JPH05204012A - Jiggle signal prediction device - Google Patents

Abstract

PURPOSE:To calculate a predictive jiggle signal by catching jiggle as a constant to predict the jiggle after a prescribed time in the midst of exposure by simple constitution and simple calculation. CONSTITUTION:Jiggling vibration is detected by a jiggle signal detection part 11 and at least one or more outputs among the outputs from the detection part 11 are stored in a jiggle signal storage part 12. Then, at least two or more kinds of predictive coefficients for predicting the jiggle signal after the prescribed time are stored in a jiggle predictive coefficient storage part 13. The output of the detection part 11 and/or the output of the storage part 12 and the output of the storage part 13 are multiplied and added by an arithmetic part 13 and the predictive jiggle signal is calculated.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a camera shake signal predicting device, and more particularly to a vibration which deteriorates image quality due to relative movement between a photographer and a subject in an optical device for storing an image of the subject such as a camera or a video camera. The present invention relates to a camera shake signal prediction device that prevents and corrects the effects of so-called camera shake vibration.

[0002]

2. Description of the Related Art Conventionally, there has been proposed a device for preventing deterioration of photographed image quality due to camera shake.

These devices use a mechanical vibration sensor (such as an angular velocity sensor or an acceleration sensor) to detect camera shake, or a time series change in the output of an image signal sensor such as a CCD or the brightness of an object. The change of is detected.
Further, as a specific method for preventing deterioration of image quality due to camera shake, correction driving of the optical system is performed so that there is no camera shake, imaging is prohibited when the camera shake is large, and the time for capturing one image is shortened. I am trying to do it. Further, a method has also been proposed in which image data is temporarily stored in a storage device, an address for reading an image is changed according to a blur signal, and image data without blur is output.

[0004] These methods for preventing camera shake require prediction of the camera shake signal when the effect is to be large. For example, in the case of detecting a blur with an image sensor such as a CCD, the integration time of the photocurrent on the pixel for capturing the image signal or the time for detecting the movement of the subject image from the time series data There is a time delay in camera shake signal detection. Further, even in the case of a mechanical vibration sensor or the like, when noise due to hum or the like is superimposed on the sensor signal, a filter for removing the harmful component is required. Therefore, the effective camera shake signal may be delayed due to this. Prediction is required to correct this delay.

In addition, when the camera shake is detected and the exposure is prohibited when the value is large and the exposure is permitted when the camera shake is small, it is necessary to know the size of the camera shake during the exposure before the exposure. There is. Therefore, it is necessary to predict the amount of camera shake that will occur during exposure before the start of exposure.

Further, in the case of a camera shake prevention device devised so as to eliminate the image blur on the image pickup system by actively moving the optical system or the like based on the camera shake signal, it is necessary to drive the optical system. A delay occurs due to the response of the actuator. It is more effective to prevent camera shake if this delay is also predicted.

Conventionally, as a method for predicting the camera shake signal, there has been considered a method of approximating the camera shake vibration to a simple vibration and detecting its cycle or frequency to predict the camera shake, or predict the timing when the camera shake will disappear. Has been. Also, a method has been proposed in which time-series data of camera shake vibration is approximated to a straight line or a higher-order regression line of quadratic or higher using a least square method or the like and extrapolated. Furthermore, ideas for applying neural networks and fuzzy inference engines to prediction have been submitted.

[0008]

By the way, in the case of approximation by the above-mentioned simple vibration, the actual vibration of the camera shake is not a clean simple vibration, and the vibration of 10 Hz or less in the dimension of the position of the image is complicatedly superimposed. is there. For this reason, the prediction based on simple vibration was inaccurate.

Further, in the case of regression line approximation, since a statistical method is used, a certain degree of accuracy is expected, but the calculation is complicated and the calculation amount becomes large, so that a camera or a video camera is required. When it is applied to such a small popular item, it is necessary to mount a computer with higher performance than necessary. Therefore, the cost becomes impractical. Furthermore, at the present time, an effective neural network or fuzzy inference chip that can be mounted on a camera or the like and is inexpensive has not been realized.

The present invention has been made in view of the above problems, and is a simple image stabilization device for correcting and preventing the vibration effect of camera shake in an optical device for recording an image of a subject such as a camera or a video camera. It is an object of the present invention to provide a camera shake signal prediction device capable of performing highly accurate camera shake signal prediction calculation with a simple configuration and simple calculation.

[0011]

That is, according to the present invention, as shown in the conceptual diagram of FIG. 1, in a camera shake prevention device in an optical device for recording an image of a subject, a camera shake signal for detecting camera shake vibration. A detection unit 11, a camera shake signal storage unit 12 that stores at least one output from the camera shake signal detection unit 11, and at least two types of prediction coefficients for predicting a camera shake signal after a predetermined time are stored. Camera shake prediction coefficient storage unit 13 and the camera shake signal detection unit 1
1 and / or the output of the camera-shake signal storage unit 12 and the output of the camera-shake prediction coefficient storage unit 13 are multiplied and added to calculate a predicted camera-shake signal.

[0012]

In the camera shake signal predicting apparatus of the present invention, the camera shake signal detection unit 11 detects camera shake vibration, and the camera shake signal storage unit 12 stores at least one output from the camera shake signal detection unit 11. It The camera shake prediction coefficient storage unit 13 stores at least two or more kinds of prediction coefficients for predicting a camera shake signal after a predetermined time. Then, the output of the camera shake signal detection unit 11 and / or the output of the camera shake signal storage unit 12 and the output of the camera shake prediction coefficient storage unit 13 are multiplied and added in the calculation unit 14 to obtain the predicted camera shake. The signal is calculated. The latest data of the camera shake signal input to the calculation unit 14 may be directly input from the camera shake signal detection unit 11 to the calculation unit 14 or may be input via the camera shake signal storage unit 12. Good.

[0013]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A camera shake signal predicting apparatus of the present invention will be described below with reference to the drawings. First, the principle of camera shake prediction in the present invention will be described.

Now, it is assumed that the camera shake signal is X (i) and the predicted camera shake signal resulting from the prediction is (Y). Here, the value of (i) takes at least values of 0 and 1, and is used up to a natural number (N) of 2 or more depending on the required accuracy of the data to be predicted. The shake signal at the present time is X (0), the shake signal at the past one time point is X (1), and the signal at the past two time points is X (1).
(2), X (i) if there is a camera shake signal at time i past
Express.

This X (0) signal is output from the above-mentioned camera shake signal detecting section 11. The past camera shake signal X (i) is sequentially stored in the camera shake signal storage unit 12, is discarded from unnecessary data that is old, and is updated with a new camera shake signal. Furthermore, the camera shake prediction coefficient storage unit 1
3 is the above-mentioned X as a coefficient used for prediction.
(0), X (1), ..., A (0) corresponding to X (i),
(1), ..., A (i) are stored. How to determine the prediction coefficient will be described later. The arithmetic unit 14 multiplies the camera shake signal X (0) by the prediction coefficient A (0) to obtain the intermediate data W (0) by the relational expression of Equation 1.

[0016]

[Equation 1] Similarly, the arithmetic unit 14 obtains the intermediate data W (i) by the relational expression of Equation 2.

[0017]

[Equation 2] However, (i) is a natural number from 1 to (N). Further, the calculation unit 14 obtains the predicted camera shake signal (Y) from the intermediate data W (i) by the relational expression of Equation 3.

[0018]

[Equation 3] Alternatively, the calculation of the intermediate data W (i) is omitted, and
It asks like a relational expression of.

[0019]

[Equation 4] Next, with reference to FIGS. 2, 3 and 4, an embodiment of the present invention will be described.

In this embodiment, a camera shake is detected by an angular velocity sensor, an actuator is operated based on the detected value,
The present invention is applied to a device that changes the position of the lens of the imaging member with respect to the optical axis in conjunction with the actuator to prevent deterioration of a photographed photograph due to camera shake.

As shown in FIG. 3, for the camera,
Set three axes of x, y, and z. Although camera shake occurs in the horizontal and vertical directions of the screen, in order to simplify the explanation, we will explain the case of correcting camera shake in the horizontal direction of the screen, but also in other directions. It is the same.

FIG. 2 is a block diagram showing the configuration of the embodiment of the present invention. The angular velocity sensor 15 detects the angular velocity due to rotation around the y-axis in FIG. 3 in order to detect camera shake in the lateral direction of the camera. The sensitivity axis is adjusted so that
The angular velocity sensor 15 has an A / D converter (not shown) at its output stage, and its output is digitized.

Here, in this embodiment, it is assumed that the amount of camera shake is predicted from four camera shake signals at 10 msec intervals including the latest camera shake signal of the angular velocity sensor. That is, the prediction is performed using the data X (0) at the time when the prediction calculation is started and the past data X (1), X (2), and X (3) at 10 msec, 20 msec, and 30 msec from this time, respectively.

The output of the angular velocity sensor 15 is taken into the camera shake signal storage unit 12 composed of a RAM every 1 msec. Including the storage area for the latest current camera shake data, 31 storage areas are required for the storage area. The past data exceeding 30 msec is updated by new data.

The camera shake prediction coefficient storage unit 13 is composed of a RAM. The coefficient (A (0)) for the current latest camera shake signal and 10 msec and 20 mse, respectively.
c, coefficients (A (1), A (2), A (3)) for camera shake signal data in the past 30 msec are stored.

In the multiplication operation unit 16, the latest camera shake signal of the present time is selected from X (0) and the camera shake signal of the past stored in the camera shake signal storage unit 12, and 10 msec, 20 ms.
ec, 30 msec past data X (1), X (2),
The calculation result of the data at each time is output as intermediate data by X (3) and the coefficients A (0) to A (3).

That is, assuming that the intermediate data is W (i), the same relational expression as in the above equation 2 W (i) = (X (i) × A (i)) (where (i) is from 0 to 3) A natural number of).

This intermediate data (W) is added by the camera shake signal prediction calculation unit 17, and the addition result is used as the predicted camera shake data. That is, assuming that the predicted camera shake data is Y, the camera shake signal prediction calculation unit 17 performs the calculation represented by the relational expression (5) to calculate the predicted camera shake data (Y).
Is output.

[0029]

[Equation 5] Note that this calculation may be directly performed by the calculation represented by the relational expression of Equation 6 without using the intermediate data (W).

[0030]

[Equation 6] The camera shake prediction coefficient storage unit 13, the multiplication calculation unit 16, and the camera shake signal prediction calculation unit 17 are configured by a CPU 18.

The actuator control circuit 19 controls the speed of the actuator 20 based on the predicted camera shake signal (Y) so as to cancel the predicted camera shake signal. Further, an image pickup unit 21 is mechanically connected to the actuator 20, and moves the position of the optical axis in the x-axis direction with respect to the lens of the camera of the image pickup unit 21 according to the operation of the actuator 20. This prevents deterioration of the image of the subject on the imaging unit 21 due to camera shake. These series of operations are performed within 1 msec, and 1 mse
The camera shake prevention is performed based on the predicted camera shake signal for each c.

Next, referring to the flow chart of FIG.
The operation of the embodiment will be described. Here, the shake prediction coefficient storage unit 13 and the multiplication calculation unit 1 are provided in the microcomputer.
6. The flow of the program when the camera shake signal prediction calculation unit 17 is incorporated will be described. This calculation subroutine is called at the initial stage of the photographing operation such as turning on the power supply and pressing the release button.

First, in step S1, the coefficient (A) necessary for predicting the camera shake is stored in the camera shake signal prediction coefficient storage unit 13.
(0), A (1), A (2), A (3)) are read.
Then, in step S2, the start address of the continuous PAM area for storing the camera shake signal is stored in the data pointer used for writing the camera shake signal in the free write / read storage area (RAM) of the microcomputer. Substitute (# V0). 31 including the latest shake signal
This is the start address of the area in which the individual pieces of data are stored.

Then, in step S3, a pre-data counter as a counter for detecting that the camera shake signal necessary for the prediction is stored in the camera shake signal storage unit 12 (RAM area) is cleared. This counter counts up until the first 31 pieces of data are stored.

Next, the output of the angular velocity sensor is read from the angular velocity sensor 15 as a camera shake signal in step S4. Next, in step S5, this angular velocity data is stored in the RAM area indicated by the above-mentioned data pointer by using a general indirect addressing method.

Then, in step S6, the value of the pre-data counter is checked to see if a sufficient number of data for prediction has already been stored in the camera shake signal storage unit 12. If sufficient data is not yet stored, the process proceeds to step S7, and if all the necessary data are available, the process proceeds to step S13 described later.

When the process proceeds to step S7, the value of the data pointer is first incremented here to indicate the area of the next data. Next, in step S8, it is checked whether or not the value of the data pointer exceeds the area for storing the camera shake signal. In the case of the embodiment, as described above, since 31 pieces of data including the latest camera shake signal are stored, when the storage area exceeds 31 pieces from the start address, the data is stored in step S9. The pointer value is reset to the start address (# V0).

Next, in step S10, it is checked whether or not the pre-data counter is insufficient for prediction, and if it is insufficient, the process proceeds to step S11 to increment the pre-data counter by one.

Then, in step S12, it is checked whether or not the processing of this routine is completed. The signal for termination is determined by the switch operation of the photographer such as release of the release button, the elapsed time from the start of control, or the like. If the termination condition is met, the processing of this subroutine is terminated. On the other hand, if the termination condition is not met, the process returns to step S4.

If it is determined in the above step S6 that the past data sufficient for prediction has been accumulated, the process proceeds to step S13, and the latest camera shake signal storage indicated by the current data pointer value is stored. The data is read from the address and assigned to the variable (X (0)). Here, if the latest captured data is still held, that value may be substituted.

Next, steps S14 to S16
At, the data of 10 msec past is read. For that purpose, first, it is checked whether the address which is smaller by 10 from the current data pointer is larger than the start address (# V0) of the RAM area storing the camera shake signal. here,#
If it is larger than V0, in step S15,
The indirect address is used to read the data 10 msec past using address information that is smaller than the data pointer by -10 minutes, and substitute it into the variable (X (1)).

In step S14, the start address (#
If the value of the data pointer -10 becomes smaller than V0), the data of 10 msec past is stored in the address 21 larger than the current value of the data pointer, so the process proceeds to step S16, Substitute the value into the variable (X (1)).

Next, steps S17 to S19
At the same time, the data in the past 20 msec is read by the same method as above and is substituted into the variable (X (2)). Further, in steps S20 to S22, 30 msec past data is read by the same method as above, and the variable (X
(3)) is substituted.

The present and the past 10 thus obtained
Using four shake signals of msec, 20 msec, and 30 msec and the prediction coefficient read in step S1,
In steps S23 to S27, the predicted camera shake signal (Y) is calculated. Here, the operations of the multiplication calculation unit 16 and the camera shake signal prediction calculation unit 17 are developed in software in a complex manner.

First, in step S23, the value of the result of multiplication (X (0) × A (0)) is substituted into the variable (Y) of the predicted camera shake signal. Then, from step S24 to step S
26, the variable (Y) is multiplied (X (1) × A
(1)) result, multiplication (X (2) × A (2)) result,
The multiplication (X (3) × A (3)) is successively added and substituted. The variable (Y) obtained as a result is the predicted shake signal value. Then, finally, in step S27, after the camera shake prediction signal is output, the process returns to step S7 and the above-described operation is repeated.

In the above-mentioned flow chart, the camera shake signal to the multiplication operation unit 16 is always fetched from the camera shake signal storage unit 12 including the latest data. However, when only the latest data is taken out from the angular velocity sensor 15 as the detection means, step S13 in the flowchart of FIG. 4 may be replaced with step S28 as shown in the flowchart of FIG. In this case, the other steps are the same as those in the flowchart of FIG. 4, and thus the description thereof is omitted here.

By the way, the coefficient stored in the camera-shake signal predicting means for the prediction calculation is determined by the amount of delay of the angular velocity sensor and the amount of mechanical delay for camera-shake correction. These delays are the time to the future we want to predict. Hereinafter, how to determine this coefficient will be described.

Now, consider the difference in the detected hand-shake signal. The difference on the first floor is Δ1, the difference on the second floor is Δ2, and the difference on the third floor is Δ3. Then, it can be expressed as the relational expressions of the formulas 7, 8 and 9, and accordingly, the formulas 10, 11 and 12
The relational expression of is required.

[0049]

[Equation 7]

[0050]

[Equation 8]

[0051]

[Equation 9]

[0052]

[Equation 10]

[0053]

[Equation 11]

[0054]

[Equation 12]

The difference between the predicted camera shake signal (Y) and the latest camera shake signal X (0) is Δ1 (Y).
Then, from the relational expressions of Equations 13, 14, and 15, Equation 1
The relational expressions 6, 6, 17 and 18 are obtained.

[0056]

[Equation 13]

[0057]

[Equation 14]

[0058]

[Equation 15]

[0059]

[Equation 16]

[0060]

[Equation 17]

[0061]

[Equation 18] Further, the time interval for detecting the camera shake signal (X) is Δt, and the time to be predicted is (S). Moreover, the ratio is set to (k). Then, the relational expression of Expression 19 is established.

[0062]

[Formula 19]

Assuming that the hand movement signal changes in time with a certain degree of smoothness, its smoothness is now 3
Assume the following derivative is a constant. Now, since a discrete camera shake signal is handled, in order for the camera shake signal to make a smooth transition, the third order difference should be proportional to the data interval time. In other words, it becomes like the relational expression of Expression 20.

[0064]

[Equation 20] From the relational expression of the equation 20 and the relational expressions of the equations 18, 17, and 16, the relational expressions of the equations 21, 22, and 23 are obtained.

[0065]

[Equation 21]

[0066]

[Equation 22]

[0067]

[Equation 23] Then, the relational expression of the equation 23 and the equations 7, 10 and 12
From the relational expression of, the relational expression of Equation 24 is obtained, and therefore, Equation 25
The relational expression of is established.

[0068]

[Equation 24]

[0069]

[Equation 25]

As expressed by the relational expression (25),
By multiplying the past camera shake signal and the current camera shake signal by a predetermined coefficient, the future camera shake signal can be predicted.
This coefficient is the prediction coefficient. In this example, the ratio of the predicted time and the time interval of the data regarding the difference of the third floor is considered to be proportional. However, in order to improve the accuracy of the prediction of the camera shake signal, a higher order difference is further considered. Of course, it is possible.

In some cases, it may be possible to judge that the accuracy of prediction is insufficient with the coefficient for prediction set by the above method. It is considered that this is because the vibrations peculiar to camera shake and the response of the actuator cannot be described in the form of a simple linear function.

In order to reduce this error, the camera shake signal is measured at the stage of designing and manufacturing the camera shake preventive device, and this coefficient is set so that the difference between the predicted camera shake signal and the actual camera shake signal is minimized. You may. As an example, by performing multiple regression analysis using the current and past camera shake signals, the coefficient of the weight of each camera shake signal can be obtained by the least square method.

Concretely, an example of optimizing the prediction coefficient when predicting the camera shake signal after the delay time of the detection of the camera shake detection sensor is as follows. The number of data and the number of measurements can be increased as necessary.

Considering the camera shake signal obtained in the experiment, it is now assumed that 200 samples obtained in 2 seconds at a sampling interval of 1 msec.
One shake signal is B (j) (j is 0 to 2000). In addition, this camera shake detection sensor outputs 30
It is assumed that it has a delay of msec.

Now, for the data at the same time, the camera shake signal sequence C0 (j) represented by the subscript (j) having the same value,
Consider C1 (j), C2 (j), C3 (j).
Here, C0 (j) is the latest camera shake signal at the time of prediction, and C1 (j), C2 (j), and C3 (j) are each 10 msec past camera shake signal, 20 mec past data, and 30 msec past. The data of
In addition, the actual amount of camera shake after the time to be predicted is D (j). When the above-mentioned actually measured camera shake signal is applied to this signal train, it becomes as shown in Table 1.

[0076]

[Table 1] The data set used for prediction and the reference data set is 1
There will be 941 pieces.

Here, the prediction coefficients are A0, A1, A2, A
Set to 3. These are 0 msec and 10 mse, respectively.
c is a coefficient for multiplying past data of 20 msec and 30 msec. An arithmetic expression for prediction is expressed as a relational expression of Expression 26, where Y (j) is a predicted value at time (j).

[0078]

[Equation 26] Error e between the actual camera shake signal and the predicted value at time (j)
(J) is expressed by the relational expression of Expression 27.

[0079]

[Equation 27] Then, the sum of squares (E) of the errors is expressed as the relational expression of Expression 28.

[0080]

[Equation 28]

Then, based on the least squares method, a combination of prediction coefficients that minimizes the sum of squares (E) of the errors is obtained. For that purpose, consider a matrix (G) represented by the relational expressions of the equations (29) and (30). (J) is 0 to 194 here
It is a value of 1.

[0082]

[Equation 29]

[0083]

[Equation 30] Then, the inverse matrix (G) ⁻¹ of the matrix (G) is obtained from the relational expression of Expression 31.

[0084]

[Equation 31]

By performing such an operation and determining the prediction coefficient, it is possible to optimize the prediction coefficient in accordance with the actual occurrence of camera shake. This coefficient may be stored in the storage means in advance. Here, an example in which the delay of the output of the camera shake detection sensor is corrected has been shown, but if the delay of the actuator and the like are taken into consideration, a more practical prediction is performed.

Further, in the above-mentioned example, an example of predicting the value of the same camera-shake detecting sensor after a certain time from the actual value of the camera-shake detecting sensor is shown. However, optimization of these prediction coefficients is performed by designing the camera. Considering that it is performed in the stage or manufacturing stage, and is stored in advance in the camera shake prediction coefficient storage means of the camera before the photographer operates it, in the prediction coefficient optimization step, a more accurate camera shake detection device is used. By using the camera shake signal as a reference and optimizing the prediction coefficient for prediction by the camera shake detection sensor in the camera, it is possible to reduce the error of the camera shake detection sensor mounted in the camera.

Further, when the focal length is changed like an interchangeable lens or a zoom lens and the relationship between the angular velocity and the blurring of the image changes, the correction can be made by storing the coefficient of this prediction according to the focal length. It will be possible. This can also be achieved by correcting the same prediction coefficient according to the focal length. Further, even when the distance changes depending on the subject distance, the accuracy can be improved by storing the coefficient according to the distance or correcting the coefficient according to the distance.

In the above-described embodiment, the angular velocity sensor is used as the camera shake signal detecting means. However, even if this is a camera shake detection using an image pickup device such as a CCD, of course, in principle, There is no difference in effect. next,
The time interval of data used for prediction will be described.

In the above-mentioned embodiment, the interval time (Δt) of this data is 10 msec. However, this time is not limited as long as it is a time that allows the sample to sufficiently trace the trajectory of the camera shake signal.
In order to sufficiently represent the camera shake, a value that can be sufficiently detected with respect to the high frequency of the camera shake signal may be used.

According to the conventional analysis of the phenomenon of camera shake, the frequency of camera shake causing image deterioration is 10 H at maximum.
Since it is required to be about z, it is sufficient if the interval can accurately detect this 10 Hz. According to the sample theorem, this may be detected at a frequency of 8 times or more. That is, the time interval of the prediction data is 12 mse
It is considered that a speed higher than c is sufficient. When the prediction time is long, the interval is set long, and when the prediction time is short, the interval is set short to improve the accuracy of the prediction.

Further, when the data used for prediction includes noise, it is possible to reduce the noise by means of software such as a low-pass filter or averaging. In this case, it is possible to prevent the signal delay from being corrected by also predicting the signal delay due to this filter.

[0092]

As described above, according to the present invention, a straight line or two lines obtained by using a complicated least square method or the like is used for signal prediction.
There is no need for complicated calculations such as approximating to a higher-order regression line of order higher or higher and extrapolation, and the load on a computer or the like is reduced when applied to a small popular product such as a camera or a video camera. At the same time, even if there is a time delay in the sensor that detects camera shake or a response delay due to the mechanical time constant of the camera shake correction mechanism, the effect of camera shake prevention can be prevented from decreasing by accurately predicting the camera shake signal, and there is camera shake. Even if the image blurs, the deterioration of the image quality can be minimized.

[Brief description of drawings]

FIG. 1 is a block diagram showing the concept of a camera shake prevention device of an image pickup apparatus according to the present invention.

FIG. 2 is a block diagram showing a basic configuration of a camera shake prevention device for an image pickup apparatus according to an embodiment of the present invention.

FIG. 3 is a diagram showing a relationship between an x-axis, a y-axis, and a z-axis with respect to a camera.

FIG. 4 is a flowchart illustrating the operation of the embodiment of the present invention.

FIG. 5 is a flowchart explaining the operation of another embodiment of the present invention.

[Explanation of symbols]

11 ... camera shake signal detection unit, 12 ... camera shake signal storage unit, 1
3 ... Hand shake prediction coefficient storage unit, 14 ... Calculation unit, 15 ... Angular velocity sensor, 16 ... Multiplication calculation unit, 17 ... Hand shake signal prediction calculation unit, 19 ... Actuator control circuit, 20 ... Actuator, 21 ... Imaging unit.

Claims (3)

[Claims] 1. An image stabilization device in an optical device for recording an image of a subject, an image stabilization device for detecting image stabilization vibration, and an image stabilization device for storing at least one output from the image stabilization signal detection device. A signal storage means and at least two for predicting a camera shake signal after a predetermined time
A camera shake prediction coefficient storage unit that stores more than one kind of prediction coefficient, and a predicted camera shake signal by multiplying and adding the output of the camera shake signal detection unit and / or the output of the camera shake signal storage unit and the output of the camera shake prediction coefficient storage unit. A camera shake signal predicting device comprising: a calculating unit for calculating. 2. The camera-shake signal predicting device according to claim 1, wherein the camera-shake signal detecting means is set at a time interval at which an output of the camera-shake signal detecting means can detect a camera shake vibration of at least 10 Hz. 3. The camera shake signal prediction apparatus according to claim 1, wherein the camera shake prediction coefficient is a coefficient optimized so that an error between the prediction data and the actual measurement data becomes small.