11 and 15 in accordance with the amount of movement of the detected image, both fields 11 and 14 have an image of the target person in the clipping frame 12. It becomes a state where it fits neatly, and the camera shake correction is realized by DA-converting this and returning it to the original state.

Actually, one video field 11 is shown in FIG.
As shown in FIG. 4, it is divided into four blocks 16 to 18 and a comparison between two video fields is performed for each block.
The camera shake amount is calculated by the representative point matching method.

FIG. 5 is a functional block diagram for explaining the motion vector detection method by representative point matching used in this embodiment, which corresponds to the contents of the camera shake detection circuit unit 8 in FIG. 1 and is a logic circuit of an LSI. It has been assembled.

In the figure, reference numeral 20 designates S / S for the video signal of the camera from the ADC 4 without lowering the detection accuracy.
A filter circuit unit 21 is provided to improve the N ratio and obtain a sufficient detection accuracy with a small number of representative points. Reference numeral 21 denotes each block 16 from the video signal that has passed through the filter circuit unit 20.
Representative point memory for extracting a plurality of representative points in a plurality of representative point areas within 19 and storing their positions and luminance signals-2
Reference numeral 2 denotes a correlation value calculator for calculating the correlation value between the luminance signal of the representative point memory 21 and other pixels in each area, and 23 cumulatively adds the correlation values of pixels at the same position in each area. The cumulative addition unit 24, which compares the correlation value of each pixel with the value of the cumulative addition unit 24 to obtain the minimum value thereof, and also obtains the average of the correlation values within one block, and further has the minimum value. It is a minimum value, an average value, and a minimum position calculation unit for obtaining the position of a pixel. Then, the cumulative addition unit 23
The minimum value, the average value, and the value obtained by the minimum position calculation unit 24 are passed to the next camera shake control circuit unit 9 as outputs.

FIG. 6 shows a functional algorithm corresponding to the contents of the camera shake control circuit section 9. The algorithm is executed by software processing of a microprocessor.

In the figure, 25 is an input from the camera shake detection circuit unit 8, and 26 is each block detected depending on whether a value obtained by dividing the minimum value of the inputs by an average value is within a certain threshold value. Correlation value judgment to judge whether or not the motion vector, that is, the detected motion vector itself is reliable, and 27 uses the correlation values of the four pixels around the pixel having the smallest correlation value in order to improve the detection accuracy of the motion vector. The minimum position interpolation for performing interpolation to interpolate the minimum position of the correlation value, 28 detects blocks 16 to 19 that are considered to be motions other than camera shake such as passing objects and moving objects, and detects this block from the camera shake correction routine. The fuzzy judgment to deviate, 29 is a vector operation for calculating the motion vector of the blocks 16 to 19 using the reliable correlation values of the blocks 16 to 3, 3
0 is an integral correction calculation for integrating the motion vector of each video field 11, 15 ... to calculate a motion vector of each video field 11, 15 ..., 31 is a motion calculated by the integral correction calculation 30 Coring correction for removing the effect of noise from the vector, 32 is pan / tilt processing for removing the effect added to the motion vector during panning or tilt operation of the video camera, 33 is effective by the vector operation 29 to pan / tilt processing 32. Block 16
It is the output of the result of obtaining the motion vector and the integral vector of the current field from the motion vectors of # 19 to 19 and the motion vectors of the immediately previous video field and the two previous video fields.

In the above algorithm, it is important to determine whether the camera shake or the passing object and the moving object enter the blocks 16 to 19 in order to correct the camera shake accurately. In the present embodiment, this judgment is made by the fuzzy judgment 28. The contents of the fuzzy judgment 28 will be described below.

FIG. 7 is a functional flow chart showing the algorithm of the fuzzy judgment 28. In this embodiment, the invalid block judgment is carried out by a triple check incorporating all three parameters of the degree of increase, the degree of passage and the degree of change which will be described later. However, as will be described later, it goes without saying that the invalid block can be sufficiently determined only by checking the individual parameters.

In FIG. 7, at 34, the correlation value whose reliability is judged by the above-mentioned 26 correlation value judgment and the motion vector of the current block to be judged which is calculated by the interpolated minimum position are inputted and a hold flag, A holding number register, a pre-previous and pre-previous video field direction flag, and an increment register are prepared respectively, and each flag and register are initialized.

Reference numeral 35 is a calculation routine of the degree of increase (integer value) that serves as an index representing the duration of time that the motion vector of the current block faces the same direction, and the detailed contents will be described in FIG. Here, in the case of normal camera shake, the degree of increase does not become so large that the value of the motion vector is positive or negative with respect to a fixed direction, but the penetration of a passing object If there is, there is a feature that the degree of increase increases because the motion vector in either direction continues.

In this increase degree calculation routine 35, first, it is judged whether the motion vector of the current block is zero or not.
5 ends at 38 (the increment value is not incremented either, so the increment value of the previous video field remains unchanged-zero in the case of the first video field). This means that it is not necessary to determine whether the current block is valid or invalid because there is no motion vector, no camera shake, and no passing object.

Here, if the motion vector has a positive or negative value instead of zero, it is judged at 39 whether or not a holding operation described later has been carried out last time. If the direction is changed (positive value to negative value, or negative value to positive value), the increment value of the previous field is 3 or more. It is judged at 41 whether it is small or not.

On the other hand, when there is no change between the motion vector of the previous video field and the motion vector of the current video field at 40, it is determined at 42 whether the degree of increase calculated in the previous video field is saturated (ie, The degree of increase is 0 to 31
In order to set to take a value between 3), it is judged whether it has reached 31), and if it is not saturated yet, the increment degree is incremented by 1 at 43, and if it is saturated, it cannot be further incremented. At 44, the value of the previous video field is left unchanged for the degree of increase as no change.

Further, when the increase degree up to the previous video field is smaller than 3 in 41, it is appropriate to consider that the passing object has passed through the current block, and in 42, the increase degree is set to 1 (minimum increase degree). ), And when calculating the degree of increase thereafter, the algorithm starting from 36 is re-executed on the basis of the direction in which the motion vector has changed.

On the other hand, when the increase degree up to the previous video field is 3 or more at 41, it is judged that a passing object has entered the current video field, but it is still early to determine,
At 46, the number of holdings until the motion vector of the previous video field is calculated is compared with a threshold value (10 in this case), and if the number is smaller than this threshold value, the increase degree is not incremented at 47 and the current increase degree is held. In this case, if the direction originally changes, the increment is basically incremented from 1 again for the first time, but if the increment in all video fields is large, even if the direction of the motion vector is reversed once. This is because it is considered that the error will be smaller if it is held without judging that the direction of the motion vector has changed immediately because of the influence of noise and the like. However, when the judgment of holding is repeated, it is set not to be influenced by noise or the like, so that the routine can be exited at an appropriate place (10 times of holding).

Further, when the process of increasing degree is suspended in the previous video field in 39, it is checked whether the motion vector of the current video field is the same as the direction of the motion vector of the previous video field. If it is the same, the saturation of the increase degree of 42 is judged, and if it is different, it is determined that the direction of the motion vector has changed from the previous video field to the true value, and the increase value is determined. Set to 2 at 49.

In this way, the degree of increase of the motion vector of each video field taken in every 1/60 seconds is calculated for each block to integrate the motion vectors in the same direction between the video fields. It means that the time is calculated, and the passing object is passing through the block at least while the increment is being incremented, and it can be understood that this block should be deleted from the camera shake correction.

The increase degree calculation routine 35 can determine whether the block is valid or invalid by calculating the increase degree. In order to make this determination deterministic, the following passage degree calculation is performed.

At this time, first, at 50, the correlation value judgment flag and the passage degree register are initialized. FIG. 9 is a time chart showing the correlation value and the degree of passage in time series, and the concept of the degree of passage will be explained along with this. When the passing object passes through the block (shown by 51), the correlation value is Suddenly higher
After this, the correlation value is kept low until the passing object leaves the block, and the correlation value suddenly rises again at the moment of exit (indicated by 52).

The idea of passing degree is to try to judge a passing object by utilizing such a rapid change of the correlation value.
The correlation value rises sharply at the rising edge 51, then increases at a constant rate until the next sharp rising edge 52 appears, then becomes constant, and then falls sharply again at the subsequent rising edge 52. By using this feature, the pass-through register is cleared when the pass-through is sharply lowered, and then the pass-through is incremented by 1 for each video field while the correlation value is low. Is calculated, and the determination is made as to whether or not the block is a block for camera shake correction.

FIG. 10 shows an algorithm of this method. In the pass degree calculation routine 53, the value obtained in the correlation value calculation 22 of FIG. 5 is first fetched, and the correlation is determined depending on whether or not this value is smaller than a predetermined value. At 54, it is judged whether the value is reliable, and if it is reliable, that is, 55,
The pass degree register is incremented by 1 and the value of this register is multiplied by an appropriate coefficient to normalize at 57. Correlation value data is fetched every 1/60 seconds, and 8 bits or more are counted in the register.

By the way, in the operation of the membership function which will be described later, since 5-bit data is handled, in the normalization of 57, which 5-bit portion of 8-bit data is to be taken as data is selected.

Finally, if there is no passing object forever, the passing degree register continues to be incremented, but since it does not make sense to have an excessively large number, it is limited at an appropriate place (maximum value 31 in the embodiment), and the routine is 58. Set it to finish.

If the correlation value suddenly increases due to the intrusion of the passing object into the block at 54, the passing degree register is cleared at 59.

When the correlation value is unreliable in this way, it can be considered that the passing object is intruding. Even if the correlation value becomes reliable thereafter, it can be considered that the object is passing for a while. Further, in the case of a passing object, as shown in FIG. 9, the correlation value becomes unreliable next time and becomes reliable again thereafter. A small period continues, but it does not have a particularly bad influence on the detection of invalid blocks. Moreover, after a considerable amount of time has passed since the correlation value became reliable, the passing level continues to take a large value.

Like the routine for calculating the passage degree, 1
Determining whether or not the correlation value in the video field captured every 60 seconds is reliable means calculating the elapsed time after determining that the correlation value is reliable, With this elapsed time, it is possible to accurately check whether or not there is a passing object in the current block. As a result, untrustworthy blocks are removed and camera shake control is performed to eliminate unnaturalness due to camera shake. be able to.

If it is desired to extract invalid blocks more carefully, the following variation degree calculation routine is performed.

At this time, first, at 60 in FIG. 7, the motion vector of each block of the current video field and the motion vector of each block of the previous video field are input.

As shown in FIG. 11, the routine 61 for the degree of change subtracts the motion vector of the previous field from the motion vector of the current field in 62, and takes the absolute value of the difference in 63. In the normal camera shake operation, no great change is seen in the absolute value between the video fields. That is, if there is a large change, it is regarded as an intrusion of the passing object into the block. Therefore, the calculation of the absolute value is, in other words, the detection of the passing object.

Since the value obtained at 63 may have a large analog value, a coefficient for normalization is multiplied at 64 to give a value from 0 to 31 as in the case of the passing degree. Further, the maximum value is limited to 65 in order to further increase the maximum value to 31.

As described above, the increment calculation routine 35,
Passage calculation routine 53, change degree calculation routine 61
At 0, the value of 0 to 31 determined by the penetration of the passing object into the block is output as a parameter, and at 66, membership calculation is performed according to the fuzzy rule described below.

Table 1 is a table defining fuzzy rules for performing the membership operation, and FIG. 12 is a diagram defining membership functions.

[0050]

[Table 1]

Table 1 is defined according to the following rules.

1. If the degree of change is small and the degree of increase is small, the block is considerably trusted even if the passing degree is very small.

2. If the degree of change is small, the degree of increase is small, and the degree of passage is not too small, the block is very reliable.

3. If the degree of increase is large and the degree of passage is not too large, the block is not very reliable even if the degree of change is small.

4. Even if the degree of increase is large, if the degree of change is small and the degree of passage is very large, the block is quite reliable.

5. Even if the degree of increase is small, if the degree of change is large and the degree of passage is not too large, the block is not trusted so much.

6. Even if the degree of change is large, if the degree of increase is small and the degree of passage is very large, the block is fairly reliable.

7. If the degree of change is large, the degree of increase is large, and the degree of passage is not too large, the block is not reliable at all.

8. Even if the degree of change is large and the degree of increase is large, if the passing degree is very large, the block is trusted to some extent.

According to this rule, it is possible to cover the erroneous determination that the object is not passing when the passing degree becomes large to some extent after the invasion of the passing object 51 described with reference to FIG. 9 by the value of the increasing degree. Has become.

The values of the degree of increase, the degree of passage, and the degree of change are converted from membership functions into corresponding membership values, and these values are substituted into Table 1 of the fuzzy rule.
In 7, the maximum value of these membership values is set as the representative membership value in the rule, and the simple sum Z in all the rules of the representative membership value and all rules of the value obtained by multiplying the representative membership value by the reliability. And the integrated sum W at.

Next, at 68, the threshold of the value T obtained by dividing the simple sum Z by the integrated sum W is less than 1 (for example, 0.
8), and it is determined whether Z is smaller than TW.

Here, based on the principle that the smaller the value of T, the more reliable the motion vector of the block,
If it is smaller, it is judged in 69 whether or not the verification is completed for all the blocks. If the determination of validity or invalidity has not been verified for all blocks, the block counter is incremented by one at 70 and the process proceeds to verification of the next block.

When Z becomes larger than TW in 68, the block is regarded as an invalid block in 71 and is excluded from the camera shake correction.

When all the blocks have been verified in 69, the fuzzy routine 28 is terminated,
The image stabilization routine after the next vector operation 29 is executed.

Now, as described in the first section, the degree of increase,
It is possible to verify the invalid block even if each of the routines 35, 53 and 61 of the passage degree and the variation degree is used independently.

FIGS. 13 and 14 show that the invalid blocks can be verified only by the degree of increase by slightly improving the algorithm of FIG. That is, it is determined whether or not the degree of increase obtained at 66 is 15 or more instead of the end 38 of FIG. 8. If it is 15 or more, it is determined that a passing object has entered this block, and the block is set to 67. To disable it.

On the contrary, if it is less than 15, it is judged that the motion vector of the block is reliable, and it is confirmed in 68 whether or not all blocks have been verified.
Then, the process returns to 36 to verify the motion vector of the next block, and if completed, ends at 70.

As a result, the invalid block can be verified only by the degree of increase, and the work from the next vector operation 29 is continuously executed.

In the case of the passing degree as well, as in the case of the increasing degree as shown in FIG. 15, instead of the END 56 of FIG. 10, it is judged at 71 whether the passing degree is 15 or more. After it is judged as reliable that the verification of all the blocks is completed at 72, either one of the steps of 73 or 74 is selected to verify the next block.

If the passing degree is less than 15 at 71, it is considered that a passing object has entered the block, and the block is invalidated and the camera shake correction is not performed.

Further, in the case of the degree of change, as shown in FIG. 16, the maximum value is limited to 31 at 65 in FIG. 11, and then it is determined at 75 whether or not the obtained degree of change is 15 or more. If it is less than 15, it is confirmed at 76 whether or not the verification of all blocks is completed. If all are completed, the routine is exited at 77, and if not yet, at 78, the next block is verified.

If the degree of change is 15 or more in 75, the block is determined to be invalid, and the block is excluded from the subject of camera shake correction.

In the above description, the method of calculating the degree of increase and the degree of passage is such that the minimum is 0 and the increment is performed, but the maximum is set to 31, for example, and conversely the decrement method is also applicable. Needless to say.

[0075]

Since the present invention is configured as described above, in the image stabilization method using the representative point matching method,
In each block of the video field, the block to be corrected and the block not to be corrected due to the intrusion of the passing object can be accurately identified, and the effect of the passing object on the image obtained by the image stabilization can be improved. Expected to be possible effect.

[Brief description of drawings]

FIG. 1 is a block diagram showing a basic configuration of a video camera.

FIG. 2 is a diagram showing a principle of electronic zoom.

FIG. 3 is a diagram showing the principle of camera shake correction.

FIG. 4 is a diagram showing block division for performing representative point matching within one video field.

FIG. 5 is a functional block diagram illustrating a method of calculating a correlation value for representative point matching.

FIG. 6 is a functional block diagram illustrating a camera shake correction method.

FIG. 7 is a flowchart showing an algorithm of the fuzzy arithmetic routine of FIG.

FIG. 8 is a flowchart showing an algorithm of the increment calculation routine of FIG.

FIG. 9 is a time-series diagram showing the relationship between the degree of passage and the correlation value.

10 is a flow chart showing an algorithm of the passage calculation routine of FIG.

11 is a flow chart showing an algorithm of a change degree calculation routine of FIG. 7. FIG.

FIG. 12 is a diagram showing the relationship between the membership function and the degree of increase, the degree of passage, and the degree of change.

FIG. 13 is a first half of a flowchart showing an algorithm of an increment calculation routine different from that of FIG. 8;

FIG. 14 is the second half of the increment calculation algorithm following FIG.

FIG. 15 is a flowchart showing an algorithm of a passage degree calculation routine different from that of FIG. 10;

16 is a flowchart showing an algorithm of a variation degree calculation routine different from that of FIG.

[Explanation of symbols]

 8 Camera shake detection circuit 9 Camera shake control circuit 11, 15 Video field 16 to 19 block 28 Fuzzy routine 35 Increase degree calculation routine 53 Passage degree calculation routine 61 Change degree calculation routine

Claims (4)

[Claims] 1. A means for dividing a video field into a plurality of blocks and calculating a motion vector for each block by a correlation value calculation means, and a means for performing necessary camera shake correction processing according to the calculated motion vector, Means for calculating an increase degree serving as an index of a period in which the direction of the calculated motion vector in the corresponding block of each video field is the same, and execution of the camera shake correction process for each block is determined based on the value of the increase degree. A camera shake control device for video cameras. 2. A means for dividing a video field into a plurality of blocks and calculating a motion vector for each block by a correlation value calculation means, and a means for performing necessary camera shake correction processing according to the calculated motion vector, Means for calculating the degree of passage, which is an index of a period in which the calculated correlation value in the corresponding block of each video field is a reliable value, and execution of the camera shake correction process for each block is determined by the value of the degree of passage. A camera shake control device for video cameras. 3. A means for dividing a video field into a plurality of blocks and calculating a motion vector for each block by a correlation value calculation means, and a means for performing necessary camera shake correction processing according to the calculated motion vector, Means for calculating the degree of change representing the absolute value of the difference between the magnitudes of the calculated motion vectors in the corresponding blocks of each video field, and the execution of the camera shake correction process for each block based on the value of the degree of change A camera shake control device for video cameras. 4. A means for dividing a video field into a plurality of blocks and calculating a motion vector for each block by a correlation value calculation means, and a means for performing necessary camera shake correction processing according to the calculated motion vector, A means for detecting an increase degree serving as an index of a period in which the direction of the calculated motion vector in the corresponding block of each video field is the same, and a period in which the calculated correlation value in the corresponding block of each video field is a reliable value. Means for calculating the degree of passage serving as an index, means for calculating the degree of change by normalizing the absolute value of the difference in the magnitudes of the calculated motion vectors in the corresponding blocks of each video field, and the increase degree calculating means. Using the calculated values of the passing degree calculating means and the change degree calculating means as parameters, the reliability of the corresponding block against camera shake is calculated according to a predetermined rule. A camera shake control device for a video camera, which includes means for determining whether to perform camera shake correction processing based on the above.