JPH10215401A - Auto-focus device for video camera - Google Patents

Abstract

PROBLEM TO BE SOLVED: To grasp the change of evaluation value that is caused by the movement of a focus lens by reducing relatively the changing rate of the evaluation value despite the mutual shakes which are caused between a video camera and its object. SOLUTION: An AF(autofocus) block 137 has an AF-IC 139 and an AF CPU 138. The AF-IC 139 includes a luminance signal generation circuit and 14 evaluation value generation circuits. Every evaluation value generation circuit generates the evaluation value by the luminance signal Y of the luminance signal generation circuit and the HD, VD and CK timing signals of a timing signal generation circuit 125. The CPU 138 performs the calculation and every processing in response to the algorithm corresponding to each evaluation value. The shake is decided based on the evaluation value showing the changing rate of the evaluation value per field and the normalized difference value of the luminance addition value. Then the speed of an AF lens is controlled based on the shake decision result.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to automatic focus adjustment (auto focus; hereinafter, abbreviated as "AF") of a news or professional video camera. Focus adjustment is to determine a so-called just focus state (hereinafter abbreviated as "JP"). More specifically, the present invention relates to an object for automatically recognizing a position of an object photographed by a broadcast station, a video camera used for business use (including a television camera, a studio camera, an ENG camera, etc.). The present invention relates to a recognition device and an imaging device that focuses on a subject using the subject recognition device.

[0002]

2. Description of the Related Art Conventionally, an AF mechanism for automatically focusing on a subject in a video camera has already been realized in the field of consumer video cameras.

At this time, it is well known that it is only necessary to determine whether the contrast of the image signal is high or low in order to detect whether or not the image is in focus. That is, when the contrast is high, the focus is on, and when the contrast is low, the focus is off. By extracting the high frequency component of the imaging signal, generating data by integrating the high frequency component present in a predetermined setting area in the screen, and using the integrated data, it is possible to determine the level of the contrast. This integrated data is data indicating how many high-frequency components are present in the set area, and this data is generally called an evaluation value. Therefore, AF can be realized by driving the focus lens so that the evaluation value is maximized, that is, the contrast is maximized.

[0004]

Consumer video cameras do not require a very precise AF mechanism due to their properties. With the AF mechanism provided in current consumer video cameras, when the subject changes or fluctuates rapidly, the time to follow it is relatively short, but after that, the state of searching for the subject continues, and the image is finely defined. According to observation, the in-focus state (JP) and the out-of-focus state (blurred state) are continuously repeated.

[0005] This is because the evaluation value of the current AF mechanism is changed not only by the change of the focus state of the target subject but also by the change of various factors such as the background and brightness of the image. is there. These factors cannot be distinguished from the value of the evaluation value, and the evaluation value extracted by such a method is by no means an accurate evaluation value representing the focus state of the subject. It is very difficult to solve this problem with a simple AF mechanism.

A major selling point is that consumer electronics are inexpensive, and there has been no request for an AF mechanism that is more precise but more expensive for consumer video cameras.

However, professional or professional equipment used in broadcasting stations and the like requires extremely high precision and definition due to its properties. For example, in a video camera used for a broadcasting station or for business use, a captured video may be transmitted to a home through live broadcasting (live broadcasting). If it is not possible to obtain a quick and accurate evaluation value during such live broadcasting, AF operation takes time, and an in-focus blurred image signal is transmitted to the home. Will be transmitted.

Therefore, a video camera used for a broadcasting station or for business use is completely different from a simple and inexpensive small AF device used in a consumer video camera. The AF mechanism of a video camera cannot be introduced as it is.

Furthermore, since a consumer video camera does not require such a clear image, the camera is used in a state where the lens is small and the depth of field is very deep, so that the focus can be easily adjusted. In other words, when the distance to the subject exceeds 10 m, even if the distance from the lens to the subject changes by a little over 1 m before and after JP, it is not determined as "blur" on the screen.

On the other hand, a camera of a broadcasting station requires a very clear image and is used in a state where a lens is large and a depth of field is very shallow. That is, even if the distance from the lens to the subject changes slightly back and forth,
It is determined to be "blurred". Therefore, it is necessary to adjust the focus more accurately by the shallow depth of the subject.

Here, "depth of field" and "depth of focus" which will be described later will be briefly described. The depth of field (also referred to as the "depth of subject") is the point closest to and farthest from the camera that is accepted as being sharp without blur when the camera lens is fixed at a certain lens set position. And the distance. That is, the total distance at which satisfactory sharpness is obtained when the lens is focused at a specific distance. In contrast to consumer video cameras, professional video cameras can have shooting conditions with a deeper depth of field. Generally, the focus adjustment characteristic of a camera is represented by the depth of field. With a camera with AF function, if the ranging error is within the depth of field of the video camera used for shooting (| depth of field |> | ranging error |), a good image that seems to be in focus is obtained. I can take it.

On the other hand, the depth of focus refers to a range of the image distance corresponding to the range before and after the object distance covered by the depth of field. In other words, this refers to the distance between the lens and the imaging surface (CCD charge-coupled device) where the focused image of the subject is allowed within a range of acceptable sharpness.

As shown in FIG. 39A, an object point (subject) P
When the focus is on the CCD surface at the position P0, the distance between the point P1 farthest from the lens (camera) and the point P2 closest to the lens (camera) which is accepted as being clear with this lens and the CCD fixed, with the lens fixed. Say. On the other hand, as shown in FIG. 39B, image distance ranges P1 'to P2' corresponding to the front and rear ranges P1 to P2 of the object distance covered by the depth of field.
Is called the depth of focus.

[0014] It should be noted that "the state in which there is no blur and it is acceptable to be sharp" means that the judgment is made with the naked eye.
It may be defined using an allowable circle of confusion. The former is a so-called sensory test with human eyes. The latter, FIG. 40A
It is assumed that a circular blur having a convergent diameter d is originally generated at one point as shown in FIG. At this time, the defocus amount δ is a small cone ab
c. The conical bottom surface when the maximum allowable blur occurs is called an allowable circle of confusion. In the case where a 2/3 inch CCD for broadcasting business is used, when the diameter d of the bottom surface of the small conical ABC is 22 [μm] or less, there is no blur and a state in which it is accepted as clear. I do. Therefore, d =
The bottom surface of the small cone ABC at 22 [μm] is called an allowable circle of confusion, and the out-of-focus amount δ at this time is the depth of focus.

The basis for setting the diameter of the permissible circle of confusion d ≦ 22 [μm] as shown in FIG. 40B is that pixels (pixels) arranged in a regular lattice of a 2/3 inch CCD used in a video camera for broadcasting stations are used. )) (Pixel pitch).
That is, if it is less than this, the image spreads to the image adjacent to the pixel at the image formation position, and does not adversely affect the adjacent pixel.

In general, a consumer video camera often makes a judgment based on a sensory test, and a business video camera often makes a judgment based on an allowable circle of confusion.

Conventionally, it has been desired to develop a very high-precision AF mechanism suitable for a professional or professional video camera.
Until now, attempts have been made to develop a high-precision AF mechanism suitable for professional or professional video cameras, but they have failed to satisfy the high demands of the broadcaster's cameraman, and all have failed. Therefore, the actual situation is that manual focusing is still performed depending on the craftsmanship of the photographer.

In the present application, as described in the detailed description of the invention, A suitable for professional or professional video cameras
F mechanism is being developed. In this AF, JP is found by searching for the peak of the evaluation value while moving the focus lens. Here, when determining the moving speed of the focus lens, if the distance is far from the peak of the evaluation value, the time is too long if the AF is too slow. On the other hand, if the speed is too close to the evaluation value peak relatively, there is a problem that the evaluation value peak is greatly passed.

Further, when there is a fluctuation between the video camera and the subject, the evaluation value is affected by the fluctuation,
There is a problem that the ratio of the change in the evaluation value due to the movement of the focus lens and the change in the evaluation value due to shaking is unknown.

Further, when the shake determination is introduced, the manual operation of the AF switch causes a slight change in the motion evaluation value of the video camera, which causes a problem that the shake determination becomes inaccurate.

Accordingly, it is an object of the present invention to provide an extremely accurate AF mechanism suitable for a professional or professional video camera in view of the above-mentioned problems. It is a further object of the present invention to provide a video camera having an extremely accurate AF mechanism suitable for a professional or professional video camera. It is a further object of the present invention to provide a very accurate AF method suitable for professional or professional video cameras.

Further, according to the present invention, when determining the moving speed of a focus lens in an AF mechanism suitable for a professional or professional video camera, the focus lens can be quickly and without exceeding an evaluation value peak in response to various conditions. It is an object to provide a moving speed determining means.

Further, according to the present invention, even when the video camera and the subject shake, the rate of change of the evaluation value is relatively reduced so that the change of the evaluation value due to the movement of the focus lens can be grasped. The purpose is to provide.

Further, according to the present invention, when swing determination is introduced,
An object of the present invention is to provide a shake determination unit that takes into account camera shake that occurs when an AF switch is manually operated.

[0025]

According to the present invention, there is provided an autofocus apparatus comprising: input means for inputting an image signal of a video camera; and an evaluation value for extracting a high-frequency component of a specific region of the image signal to generate an evaluation value. Generating means, signal processing means for calculating a command value to be given to a focus drive unit of the lens of the video camera according to the evaluation value, and output means for sending the command value to a drive unit of the lens of the video camera. An auto-focusing device comprising: a circuit for calculating a value obtained by adding the luminance of the image signal in the specific area,
The evaluation value and the rate of change of the luminance addition value during a predetermined period are checked to determine whether or not the video camera and the subject shake, and the speed of the focus lens during autofocusing is controlled in accordance with the presence or absence of the shake. ing.

Further, in the auto-focusing device according to the present invention, in the above-mentioned auto-focusing device, when it is determined that there is further shaking, the speed of the focus lens is made relatively high, and conversely, it is determined that there is no shaking. If
The lens moves from the blurred state to the vicinity of the just pin at a first lens speed, and moves from the vicinity to the just focus at a second lens speed relatively lower than the first lens speed.

Further, in the autofocus apparatus according to the present invention, in the above autofocus apparatus, switching from the first lens speed to the second lens speed is performed in accordance with the rate of change of the evaluation value. .

Further, in the autofocus method according to the present invention, an image signal of a video camera is input, a high-frequency component of a specific region of the image signal is extracted to generate an evaluation value, and an evaluation value is generated within the specific region of the image signal. A value obtained by adding the luminance is calculated to generate a luminance addition value, and the evaluation value and the rate of change of the luminance addition value during a predetermined period are checked. This is an autofocus method for a video camera including various steps of judging the presence or absence of a shake from one of the methods.

[0029]

Embodiments of the present invention will be described below in detail with reference to the accompanying drawings. In this embodiment, a description will be given of a hardware configuration and an operation algorithm of an AF of a broadcasting station or a commercial video camera developed this time. Further, a description will be given of the hardware configuration for creating each evaluation value of AF, the types and properties of the evaluation values, and the like.

Since the description of the embodiment is long, a table of contents will be described first, and then the description will be made in the order of the table of contents. Also, the symbols used in the drawings are basically 4
The upper two digits correspond to the drawing number, and the lower two digits are sequentially given to each element. However, with respect to the figure numbers 1 to 9, the thousands place zero is omitted. Also,
Each step in the flowchart is further identified with an "S" at the top.

Table of Contents 1. Hardware configuration Evaluation value 3. AF algorithm 3.1 Overall flowchart 3.2 Shaking judgment (when the lens is stationary) 3.3 Wobbling 3.4 Initialization of hill-climbing parameters 3.5 Update of maximum and minimum values 3.6 Lens speed setting 3.7 Saturation luminance judgment 3 .8 False Mountain Judgment 3.9 Shake Judgment (Lens Operation) 3.10 Direction Judgment (Climbing Up) 3.11 Reverse Feed Judgment 3.12 Reverse Feed Inversion Process 3.13 Edge Reaching Determination 3.14 Edge Inversion Process 3. 15 Midway judgment method 3.16 Peak selection 3.17 Mountain descent judgment 3.18 Peak position calculation (centroid processing) 3.19 Move to peak position 3.20 Long filter

1. Hardware Configuration FIG. 1 shows the overall configuration of a video camera according to the present invention. The video camera has a lens block for optically condensing incident light on the entire surface of the image sensor, and converts the incident light from the lens block into electrical image signals of R, G, and B (red, green, and blue). An imaging block for conversion and a signal processing block for performing predetermined signal processing on the imaging signal are separated. As a control unit, a lens block control C for controlling a lens block is used.
PU114 and a main CPU 141 for controlling a video camera body including an imaging block and a signal processing block
And an AF block 137 having an AF dedicated CPU 138 for generating an evaluation value, which is a parameter used for AF, from the image pickup signals RGB.

This lens block is moved as an optical element in the direction of the optical axis, so that the focus lens 111 drives the focus on the target subject (target) to JP, and the focus lens should advance at the start of AF. A wobbling lens 112 used for a wobbling operation described later to find the direction;
An iris mechanism 113 for adjusting the opening amount.

The lens block further includes a focus lens position detection sensor 11 for detecting the position of the focus lens 111 in the optical axis direction of the focus lens.
1a, a focus lens drive motor 111b for moving the focus lens in the optical axis direction, and a focus lens (FL) drive circuit 111c for providing a drive control signal to the focus lens drive motor. Similarly, for the wobbling lens 112, a wobbling lens position detection sensor 112a for detecting the lens position of the wobbling lens in the optical axis direction, a wobbling lens drive motor 112b for moving the wobbling lens in the optical axis direction, and a wobbling lens drive A wobbling lens (WOB) drive circuit 112c for providing a drive control signal to the motor. Similarly, for the iris mechanism 113, an iris position detection sensor 113a for detecting an opening position, an iris mechanism driving motor 113b for opening and closing the iris mechanism,
An iris drive circuit 113c for supplying a drive signal to the iris drive motor.

It should be noted that the lens block CPU 114
The focus lens driving circuit 111c, the wobbling lens driving circuit 112c, and the iris driving circuit 113c are electrically connected to each other and supplied with a control signal. The detection signals from the focus lens position detection sensor 111a, the wobbling lens position detection sensor 112a, and the iris lens position detection sensor 113a always
It is sent to the lens block CPU 114. Also,
The lens block CPU 114 controls the focus lens 1
11 and the focal length data of the wobbling lens 112,
It has a ROM (or an EEPROM) in which aperture ratio data, the manufacturer name and serial number of the lens block are recorded, and these data are stored in the CPU 114 for the lens block.
It is read based on a read command from. Further, an AF switch 115 for instructing activation of AF is attached to the lens block CPU 114, and the lens block CPU 114 transmits switch information to the main CPU 141 and the AF block 13 of the camera body.
7 is transmitted.

The imaging block further includes a color separation prism 121 for separating the incident light from the lens block into three primary colors of R, G, and B, an R component, a G component, and an R component separated by the color separation prism. B component light is imaged on the imaging surface,
The imaging light of each formed color component is converted into electrical imaging signals R and G.
And image sensors 122R and 12 which convert and output the images to B and B, respectively.
2G and 122B. For example, the imaging devices 122R, 122G, and 122B are composed of CCDs (charge coupled devices).

The imaging block includes preamplifiers 123R, 123G, and 123B for amplifying the levels of the imaging signals R, G, and B respectively output from the imaging device, and performing correlated double sampling for removing reset noise. Have.

Further, the image pickup block, based on a reference clock from a reference clock generation circuit provided therein,
A timing signal generation circuit 125 for generating a VD signal, an HD signal, and a CLK signal serving as a basic clock when each floor in the video camera operates, and a VD signal, an HD signal, and a CLK signal supplied from the timing generation circuit. C for giving a drive clock to the image sensor 122R, the image sensor 122G, and the image sensor 122B based on the
And a CD drive circuit 124. The VD signal is a clock signal representing one vertical period, the HD signal is a clock signal representing one horizontal period, and the CLK signal is a clock signal representing one pixel clock. And a timing clock comprising the CLK signal are supplied to each circuit of the video camera via the main CPU 141, though not shown.

The signal processing block is a block for performing predetermined signal processing on the imaging signals R, G, and B supplied from the imaging block. The signal processing block includes A / D conversion circuits 131R, 131G, 131 for converting the imaging signals R, G, B from an analog format to a digital format, respectively.
B and gain control circuits 132R, 132G, 1 for controlling the gains of the digital video signals R, G, B, respectively, based on the gain control signals from the main CPU 141.
32B and signal processing circuits 133R, 133G, which perform predetermined signal processing on the digital video signals R, G, B,
133B. These signal processing circuits include, for example, a knee circuit for compressing a video signal at a certain level or higher, a gamma correction circuit for correcting the video signal level according to a set gamma curve, and a predetermined black level and a predetermined level or higher. And a B / W clipping circuit for clipping the white level. The signal processing circuit 133
R, 133G, 133B are a knee circuit, a γ correction circuit, and B
/ W clip circuit, a known black γ correction circuit,
It may have an outline emphasis circuit, a linear matrix circuit, or the like.

The signal processing block further receives the video signals output from the signal processing circuits 133R, 133G, and 133B, respectively, and outputs the luminance signal Y and the color difference signals (RY), (B) from the video signals R, G, and B. −Y).

The signal processing block further receives the video signals R, G, and B output from the gain control circuits 132R, 132G, and 132B, respectively.
AF for performing AF based on 2R, 132G, 132B
Block 137, signal processing circuits 133R, 133G,
An iris control circuit 135 that receives the video signals R, G, and B output from the 133B and controls the iris based on the signal levels so that the amount of light incident on the image sensor becomes an appropriate amount; and a signal processing circuit 133R. , 133
And a white balance control circuit 136 for receiving video signals output from the G and 133B and performing white balance control based on the signal levels.

The AF block 137 includes a newly developed autofocus integrated circuit (AF-IC) 139 and an AF CP.
U (model SH7034 of the applicant) 138, and the AF-IC (139) includes a luminance signal generation circuit (not shown) and an evaluation value generation circuit (not shown) prepared for each evaluation value. )). Each evaluation value generation circuit receives the luminance signal Y from the luminance signal generation circuit and each of the HD, VD, and CK timing signals generated by the timing generation circuit 125, and generates each evaluation value. AF
The CPU 138 according to the AF algorithm (described later) stored in a storage device (not shown) in advance.
The calculation and various processes according to the AF algorithm are performed on each evaluation value, and the data calculated by the calculation is updated once in one field and accumulated. Details of the type and role of the evaluation value will be described after the item “2. Evaluation value”.

The AF CPU 138 includes an AF-IC (13
For 9), data such as a fast-distance area size (evaluation frame size) and an image type (NTSC / PAL, pixel clock, etc.) are set, and an evaluation value is received from the AF-IC (139). Also, various kinds of data (focus position, iris value, etc.) of the camera are received through serial communication with the CPU 114 on the lens block side, focus command data is calculated, and the lens block side C through serial communication.
The focus instruction data is sent to the PU 114. The AF algorithm that is the operation software of the AF CPU 138 will be described in the item “3. AF software”.

Although not shown in the figure, the iris control circuit 135 selects a NAM circuit for selecting a signal having the highest signal level among the supplied video signals RGB and an area on the screen of the selected signal. It has a total integration circuit that divides and integrates the video signal for each area. The iris control circuit 135 determines all lighting conditions, such as back light proof, normal light proof, flat lighting, and spot lighting, of the subject based on the integrated data for each area, and controls an iris control signal for controlling the iris. And sends this iris control signal to the main CPU 141.
The main CPU 141 sends a control signal to the iris drive circuit 113C via the lens block side CPU 112 based on the iris control signal.

From the supplied video signals R, G and B, the white balance control circuit 136 calculates (RY) = 0,
A white balance signal is generated so that (BY) = 0, and this white balance signal is generated by the main CPU 141.
To send to. The main CPU 141 controls the gain control circuits 132R, 132R based on the white balance signal.
G and 132B are supplied with a gain control signal.

Generally, among video cameras, a lens block is manufactured by an optical device manufacturer, and a video camera body including other imaging blocks and signal processing blocks is manufactured by an electronic device manufacturer such as the present applicant. The lens block and the video camera body are interchangeable lens types that can be attached to the camera body with lenses from any manufacturer as long as they correspond to a predetermined format, protocol, and command data content. That is, the lens block side CPU 114
Transmits information corresponding to various information requests (focus position, iris value, etc.) sent from the video camera main body through serial communication to the main CPU 141 and the AF of the camera main body.
To the CPU 138 for use. Conversely, the lens block CPU 114 communicates with the main CPU 141 and the AF CPU 137 of the camera body through a serial communication line.
In response to the focus control command and the wobbling control command sent from the controller, the focus lens driving circuit 111c, the wobbling lens driving circuit 1122c, and the iris driving circuit 113c are respectively controlled.

Next, the AF block 137 will be described in detail with reference to FIG. AF-IC (139)
Are 14 evaluation values (ID0 to ID0, as described later).
It is specified by ID13. ), A luminance signal generation circuit 201 and an evaluation value generation circuit 20 for each evaluation value.
Has two. The luminance signal generation circuit 201 is a circuit that generates a luminance signal Y from the supplied video signals R, G, and B. In order to determine whether the image is in focus or out of focus, it is sufficient to determine whether the contrast is high or low. Since the change in the contrast is not related to the change in the level of the color signal, it is possible to determine whether the contrast is high or low by detecting only the change in the level of the luminance signal Y. The luminance signal generation circuit 201
A luminance signal Y can be generated by performing a known operation (that is, Y = 0.3R + 0.59G + 0.11B) on the supplied video signals R, G, and B.

The evaluation value generation circuit 202 has 14 evaluation value generation circuits for generating 14 types of evaluation values.
The evaluation value generation circuit includes a horizontal direction evaluation value calculation filter circuit, a total integration type horizontal direction evaluation value calculation filter circuit, and a saturated luminance number calculation filter, and details thereof are described in “2.2.
This will be described in connection with "type of evaluation value".

2. Evaluation Value The following describes the nature of the evaluation value used in the AF of the present embodiment. The evaluation value is basically the sum of high-frequency components in the distance measurement area of the image, and is generally an amount that has a correlation with the degree of JP and inversely correlates with the degree of blur of the image. In this AF system, the evaluation value is an AF-IC (autofocus integrated circuit) 139 that has been developed exclusively.
Is calculated by AF includes an operation of moving the focus and driving the focus to JP while judging an increase or decrease of the evaluation value accompanying the focus.

There are 14 types of evaluation values used in the actual AF, which are specified by ID0 to ID13. These evaluation values are listed. The evaluation values ID7, ID8, ID1
Since 0 and ID11 are not related to the high-frequency components of the image data, they cannot be said to be the original evaluation values from the definition of the evaluation values described above, but the focus is driven into JP while judging whether the evaluation values increase or decrease. Since they play an auxiliary role above, they are handled as evaluation values in this embodiment.
The ID # is an identification symbol of the evaluation value given on the program, and is used in this specification when specifying the evaluation value.

ID # evaluation value name Remarks 0: IIR1-W1-HPeak Basic evaluation value 1: IIR1-W2-HPeak 2: IIR4-W1-HPeak 3: IIR4-W3-HPeak 4: IIR0-W1-Vintg5: IIR5 -W1-VIntg 6: IIR1-W1-HIntg 7: Y-W1-HIntg Brightness addition value 8: Y-W1-Satul Saturation brightness number 9: IIR1-W3-HPeak 10: IIR1-W4-HPeak 11W: IIR -HPeak 12: Y-W3-HIntg luminance addition value 13: Y-W5-HIntg luminance addition value

The name of the evaluation value assigned to each evaluation value of 1D0 to ID13 is "use data-
Frame Size—Evaluation Value Calculation Method ”. For example, as the evaluation value ID0, the data specified by “IIR1” is used, the evaluation frame size of “W1” is used, and “HPea” is used.
The evaluation value obtained by using the evaluation value calculation method of “k”.
As described above, the evaluation value name directly indicates the attribute of the evaluation value.

The usage data of the evaluation value name is roughly classified into "II
R "and" Y ". There is an IIR4 that uses high-frequency component data extracted from a luminance signal using an HPF (high-pass filter), and a Y that uses the luminance data of an imaging signal without using an HPF.

When an HPF is used, as shown in FIG. 3, an IIR type (infinite length impulse response type) HPF is used. Depending on the type of HPF,
The evaluation values IIR0, IIR1, IIR3, and IIR4 represent HPFs having different cutoff frequencies.

The reasons for setting HPFs having different cutoff frequencies are as follows. For example, an HPF with a high cutoff frequency is very suitable near JP. This is because the rate of change of the evaluation value is relatively large with respect to the movement of the lens near JP. Conversely, even if the lens is moved where the focus is largely shifted, the rate of change of the evaluation value is small, so that an HPF with a high cutoff frequency is not suitable where the focus is greatly shifted.

On the other hand, the HPF having a low cutoff frequency is
It is suitable for places where the focus is greatly shifted. This is because if the lens is moved even when the focus is largely shifted, the rate of change in the evaluation value is relatively large. Conversely, even if the lens is moved in the vicinity of JP, it is not suitable because the rate of change of the evaluation value is small. Since the rate of change in the evaluation value is relatively small with respect to the movement of the lens near JP, an HPF with a low cutoff frequency is not suitable near JP.

In order to generate evaluation values having different properties by using a plurality of filters having the opposite properties and to select an optimum evaluation value at that stage in the AF process, different cutoff frequencies are used. Is set.

When the HPF is not performed, the Y signal is a luminance signal Y without a filter of the image pickup signal.

Next, the frame size, that is, the size of the evaluation frame is
This is the size of the image area used for generating the evaluation value.
Also called an evaluation frame, a distance measurement frame, or a specific area. As shown in FIG. 7, the frame sizes used in this embodiment include five types of frame sizes W1 (evaluation frame 1) to W5 (evaluation frame 5).
Each has a different size. Evaluation frame W1 to evaluation frame W5
Center corresponds to the center of the screen.

Evaluation frame size (horizontal × vertical unit: pixels) Remarks W1 116 × 60 Basic frame size W2 96 × 60 W3 232 × 120 W4 192 × 120 W5 576 × 180 However, the screen size of one field is 768 pixels × 24
0 pixels.

Of these frame sizes, the AF developed this time
In software, AF processing is performed using W1, which is the smallest possible size, in order to faithfully focus on a target object. W1 is about 1 / screen size
It is about 6.6, and this frame size W1 is referred to as a basic frame size.

As described later, in a specific evaluation value using W1 or W2, basically, W1 or W2
Is set, but under certain conditions, the frame size W
1, W2 may be enlarged and changed to the size of the frame size W3, W4 or the frame size W5 in some cases. That is, "false mountain"
(A phenomenon in which the evaluation value increases when the focus advances in the blurring direction.) Occurs, and J remains in the minimum frame W1 or W2.
In some cases, it is not possible to keep up with P. In such a case, the evaluation value is used to enlarge and change the frame size originally used.
The AF process is performed again. Details will be described later.

As described above, by setting a plurality of types of frame sizes, different evaluation values corresponding to each frame size can be generated. Therefore, an appropriate evaluation value can be obtained from any one of the evaluation values ID0 to ID13 regardless of the size of the target subject.

The evaluation value calculation method includes HPeak, HIntg, VIn
There are tg and Satul systems. The HPeak method indicates the horizontal evaluation value calculation method of the peak method, the HIntg method indicates the horizontal evaluation value calculation method of the total integration method, the VIntg method indicates the vertical evaluation value calculation method of the integration method, and the Satul method indicates the number of saturation luminance. . Each will be described.

In the HPeak method, a horizontal image signal is converted to an H signal.
This is an evaluation value calculation method for obtaining a high frequency component using a PF.
Evaluation values ID0, ID1, ID2, ID3, ID9, ID
10 and ID11. Therefore, these evaluation value generation circuits in FIG. 2 have a horizontal direction evaluation value calculation filter described below. These evaluation values differ in the cutoff frequency and frame size of the HPF used for generating the evaluation value.

FIG. 3A shows a circuit configuration of a horizontal direction evaluation value calculation filter used in the HPeak system.
The horizontal direction evaluation value calculation filter includes an HPF 202 that extracts only a high frequency component from the luminance data Y of the luminance signal generation circuit (reference numeral 201 in FIG. 2), an absolute value processing circuit 203 that obtains an absolute value of the high frequency component, and an absolute value conversion circuit. A multiplication circuit 204 for multiplying the high frequency component by a horizontal frame control signal; a line peak hold circuit 205 for holding one peak value per line; and a vertical integration of each peak value for all lines in the evaluation frame. And a vertical direction integration circuit 206.

High-frequency components are extracted from the luminance data Y by the HPF 202, converted into absolute values by the absolute value processing circuit 203, and are all positive data. Next, the frame control signal in the horizontal direction is multiplied by the multiplying circuit 204 to obtain an absolute value high-frequency component in the evaluation frame. The multiplication of the frame control signal
By specifying the range to be evaluated in relation to the target subject with respect to the operation as an evaluation frame, and using only the high-frequency components in the frame, the evaluation value based on the video information that enters and exits outside or around the frame is obtained. This is to eliminate noise, sudden changes, and the like. 1 in the frame by the line peak hold circuit 205
.., H, one peak value hp1, hp2,.
pn is held respectively.

One peak value is held for each line, and this value is held in the vertical direction integration circuit 206 in the vertical direction with respect to the line in the evaluation frame based on the vertical frame control signal.
Add from p1 to hpn. A value obtained by summing each peak value for a specific evaluation frame of one screen for a line Σ (hp1 + ... +
hpn). This method is referred to as an HPeak method because the peak in the horizontal direction (H) is temporarily held.

The following contrivance has been made for the frame control signal. FIG. 3 shows a horizontal frame control signal 301 and a vertical frame control signal 30 corresponding to the evaluation frame 303.
2 is shown. Here, the vertical frame control signal 302 is a square wave, but the horizontal frame control signal 301 is not merely a square but gives a characteristic like a roof of a house, and the contribution rate of the 30 pixels at both ends is gradually increased. It is declining. The reason for this is to reduce the influence of the intrusion of the edge outside the frame (high-luminance edge around the evaluation frame) located around the frame as the focus advances and the fluctuation of the evaluation value due to the shaking of the subject. is there. The setting of the frame size and the generation of the frame signal characteristics are configured to be freely performed by the CPU using the AF-IC developed this time.

Here, the difference in the cutoff frequency of the HPF 202 is as shown in FIG.
(Z) = coefficient α in (1−z ⁻¹ ) / (1−αz ⁻¹ )
Is determined by the value of The difference of the coefficient α of the cutoff frequency of each evaluation value is as follows.

[0071] HPF alpha value Remarks IIR0 1/2 V direction using IIR1 1/2 H direction using IIR3 7/8 (= 1-1 / 8) used in the V direction IIR4 7/8 (= 1-1 / 8) Used in H direction

The cutoff frequency of the HPF specified by the coefficient α = 1/2 is relatively high near fsc (subcarrier frequency, which is １ ／ of the sampling frequency), and the sensitivity is relatively low. So it doesn't pick up much noise.
On the other hand, the cutoff frequency of the HPF specified by the coefficient α = 7/8 is relatively low, so that the sensitivity is relatively high, but noise is easily picked up. Further, since the phase delay is much larger than the case where the coefficient α = １ ／, it is easy to be affected outside the frame.

Note that IIR0 and IIR1 with α = 1/2,
Further, the reason why IIR3 and IIR4 of α = 7/8 are separated is due to the difference between the V direction (vertical direction) and the H direction (horizontal direction) in the evaluation value calculation method described later. The reason is that the spatial frequency is relatively low because the H direction is continuous with respect to the line of the image pickup signal, while the V direction is discrete because it is a jump line in field units. Therefore, the used data names are different so that they can be identified even if the same α is used.

The HIntg method is a method for calculating an evaluation value in the horizontal direction obtained by the total integration method. For the HIntg method, the evaluation value ID
6, ID7, ID12 and ID13.
Compared to the HPeak method, the HPeak method obtains one peak value hp1 to hpn per line and adds them in the vertical direction Σ (hp1 +... + Hpn), whereas the HIntg method uses each peak value. All luminance signals y1 to y
n or their high-frequency components h1 to hn and add them vertically (垂直 (y1 +... + yn), Σ (h1 +.
n) is different. The HIntg method is classified into IIR1 in which use data uses a high-frequency component and Y in which the luminance signal Y itself is used as it is.

In the HIntg method, II is used as data to be used.
When R1 is used, the high-frequency components extracted from the luminance signal by the HPF specified by IIR1 are totally integrated. Used for evaluation value ID6. Therefore, the evaluation value ID6 generation circuit in FIG. 2 has the following total integration type horizontal evaluation value calculation filter. FIG. 5 shows a circuit configuration of this total integration type horizontal evaluation value calculation filter. This filter includes an addition circuit 501 that outputs luminance data Y from a luminance signal generation circuit (reference numeral 201 in FIG. 2), an HPF 502 that extracts only high-frequency components from the luminance data Y, an absolute value processing circuit 503, A multiplication circuit 504 for multiplying the component by the horizontal frame control signal; a horizontal addition circuit 505 for adding the absolute value high-frequency component to all the lines in the frame; and all the components in the frame based on the vertical frame control signal. A vertical integration circuit 506 that integrates the absolute value of the absolute value high frequency component of the line in the vertical direction.

Compared with the HPeak horizontal evaluation value calculation filter of FIG. 2, the total integration horizontal evaluation value calculation filter is the same up to the horizontal frame signal multiplication circuit 504, but the horizontal addition circuit 505 The difference is that all the horizontal data after the multiplication are added, and then the vertical integration circuit 506 integrates all the lines in the frame in the vertical direction. Therefore, the HIntg method has a relatively higher evaluation value than the HPeak method and is effective for a low-contrast subject, but has a disadvantage that a false mountain easily occurs. Therefore, in the AF of the present embodiment, the HP Peak method and the HIntg
The system can be used in a timely manner.
The use of the tg method will be described later.

When the luminance data Y is used as the use data in the HIntg method, all the luminance data Y are added in the horizontal direction, and the total is integrated within a set evaluation frame. The evaluation value in this case is also called a luminance addition value. Evaluation value ID
7, used for ID12 and ID13. The luminance addition value is obtained by a luminance addition value calculation filter circuit in which the HPF 502 is removed from the total integration method horizontal direction evaluation value calculation filter in FIG.

Since the high-frequency component is not extracted from the luminance addition value, the original evaluation value using the high-frequency component directly related to the blur phenomenon, that is, the HPeak method, the HIntg method using the IIR, and the Vintg method described below. It is different from the evaluation value of the method. Actually, when the subject does not move into or out of the evaluation frame without the subject shaking or disturbance (noise), the luminance addition value is almost the same regardless of whether the image is in the JP state or the blurred state. . However, when the subject shakes or disturbs to enter the frame due to disturbance or the like, the luminance addition value greatly changes. Therefore, the luminance addition value is effective as an evaluation value mainly used for monitoring disturbance, shaking, and the like by utilizing such properties. Thus, the luminance addition value is J
Since it plays an auxiliary but effective role in driving P, it is treated as an evaluation value in this embodiment.

The VIntg method is a vertical direction evaluation value calculation method of the total integration method. Used for evaluation values ID4 and ID5. Therefore, these evaluation value generation circuits in FIG. 2 have a vertical evaluation value calculation filter described below. HPe
In both the ak method and the HIntg method, an evaluation value is generated by adding in the horizontal direction, whereas the VIntg method is an evaluation value generated by adding a high-frequency component in the vertical direction. For example, if the upper half of the screen is white and the lower half is black, or the image of a horizontal line is present in some scenes, only the high frequency component in the vertical direction is present and no high frequency component in the vertical direction, the HPeak method horizontal direction evaluation value does not function effectively. . Therefore, the evaluation value of the VIntg method is determined so that AF can function effectively even in such a scene.

In general, in order to add high-frequency components in the vertical direction, it is necessary to temporarily store high-frequency component data in a line memory or a frame memory, and to devise the reading order. And it becomes complicated. However, the VIntg method defined here obtains a vertical evaluation value by a simple original configuration that does not require such a circuit configuration.

FIG. 6A is a diagram showing the circuit configuration of the vertical direction evaluation value calculation filter. The vertical direction evaluation value calculation filter includes a vertical direction evaluation value calculation filter 601, a timing generator 602, an IIR HPF 603, an absolute value circuit 604, and an integration circuit 605.

The vertical direction evaluation value calculation filter 601 averages 64 pixels at the center of the evaluation frame of each line of the luminance data Y from the luminance signal generation circuit (201 in FIG. 2) (the same applies to the total value). ) Is calculated and output once per 1H.
Here, the reason why the number of pixels is 64 at the center is to remove noise around the evaluation frame. Here, since only one pixel is sequentially accumulated and finally one total value is output, a simple configuration that does not require a memory device such as a line memory or a frame memory is provided. Next, high frequency components are extracted by the HPF 603 in synchronization with the line frequency. The high frequency component at the center is converted into absolute value high frequency data by an absolute value circuit 604,
The integration circuit 605 performs total integration for all lines in the frame.

The vertical evaluation value is particularly effective for an image having no horizontal change, such as a horizontal line scene. Conversely,
In the case of an image having a change in the horizontal direction, this becomes noise and the validity of the evaluation value decreases.

The Satul method is a calculation method for calculating the number of saturated luminance data Y in a frame (specifically, the luminance level is equal to or more than a predetermined amount). Used for the evaluation value ID8, the evaluation value ID8 is also referred to as a saturation luminance number. Therefore, these ID8 evaluation value generation circuits in FIG. 2 include a saturation luminance number calculation circuit described below. When the saturation luminance number is large, false mountains often occur, and special processing is required as AF for the occurrence of false mountains. It is defined as an evaluation value for detection.

FIG. 6B shows an example of the saturation luminance number calculation circuit. This saturation luminance number calculation circuit calculates the luminance data Y and the threshold value α.
And the luminance data Y is greater than or equal to the threshold α (saturated luminance)
, A comparator 606 that outputs “1”, an AND circuit 607 that inputs an output of the comparator, a horizontal / vertical frame control signal, and a clock signal CLK representing one pixel clock, and a clock input that outputs the output of the AND circuit. And a VD signal representing one vertical period is input to the set terminal, and the counter 6 counts the number of saturation luminance data for one field.
08. Next, each evaluation value will be described individually.

Next, the individual evaluation values will be described. (1) ID0 (IIR1-W1-HPeak) The evaluation value ID0 is the most basic evaluation value in the AF of this embodiment, and is also referred to as a basic evaluation value. A of this embodiment
F executes the process by preferentially judging the increase or decrease of the evaluation value ID0. The evaluation value ID0 is the HPF specified by IIR1.
Is relatively high around fsc (1/4 of the sampling frequency, which is the frequency of the subcarrier), and the probability that the evaluation value peak matches JP is high.

On the other hand, the evaluation value ID3 (IIR4-
W1-HPeak) has a relatively high sensitivity because the cutoff frequency of the HPF specified by IIR4 is relatively low,
The evaluation value peak is likely to be affected by factors outside the evaluation frame.
There is a danger that there will be no overlap with this, and as a result, there will be a mistake in running the JP.

On the other hand, the evaluation value ID0 has a relatively low sensitivity, so that a false mountain is less likely to occur as compared with other evaluation values. And there is a disadvantage that a significant value cannot be obtained.

The circuit configuration for generating this evaluation value is based on the horizontal evaluation value calculation filter shown in FIG. HPF shown in FIG. 3B
Is 1/2. The size of the frame is the basic size evaluation frame W1 (116 × 60), which is about 1 /
It is 6.6. The evaluation frame W1 has a frame size relatively smaller than the evaluation frame used for AF of the consumer video camera. The reason why the size of the frame is smaller than that of a consumer video camera is to focus on a target object. On the other hand, when the frame size is reduced, the evaluation value tends to fluctuate due to camera shake, subject shake, and the like, and AF becomes more difficult. Therefore, the evaluation value ID0
In this example, the evaluation frame W1 is not fixedly used, but is expanded to another evaluation frame having a line segment ratio doubled or tripled under certain conditions in false mountain determination and saturation luminance determination described later. We are elaborate to use it after changing it. Details will be described later.

(2) ID1 (IIR1-W2-HPeak) Compared with the evaluation value ID0 (IIR1-W1-HPeak),
The evaluation value ID1 is generated in an evaluation frame W2 (96 × 60) having a slightly smaller size only in the horizontal direction. ID1
Is that the behavior of the evaluation value curve is almost the same because the other conditions are all the same except that the frame size is slightly smaller than ID0.
It is provided to perform “false mountain judgment” described later by utilizing the difference in the horizontal frame size from IDO.

(3) ID2 (IIR4-W1-HPeak) ID2 (IIR1-W1-HPeak)
Means that the cutoff frequency of the HPF in the generation of the evaluation value is 1
/ 8, which contributes to the generation of the evaluation value up to low frequency components, and thus has a relatively high sensitivity property. Since the frame size is the same basic frame size W1 as ID0, it takes a significant value even in the blurred state compared to ID0,
Since the phase delay of the HPF is much larger than that of ID0, there is also a drawback that the phase is easily affected outside the frame and a false mountain easily occurs.

FIG. 2 shows the tendency of the evaluation value ID0 and the evaluation value ID2. The evaluation value ID0 curve has a sharp shape because only the high-frequency component whose cutoff frequency is relatively high contributes.

(4) ID3 (IIR4-W3-HPeak) Compared with ID2 (IIR4-W1-HPeak), the evaluation value ID3 has a different frame size and an evaluation frame W3 (232 × 232 × Since 120) is used, the number of lines increases by that amount, and the evaluation value becomes relatively high.
"Wobbling direction determination" and "reverse transmission determination" described later
And "checkdown determination".

(5) ID4 (IIR0-W1-VINtg) The evaluation value ID4 is an evaluation value for obtaining a high frequency component in the vertical direction. The feedback coefficient of the IIR HPF is IIR
Α = １ ／, which is the same as 1, but since it is a jump line in a field unit, it is lower than IIR1 in terms of spatial frequency. The frame size is the same basic frame size W1 as ID0.
The evaluation value ID4 is an evaluation value having a correlation with the blurring phenomenon of the subject when the video has no high-frequency component in the horizontal direction and has a high-frequency component in the vertical direction. It has the property of easily occurring.

(6) ID5 (IIR3-W3-VINtg) Evaluation value ID5 is a vertical evaluation value similar to ID4 (IIR0-W1-VINtg). However, this evaluation value was not used at the time of filing the present application. Usage data IIR3 is:
The feedback coefficient α of the IIR HPF is α = 7/8, which is the same as IIR4, but is lower than IIR4 in terms of spatial frequency because it is a jump line in the field unit. The frame size is the same evaluation frame W3 as ID2.

(7) ID6 (IIR1-W1-HIntg) The evaluation value ID6 is ID0 (IIR1-
W1−HPeak), but the sensitivity of the evaluation value is high because all the data after the absolute value processing are fully integrated in both the H and V directions. Therefore, ID6 is effective for dark scenes. However, there is a disadvantage that noise is easy to be picked up due to the high sensitivity of the evaluation value, and false noise is apt to be generated because all noises generated in an extremely dark scene are added. The evaluation value ID6 is mainly prepared for a low-contrast subject.

(8) ID7 (Y-W1-HIntg) The evaluation value ID7 is a value obtained by simply summing the luminance data Y in the frame, and is also called a luminance addition value. The frame size is the basic frame size W1. If the subject is stationary, the brightness addition value does not change much even if the focus changes, so that it is not used for determining the blur state of the subject. However, the luminance addition value has a property of greatly changing when there is a shaking state of a subject or a disturbance enters a frame, and is used for monitoring these.

(9) ID8 (Y-W1-Satul) The evaluation value ID8 is the number of pixels having a luminance higher than a predetermined luminance level set by the CPU. ID8 is also called a saturation luminance number. The predetermined luminance level is set by the CPU, and a value close to a substantially maximum luminance level is set.
The frame size is the basic frame size W1. When the saturation luminance number is large, a false mountain often occurs. In the AF software, the saturation luminance number ID8 is monitored, and when the number exceeds a predetermined number, the process is switched to a special process to shift the focus to JP.

(10) ID9 (IIR1-W3-HPeak) The evaluation value ID9 is different from ID0 (IIR1-W1-HPeak) only in the frame size, and is used as an enlarged value of twice the line segment ratio. The frame size is the evaluation frame W3 (232 × 12
0). It is used for peak determination in “sway determination”, “false mountain determination”, and “insurance mode”.

(11) ID10 (IIR1-W4-HPeak) The evaluation value ID10 is ID9 (IIR1-W3-HPeak).
The evaluation frame W has a slightly smaller size in the horizontal direction compared to
4 (192 × 120). Regarding the frame size, the relationship between ID10 and ID9 is
This is similar to the relationship for 0. Used for peak determination in “false mountain determination” and “insurance mode”.

(12) ID11 (IIR1-W5-HPeak) The evaluation value ID11 is ID0 (IIR1-W1-HPeak).
Compared with, only the frame size is approximately 5 times the width and 3 heights in line segment ratio.
The evaluation value is doubled. Frame size is the largest evaluation frame W
It is 5 sizes. Like the evaluation value ID10, it is used for the peak determination in the "sway determination", the "false mountain determination", and the "insurance mode".

(13) ID12 (Y-W3-HIntg) Evaluation value ID12 is an evaluation value obtained by enlarging only the frame size by a line segment ratio twice in the horizontal and vertical directions as compared with ID7 (Y-W1-HIntg). It is. . The frame size is the evaluation frame W3 (232 × 12
0). This is used to determine the shaking state of the subject and the intrusion of the disturbance into the frame.

(14) ID13 (Y-W5-HIntg) The evaluation value ID13 is larger than the ID7 (Y-W1-HIntg) by expanding only the frame size by approximately 5 times in the line segment ratio and 3 times in the vertical direction. This is the evaluation value obtained. The frame size is the maximum evaluation frame W5. Like the evaluation value ID12, it is used to determine the shaking state of the subject and the intrusion of the disturbance into the frame.

(15) Relationship between evaluation values HPeak (HIntg) using horizontal high-frequency components extracted from luminance data Y according to use data
And Y, which uses all the luminance data Y, and VIntg, which integrates in the vertical direction using all of the high-frequency components in the center of the horizontal direction data.

As for HPeak, as shown in FIG. 9A,
For the basic evaluation value ID0, there are ID9 whose evaluation frame is expanded four times in area ratio and ID11 whose area ratio is expanded almost 15 times, and ID1 whose horizontal direction is slightly narrower than IDO.
And ID10 whose horizontal direction is slightly narrower than ID9. Further, the HPF is used for the basic evaluation value ID0.
ID2 whose cutoff frequency is low (high sensitivity) and I
ID3 having a lower cutoff frequency of the HPF than D9 is prepared. Further, in particular, unlike the basic evaluation value ID0, there is an ID6 for performing the total integration HIntg.

As for Y, as shown in FIG. 9B, the luminance addition value ID7 and ID1 obtained by expanding the evaluation frame to an area ratio of 4 times.
2 and ID13 which has been enlarged approximately 15 times in area ratio.
As for VIntg, as shown in FIG. 9C, ID4 and ID5 having an enlarged evaluation frame and a low cutoff frequency are prepared. In addition, there is ID8.

Next, the operation of the video camera will be described with reference to FIGS. 3. AF algorithm 3.1 Overall flowchart The AF mechanism developed this time has been developed based on the concept of adopting a one-shot type. The specification is "one-shot type AF", which is activated when the AF switch (reference numeral 115 in FIG. 1) is pushed (AF start) and ends the operation when the focus is moved to JP. That is, once the AF operation switch is pressed, the focus becomes JP.
The lens in the state becomes immobile until the AF operation switch is pressed next time. This is to avoid the danger when the focus fluctuates during the broadcast.

In many shootings in an actual broadcasting station studio or various events, a plurality of video cameras are prepared and switched to be broadcast as appropriate. Therefore, the distance between a specific camera and a subject is almost constant, and the one-shot type AF is a mode suitable for such an image. On the other hand, it is not suitable for photographing a subject whose distance to the camera changes continuously, such as when photographing a train running from here to a distant place outdoors.

FIG. 10 is a flowchart of the AF software. This AF software generally consists of a swing determination stage, W
It consists of three stages: an OB stage and a hill climbing stage. In the shake determination stage, the shake of the subject is determined, and a mode is set according to the determination. In the WOB stage, wobbling processing is performed by moving only the wobbling lens, and an initial direction in which the focus lens should advance is determined. The focus lens is not moved until the end of the wobbling process. In the hill-climbing stage, a process of actually moving the focus lens, determining the increase or decrease of the evaluation value accompanying the focus lens, and driving the focus to JP is performed.
In the AF flowchart, normally, the hill-climbing loop is set to one.
The evaluation value peak is detected while making several rounds at one time in the field, and is driven into JP.

In addition to the AF flow, as the data processing for each field, the next data is sequentially updated and recorded in a predetermined loop buffer memory (for 32 fields). The data stored in the loop buffer is used for various processing and determination of AF software. (a) Raw data of 14 evaluation values (b) Moving average of each of the three fields of each evaluation value data of (a) (14 types) (c) Lens position (d) Iris value and focus value ( (Alternately stores one field at a time.)

Hereinafter, the steps of the AF flow of FIG. 10 will be described in order.

3.2 Judgment of shaking (when the lens is stationary) When the start switch attached to the lens is pressed,
The AF mechanism starts operating. In step S1001, a shake determination when the lens is stationary, that is, whether or not there is a shake is determined.
“Shake” is a relative movement between the camera and the subject.
This swing determination has a feature that the evaluation value data from when the AF switch is pressed until a predetermined period elapses does not contribute to the calculation for the swing determination.

FIG. 11 shows that the camera is stopped by panning and
The change of the evaluation value when the switch of F is turned on is shown. In the swing determination, the swing determination period is set so as to eliminate the influence of the pan and the switch ON on the evaluation value. As shown in FIG. 11, after the period 1105 during which the evaluation value fluctuates due to panning, the period from the time when the AF mechanism is switched on (SW ON) 1101 until the lapse of 20 fields, excluding the period of the immediately following 8 fields, fluctuates. Judgment period (between 12 fields of 9 to 20 fields) 11
03, and the state of shaking is determined using the evaluation value data in this period.

The eight fields (1104) immediately after are excluded in order to remove the influence of camera shake at the time of switch push. Thus, the swing determination period 100
By setting 3, an evaluation value that is not affected by the switch ON can be obtained. In addition, "pan" (pan, pan
"Ning)" is a type of camera work, which means taking a picture while moving the camera left and right. Typical breads include follow breads and surveying breads. The image of a car speeding in a car race appears to be stationary, but the background flowing horizontally is the result of following the camera in accordance with the speed of the car.

As a result of the swing determination, if there is a swing, step S
Proceed to 1002 to perform “swing mode setting”. When the swing mode is set, the moving speed of the focus lens and the operation of the wobbling lens are changed as described later. Up to this stage, the wobbling lens and the focus lens are not moved.

Here, the value of each evaluation value is represented by e, and in particular, the value of the evaluation value IDi is represented by e [i]. The evaluation value of the current field is e0, and the evaluation value of the jth field before is ej.
,. Therefore, for example, e2 [0] means a value two fields before the evaluation value ID0. Further, as described below, at each stage of the AF, the value e [i] of each evaluation value is variously determined by a predetermined threshold. The threshold is set to th, and particularly, the threshold of the evaluation value IDi is set to t.
h [i]. If there are a plurality of thresholds, the first threshold is set to th1 [i], and the second threshold is set to th2
[I].

In the present embodiment, each threshold value and other determination criteria will be described with specific numerical values as much as possible, but these numerical values are illustrative and not restrictive. It should be appreciated that the essence of the invention is that a threshold or other criterion is set for an evaluation value having a specific property to determine the focus state at each stage or an event related thereto.

The determination of the swing uses the normalized difference value of the evaluation value ID0 and the normalized difference value of the evaluation value ID7 (brightness added value). The normalized difference value [%] of the evaluation value ID0 is the value e0 [0] of the current field and the value e2 two fields before that.
Using [0], it is defined as 50 × | e0 [0] -e2 [0] | / e0 [0]. Similarly, the normalized difference value [%] of the evaluation value ID7 is calculated by using the value e0 [7] in the current field and the value e2 [7] two fields before that, as follows: 50 × | e0 [7] -e2 [ 7] | / e0 [7].

The normalized difference value means the rate of change of the evaluation value per field. Here, the reason why the evaluation value of the current field is compared with the evaluation value of two fields before is to remove the influence of the evaluation value fluctuation due to the difference between the odd field and the even field, by setting the distance between two fields as odd or even. This is because only one of the fields is used. Also, regardless of percentage,
It is for the same reason that it is multiplied by 50.

Further, in the AF mechanism actually manufactured as a prototype, e0 [0], e2 [0], e0 [7], e2 [7]
Employs a moving average of three fields.
This is because the fluorescent lamp in the room blinks at 50 Hz, while the camera operates at 60 Hz, so that the flicker of the fluorescent lamp is 20 Hz.
Hz to eliminate this effect. Similarly,
The value e of each evaluation value described below is an average value of three fields unless otherwise specified.
Please be aware that there is no influence of flicker of fluorescent light.

A normalized difference value is calculated using the evaluation value ID0 stored in the loop buffer over the twelve fields of the swing determination period, and the maximum value among the normalized difference values is determined as the maximum normalized difference value (hereinafter, referred to as “the maximum normalized difference value”). ndiff e [0] ”), and the determination of the swing is executed.

However, when the value of the evaluation value ID0 is low, “ndiff e [0]” becomes a large value even if the subject does not fluctuate due to noise fluctuation that is constantly present, and exceeds the threshold value for the vibration determination. Sometimes.

Therefore, when the value of the evaluation value ID0 is low, each normalized difference value of 12 fields is calculated using the luminance added value ID7 instead, and the maximum normalized difference value “ndiff e” is calculated from the maximum value. [7] ”, and the swing determination is performed using the created [7]. The reason for not using the maximum normalized difference value “ndiff e [0]” of the evaluation value ID0 in all cases is that
If the evaluation value ID0 is equal to or less than a certain threshold value, “ndiff e
This is because [7] is considered to correspond to the shaking of the subject. The specific criteria of the swing determination are shown below.

The average value of the evaluation value ID0 (the swing determination period 12
If the average value of the field is 200 or more, ndiff e [0] <3% → still mode ndiff e [0] ≧ 3% → swing mode 1 ndiff e [0] ≧ 30% → swing mode 2

If the average value of the evaluation value ID0 is less than 200, ndiff e [7] <7% → still mode ndiff e [7] ≧ 7% → sway mode 1 ndiff e [7] ≧ 12.5% → sway mode 2 I do.

The swing determination procedure and the determination reference value were determined by photographing a large number of subjects. Initially, if possible, we wanted to make a swing determination using the evaluation value ID0,
Is also used, in the case of a scene with low luminance and close to noise (less than 200), “nd” based on ID7
This is because the use of "iff e [7]" provided a preferable result. Here, as a reference value, the average value of the evaluation values ID0 was divided into 200 or more or less, and 3%, 30
The adoption of the thresholds of%, 7% and 12.5% was determined experimentally by photographing a large number of subjects.

When the screen is in a large blur state, the screen becomes nearly uniform, so that it becomes difficult to detect even if the subject is shaken. Therefore, if the AF operation is started from a large blur state, the shake determination will determine the “stationary mode” regardless of the presence or absence of the shake. Therefore, when the evaluation value becomes significant even in the hill-climbing stage in which the focus lens moves, which will be described later, the swing determination (S1014) is performed again.

The “long filter judgment” in step S1003 and the “long filter processing” in S1004 will be described last for the sake of explanation.

3.3 Wobbling Wobbling in step S1005 (hereinafter referred to as "WOB")
Abbreviated. ) Is determined to be possible or not, and if possible, the process proceeds to the WOB operation (S1006); otherwise, the process directly proceeds to the hill climbing stage.

First, the usefulness of WOB will be described.
When AF is requested, a method of immediately moving the focus lens can be considered. In general, a WOB lens has a smaller mass than a focus lens, so that the focus can be moved to a depth of focus in a relatively short time. Of course, the focus can be moved by the focus lens corresponding to the depth of focus, but it requires a longer time than the WOB lens. Further, under lens photographing conditions with a large depth of focus (the iris tends to close, at the telephoto end), the focus lens relatively far from the CCD needs to be moved further.

Here, the "tele end" refers to an enlarged state, a state in which an extremely narrow range is in focus, and FAR
The evaluation value varies extremely from to NEAR. Conversely,
The “wide end” refers to a zoom-up state, and a wide range is in focus at the tele end. Therefore, the evaluation value does not change from Far to Near. WOB becomes effective as the tendency of the tele end increases.

When the focus lens is moved without knowing the direction in which it should proceed, and if it is determined that the evaluation value is decreasing, the focus lens has poor response characteristics of reversing the moving direction, and the focusing time is short. It will take a long time. Proceed in such a blurring direction and then reverse to J
In the operation of moving in the P direction, the probability that occurs when the screen is not looked good and the user moves without knowing the direction in which to move is not exactly determinable, but may be said to be half the probability. Also, in WOB, compared to the method of moving the focus lens, a change in the screen can be performed with almost no intelligibility.
Focus state (just pin, small blur or large blur)
WOB is useful if the direction in which the evaluation value changes, that is, the direction in which the focus lens should travel, can be accurately known.

Therefore, if the focus lens is a front lens feeding type lens and has a separate WOB lens, it is determined in advance whether WOB is possible or not, and then the WOB lens is determined in advance.
The B operation is performed, the direction in which the focus lens advances is searched, and then the focus lens is moved.

At step S1005, a WOB possibility judgment process is performed. In the WOB availability determination process, it is determined whether the WOB is valid (possible), that is, whether the WOB is to be performed. WOB is not always valid. WO
B is invalid when the evaluation value ID0 is equal to or less than a predetermined threshold value (for example, 250), in a swing mode, or when the saturation luminance number ID8 is equal to or more than a predetermined amount. In these cases, WOB is not executed. .

When the evaluation value ID0 is a predetermined threshold (for example, 2
50) In the following cases When the evaluation value IDO is low, WO in the JP or large blur state
Since a change in the evaluation value cannot be obtained by B, the direction determination for determining the moving direction of the focus lens cannot be performed. Therefore, a predetermined threshold is provided, and if it is less than that, WOB is not performed. The threshold value 250 was experimentally determined using various subjects. Here, for example, the use of other evaluation values such as ID3 having a low cutoff frequency and a large frame size was also considered. However, in many cases, the determination of the direction of the focus lens was erroneous due to the occurrence of a false mountain. ID
Only 0 is used as a reference.

In the case of the swing mode When the swing mode (when the lens is stationary) in step 901 is determined to be the swing mode 1 or 2, the evaluation value fluctuation due to the swing is large, and the direction determination is often erroneous even in WOB. Therefore, WOB is not performed.

When the saturation luminance number ID8 is equal to or more than a predetermined amount When the saturation luminance data is equal to or more than a predetermined amount in the evaluation frame,
Even when WOB is performed, the evaluation value tends to remain unchanged. This is because in the case of saturation, the sharpness of the edge does not change even if it is slightly blurred.

When the evaluation value ID0 is equal to or more than the predetermined threshold value, the mode is the stationary mode, and the saturation luminance number ID8 is equal to or less than the predetermined amount (when none of the above conditions applies), WO
The process proceeds to the WOB operation (S1006) to perform B. ~
If any of the above apply, WOB is not valid,
Proceed directly to the hill climbing stage without performing WOB.

In step S1006, a WOB operation is performed.
This is because the AF block 137 and the lens block side CP
This is performed by giving a WOB lens control designation to U114 (see FIG. 1). FIG. 12 shows a flowchart of the WOB operation. When the WOB routine is started, in step S1201, all 14 types of evaluation value (e0 [i], i = 1 to 13) are stored. In step S1202, the WOB lens is moved from the current position to Ne.
After moving toward the ar direction and moving to a position closer to the Near direction by a half-depth of focus, the evaluation value is stabilized.
At 1203, 14 kinds of evaluation values (enear [i], i = 1 to 1)
13) is stored. In step S1204, the WOB lens is moved in the Far direction so as to be shifted from the initial position by a half depth of focus in the Far direction, and the evaluation values are stabilized. Then, in step S1205, 14 types of evaluation values (efar) are set.
[I], i = 0 to 13) are stored.

At step S1206, the WOB lens is returned to the neutral position (initial position), and after the evaluation values are stabilized, at step S1207, 14 types of evaluation values (enuet) are set.
ral [i], i = 0 to 13) is stored. The reason for returning to the initial position again and setting it as the neutral position and sampling 14 kinds of evaluation values is to increase the calculation accuracy performed with efar [i]. Therefore, step S1206 is omitted, and e0 [i] of step S1202 is removed.
May be used instead.

Here, the Far direction and the Near direction will be described. The direction in which the lens moves away from the CCD (ie,
The direction to approach the subject. ) Is called the Near direction. In other words, the direction in which a subject located near (near) moves so as to be JP is called a near direction. The Far direction is a direction opposite to the Near direction. Also, Near
Regarding the end and the far end, the lens position where the subject at the closest possible distance in JP is in the JP state is called the Near end, and the lens position where the subject at infinity is in the JP state is called the Far end.

A time chart of the WOB operation is shown in FIG.
3 is shown. In FIG. 13, wtime is the sum of the moving time of the WOB lens and the phase delay time (one field) of the moving average filter. wtime is different depending on the iris value F. This is because moving the WOB lens by the half depth of focus requires more time as the iris is stopped down.

Either the movement to the Near side in step S1202 or the movement to the Far side in step S1204 may be performed first.

The evaluation value e0 thus stored
[I], efar [i], enear [i], eneutral
The process proceeds to the direction determination process of step S1208 for determining the focus state using [i], (i = 0 to 13). FIG. 12B shows a detailed flow of this direction determination processing.

The flow of the direction determination process shown in FIG.
It consists of a first stage determination and a second stage determination. First, the first-stage determination is performed using the evaluation value of the use data IIR1 having a relatively high cutoff frequency to determine the direction. If the direction is not known, the evaluation value contributing to the direction determination is changed. And the cutoff frequency is comparatively low.
The second stage determination is continued using the evaluation value of the R4 system. The first stage determination is the main direction determination. In the first stage determination, a just pin (JP) determination (S120)
9), Near determination (S1211), Far determination (S1
Each determination process is executed in the order of 213).

In step S1209, the focus is set to JP.
JP determination is made as to whether or not. JP means that the focus is shifted from the JP position within a half depth of focus (half the depth of focus). If the JP is determined as JP in the JP determination, the focus has already been set and there is no need for AF.
In step S1210, there is no moving direction (just),
The mode is set to normal, and the WOB routine ends. If it is determined that it is not JP, Near in S1211
Proceed to determination.

In step S1212, a Near determination is made as to whether the direction in which the evaluation value increases is the Near direction. If Near is determined in the Near determination, the moving direction is set to Near and the mode is set to Normal in step S1212, and the WOB routine ends. If it is determined that it is not Near, the process proceeds to Far determination in step S1213.

In step S1213, a Far determination is made as to whether the direction in which the evaluation value increases is the Far direction. F
If it is determined to be Far in the ar determination, step S1214
Set the movement direction to Far, set the mode to Normal, and
The OB routine ends. If it is determined that it is not Far, the first-stage determination ends, and the process proceeds to the evaluation value weight change processing in step S1215.

Either the Near determination in step S1211 or the Far determination in step S1213 may be performed first.

If the first-stage determination does not determine any of JP, Near, and Far, and the moving direction of the focus lens cannot be determined, the weight of the evaluation value is changed in step S1215, and the second-stage determination is performed. move on. The “weight of the evaluation value” is a weight for each input of the fuzzy filter type synapse processing (see FIG. 19) performed in the Near determination and the Far determination.

In the second stage determination, step S1216
, The Near determination is performed again. If Near is determined in the Near determination, the moving direction is tentatively determined to be Near in step S1217, the mode is set to flat, and the WOB routine ends. If it is determined that it is not Near, the Far determination is not performed and step S128 is performed.
Proceeding to 3, the moving direction is tentatively determined to be Far, the mode is set to flat, and the WOB routine ends. That is, all are set to Far except for the case where it is determined to be Near in the Near judgment. This is because when the camera is actually used, the object is often a few meters or more away, so when it is not clear what direction it is, it is considered that if it goes to Far, it is likely to be the direction approaching JP. is there. As described above, the WOB operation ends, and the process proceeds to JP determination (S1007).

After the first stage determination, the weight of the evaluation value is changed in the second stage determination and the same determination is performed again for the following reason. In the first stage determination, the weight of the evaluation value is actually given by W [] = {10,0,0,0,0,0,0,0,0,0,0,0,0,0} where Each number in parentheses is a weight for each of the evaluation values ID0 to ID13, and only the evaluation value ID0 of the use data IIR1 having a relatively high cutoff frequency contributes to the determination. The reason why the first-stage determination is performed by narrowing down only to the evaluation value ID0 is that if the direction determination is erroneous, the AF operation is hindered, and therefore, erroneous determinations in JP, Near, and Far determinations are reduced as much as possible.

In the first stage judgment, JP, Near or Far
If neither of the above is determined, it is not possible to determine the direction in which the focus lens should proceed, that is, it is determined that the subject is largely blurred. In this case, whether to move the focus to Near or Far cannot be determined for a clear reason. Therefore, in step S1215, the weight of the evaluation value is changed in order to slightly increase the sensitivity, and the process proceeds to the second stage determination. The weight is changed as follows: W [] = {5,10,20,20,0,0,10,0,0,5,10,5,0,0}. This weighting is based on ID2 (IIR4−) of the usage data IIR4 having a relatively low cutoff frequency.
W1-HPeak) and ID3 (II
R4-W3-HPeak).

As described above, in the first-stage determination, the probability of an erroneous determination is low, but the sensitivity of the evaluation value is relatively low, and a significant value cannot be obtained even in a slight blur state. The second-stage determination has good sensitivity, but is likely to cause false mountains and erroneous determination. Here, the evaluation value ID0 of the usage data IIR1 having a relatively high cutoff frequency in the first-stage determination, and the evaluation value ID2 or ID of the usage data IIR4 having a relatively low cutoff frequency.
It is also conceivable to determine the direction by adding 3 to 3. However, even in this case, it was found that the latter false mountain easily occurs and erroneous determination easily occurs. Therefore, in the present embodiment, the first-stage determination is first given priority, and if the direction cannot be determined here, the backup is made in the second-stage determination, so that the WOB erroneous determination probability is kept low while maintaining the WOB erroneous determination low.
Was able to raise the sensitivity.

Now, the specific contents of the first-stage determination and the second-stage determination, namely, the just-pin determination, the Near determination, and the Far determination will be described.

In the just-pin determination, how the evaluation value changes when WOB is performed to determine “JP” will be visually described. When WOB is performed, the evaluation value changes as shown in FIGS. 15 and 16 according to the focus state. FIG. 15A shows a change in the evaluation value when the starting point (0) matches the JP. WOB ((0) →→→
), The change of the evaluation value becomes a W-shape that decreases twice.

FIG. 15B shows a case where the starting point (0) is slightly shifted from JP. Due to WOB ((0) →→→), the evaluation value decreases twice, but becomes unbalanced between the first and second times. The state in FIG. 15B is also JP.

FIG. 15C shows a case where the starting point (0) is shifted near the half depth of focus. According to WOB ((0) →→→), the evaluation value becomes two types, that is, the value increases after the decrease and the value decreases after the increase. In each case, the difference in decline is greater than the difference in rise. The ratio between the difference between the decrease and the difference between the increases is determined. As shown in FIG. 17, in the present embodiment, the decrease (b): the increase (a) =
Adopt 5: 1 or more. This is also an experimental value.

FIG. 16D shows a change in the evaluation value when the starting point is shifted from JP by more than a half depth of focus. WOB
As a result of ((0) →→→), the evaluation value becomes two types, that is, the value increases after the decrease or the value decreases after the increase, and the difference between the decrease and the difference between the increases is substantially the same. The state in FIG. 16D is not determined to be JP, and Nea
Proceed to r determination.

FIG. 16E shows a change in the evaluation value when the starting point is located at a large blur position greatly deviated from JP. In this case, even if WOB ((0) →→→) is performed, there is almost no change in the evaluation value. Even if there is a change, the noise level is equal to or lower than the stationary noise of the evaluation value.

FIG. 16F shows a change in the evaluation value in the case of a subject having little detail in the JP state. When the detail is small, the change of the evaluation value is hardly obtained in WOB ((0) →→→), and the behavior is the same as the large blur in FIG. 16E. However, it is in JP.

Here, the change in the evaluation value is similar between FIG. 16E and FIG. 16F, and it is difficult to distinguish between the two only by the change in the evaluation value. This means that it is difficult to determine from a WOB the large blur state of a subject having details and the JP state of a subject having small details such as a wall that is nearly uniform. This indistinguishability is one factor that makes image processing AF difficult.

In this embodiment, in the case of FIG. 16A to FIG. 16C, the WOB
Do not do. FIG. 16D to FIG. 16F determine that it is not JP, and proceed to Near determination. Actually, in the case of FIG.
However, since it is indistinguishable from FIG. 16E, it must be determined that it is not JP. In order to numerically realize the above JP determination, e stored in the WOB flow
The calculation methods shown in FIGS. 18 and 19 are executed using the values of 0, enear, efar, and eneutral.

The just pin determination in the direction determination in FIG. 18 will be described. Neutral position of WOB lens and Fa
Evaluation value difference (e0) from the position separated by the depth of focus in the r direction
[0] -enear [0]) 1801 and the extended matching degree m1, m2 from the neutral position and the evaluation value difference (eneutral [0] -efar [0]) 1806 between the neutral position and the position separated by a half depth of focus in the Near direction. Is calculated.

As shown in the details of the calculation of each extended matching degree in the lower part of the figure, the calculation of the extended matching degree m1 is based on the evaluation value difference (e
0 [0] −enear [0]), which is m1 on the vertical axis. (1) When the evaluation value difference is from 0 to the first first threshold th1 [0], m1 = 0. (2) When the evaluation value difference is the first threshold th1 [0], m1 = 0
And when the evaluation value difference is the second threshold value th2 [0],
m1 = 100 [%] is taken. (3) When the evaluation value difference is equal to or larger than the first threshold th1 [0], (2)
Under the condition of (4) If the evaluation value difference is 0 or less, the evaluation value is line-symmetric with respect to the vertical axis.

After all, the evaluation value difference and m1 are in a linear function, but have a width near 0, and the subtle difference is that m1 is zero. The reason for this is that a width near zero is provided in order not to contribute to the determination with a small steady noise (ripple component) of the evaluation value.

The calculation of the extended matching degree m2 is similarly performed in accordance with the evaluation value difference (eneutral [0] -efar [0]) on the horizontal axis.

If both m1 and m2 are 100% or less, the switches 1603 and 1610 take the upper route (1804, 1811), and m1 and m2 are the adders 1
At 807, the sum is compared with a constant β at a comparator 1814. In the present embodiment, as described in FIG.
Since the amount of decrease: the amount of increase: 5: 1 or more, β ≧ 80%. If the sum is greater than β, the comparison result is “JP”
, And if less than β, the result is “not JP”.

According to the example of FIG. 15C, for example,
The evaluation value difference (e0 [0] -ene
ar [0]), the evaluation value difference (eneutral [0] -efar [0]) due to the movement from the matching degree m1 = + 90% to
It is assumed that the matching degree m2 = −10% is obtained from When both are added, the result is 80%, which satisfies β ≧ 80%, and the comparison result is “JP”.

One of m1 and m2 is 100
If the value exceeds%, it is necessary to look at the ratio, so both switches 1803 and 1810 take the lower route,
Both m1 and m2 are normalized with the larger absolute value of m1 or m2 (1805, 1812). That is, the larger absolute value of m1 or m2 is defined as 100. Then, the adder 18
The sum is obtained at 07, and a comparison operation is performed at the comparator 1614. Similarly, if the sum is larger than β, the comparison result is “JP”, and if the sum is smaller than β, the result is “non-JP”.

In the case of FIGS. 15A, 15B and 15C, m1 and m2
Both are positive, and a constant β (for example, 0.8) is experimentally determined so that if the evaluation value difference in the WOB is equal to or more than a predetermined value, the determination becomes the JP state. In the case of FIG. 16D, one of m1 and m2 is positive and the other is negative. When the absolute value of both m1 and m2 is 100% or less, it is usually (m1 + m2)
Are canceled out to become a value close to zero, and thus become equal to or less than β, so that the judgment is “not JP”.

However, the absolute value of both m1 and m2 is 1
When (m1 + m2) is positive and | m1 / m2 | is close to 1 (this is a typical small blur state), the judgment result may be JP. is there. So after normalizing with the larger of m1 and m2,
Misjudgment is prevented by summing and comparing.

In the case of FIGS. 13E and 13F, m1 and m2 are both zero or a small value, and the judgment is "no JP state". Although the coefficient β is set to 80%, as shown in FIG. 17, the JP state is recognized until the ratio of the evaluation value decrease (b) to the evaluation value increase (a) becomes 5: 1 as shown in FIG. . This value was obtained by adjusting experimentally using many subjects.

Next, the Near determination in the direction determination will be described. The Near determination has a fuzzy filter type synapse configuration shown in FIG. 19A. In general,
Fuzzy means ambiguous, and synaptic configuration is essentially a junctional configuration that transmits between nerve cells.
It is named this way because it resembles this neural cell junction configuration.

In this configuration, each evaluation value ID [i] (i = 0
With regard to (13) to (13), the degree of coincidence “of the likelihood that the evaluation value increases in the Near direction” with respect to the change (efar [i] −enear [i]) (1501) of the evaluation value ID0 [i] accompanying the WOB is obtained 1502), each evaluation value I
Each D [i] is multiplied by a predetermined weight coefficient W [i] (1503). The sum of these is calculated (150
4) Then, “coefficient α × total weight” (α × ΣW
The comparison with [i]) is made (1505) to determine Near or not. Refer to the detailed diagram of FIG. 18 for the calculation of the degree of extension matching. This processing is equivalent to obtaining the “likelihood that the evaluation value increases in the Near direction” from each evaluation value and performing a weighted majority decision.

The Near determination is performed in step S1211 of the first stage determination and step S1216 of the second stage determination.
Executed in In both cases, the weight of the evaluation in the first stage determination is W [] = {10,0,0,0,0,0,0,0,0,0,0,0,0,0} On the other hand, the weight of the evaluation in the second stage determination is W [] = {5,10,20,20,0,0,10,0,0,5,10,5,0,0} Differs in that

Next, a description will be given of the Far determination in the direction determination. As shown in FIG. 19B, the Far determination determines that the change of the evaluation value ID [i] is (eneutral [i] -e
Except for using far [i]), the same determination as the Near determination is performed.

In step S1007, “JP determination” is performed based on the data obtained by the WOB operation. If the first stage of the WOB operation determines JP, the AF operation is terminated without moving the focus lens.

Otherwise, the process proceeds to the hill climbing stage. The first lens movement direction is determined by WOB (S1006). However, if it is determined to be Near or Far, the process proceeds to the hill-climbing stage as a “normal mode” (a mode in which the JP is relatively close) at a focus position close to the JP.
Proceed to 1008). If the second-stage judgment is made without judging any of JP, Near or Far, Ne
The ar or Far is provisionally determined, and the process proceeds to the hill climbing stage as a “flat mode” in a large blur state (a mode in which the JP is so far away that the direction cannot be determined). Therefore, the process proceeds to initialization of hill climbing parameters (S1008). . The difference between the “normal mode” and the “flat mode” will be described later.

3.4 Initialization of Mountain Climbing Parameters In step S1008, before starting AF mountain hill climbing, initialization of each parameter in a program related to mountain hill climbing is performed. The parameters for initialization are shown below.

When reaching the boundary value count end, the count for the waiting time until the inversion command is issued is reset to zero. End arrival flag Resets the flag indicating that the end has been reached. This flag is used to determine the arrival at both ends. Maximum value of evaluation value Set the maximum value of each evaluation value to 0. This is used in the insurance mode peak selection described later. Minimum value of evaluation value Set the minimum value of each evaluation value to the evaluation value of the current field. This is used in the insurance mode peak selection described later. Maximum value position of evaluation value The maximum value position of each evaluation value is set to the lens position of the current field. This is used in the insurance mode peak selection described later. Maximum acute angle ratio The maximum acute angle ratio of each evaluation value is reset to zero. This is used in the insurance mode peak selection described later. Acute angle rate maximum value position The acute angle rate maximum value position of each evaluation value is set to the lens position of the current field. This is used in the insurance mode peak selection described later.

3.5 Maximum / Minimum Value Update Processing In step S1009, when hill climbing is started, the maximum value and minimum value of the evaluation value are updated for each field (see FIG. 14). Here, the updating process of the maximum value and the minimum value of the evaluation value means that if the maximum value of the evaluation value in the field is less than the previous maximum value, the maximum value is not updated. Is not updated if exceeds the previous minimum value.

Each evaluation value (14 types of ID0 to ID13)
The maximum value max.e [i] The minimum value of each evaluation value (14 types) min.e [i] The lens position where each evaluation value (14 types) becomes the maximum These values will be described later. Used for peak selection in insurance mode.

3.6 Setting of Lens Speed In step S1010, the lens speed is set. It controls the lens speed from when the focus lens starts moving until the peak position is detected. FIG. 20 shows a flow for determining the lens speed.

In the AF of this embodiment, first, in step S1
At 001 and 1002, the swing is determined and the mode is set. Thereafter, the lens speed at the time of AF is controlled for each field in accordance with the presence or absence of the shake as a result of the shake mode set and determined at steps S1014 and 1015. I have.
Further, it is determined by the WOB operation (S1007) whether the mode is the normal mode determined in the first stage determination or the flat mode performed up to the second stage determination. The lens speed is controlled using this result.

It is determined whether or not the subject has shaken. If there is a shake, the lens is operated at a high speed, and the low-speed switching is not performed. When there is no shaking, the speed is set to be high until approaching the peak position, and the speed is switched to be low when approaching the peak position. The lens speed has two stages, low speed and high speed. This is because it is considered that not providing an intermediate speed between the high speed and the low speed is efficient for the AF convergence time and is a good measure.

The low speed is 2Fs / field, that is, the speed at which the field advances by twice the depth of focus Fs per field. High speed is 12 Fs / field or the slowest of the maximum speeds of the lens. Here, the maximum speed is 12Fs / fiel
The reason for limiting with d is that if the speed is further increased, the collapse of the evaluation value curve becomes conspicuous, and the determination of speed switching from high speed to low speed will not be successful. Also, if the speed switching does not work well or does not work (explained later, the speed switching is not performed in the case of shaking), if the speed is set to a speed higher than this, the amount of overshoot after passing the peak will be reduced. We use this restriction, partly because it grows. 2Fs / field, 12Fs / fiel
The value of d is an experimentally determined value.

The lens speed determination flow of FIG. 20 will be described. In step S2001, the swing is determined. Initially, the shake determination and the shake mode setting when the lens is stationary (S1001,
As a result of S1002), from the next time, as a result of the shake determination and the shake mode setting (S1014, S1015) at the time of the lens operation, the shake mode becomes 1 or 2, and if there is a shake, the process proceeds to step S2002. Then, the process proceeds to step S2003. Here, since the swing determination and the swing mode setting are performed in advance when the lens is stationary, the lens speed can be set before the AF operation starts.

In step S2002 (when there is shaking), the lens is operated at high speed, and low-speed switching is not performed. The reason is that the evaluation value fluctuates even if the lens is stationary when there is shaking, so when AF is started, the evaluation value shows the contribution due to shaking and the evaluation value change due to the focus changing by moving the lens. , And are combined.

In the AF, since the focus is moved to search for the peak of the evaluation value, a malfunction such as "AF converges at a blurred focus position" occurs if the contribution of the above-mentioned fluctuation is large. In the current AF system configuration that calculates and uses an evaluation value from camera image processing, in order to avoid the above-described malfunction, the lens speed is increased,
By moving the lens and increasing the contribution of the change in the evaluation value due to the change in focus, the contribution due to the shaking is relatively reduced, and an evaluation value characteristic that does not affect the shaking is obtained.

Therefore, in the case of shaking, low-speed switching is not performed until the peak position is detected.
When low speed switching is performed, the above
Is increased, and the probability of malfunction due to shaking increases. In such a case, since the lens speed does not decrease even when the camera comes close to the just pin, erroneous determination due to the shake can be reduced.
As a result, it is possible to reduce the erroneous operation that the AF ends in the blurred state.

In the case where there is no shaking, in step S2003, division is made according to the elapsed time t1 from the start of the operation of the focus lens. If the elapsed time has passed the initial low-speed period t1, the process proceeds to step S2004. If the elapsed time is within t1, the process proceeds to step S2005 to switch to low speed and move the lens. The reason why the initial low speed period t1 is provided is that when the direction is incorrectly determined in the WOB operation, the operation proceeds at a high speed from the beginning of the operation and does not go too far. Here, the initial low speed period t1 is as follows.

T1 = 12 fields (for flat mode) = 36 fields (for normal mode)

In the case of the flat mode, since the direction cannot be determined by the WOB direction determination (Near, Far), it is determined that JP is far at the start of the lens operation. However, the reason why t1 is not set to zero is to take care of a case in which it is erroneously determined that there is a flat screen when it happens to be in JP. Even if there is such an erroneous determination, if t1 is set to 12 fields, even if the lens operation is started from JP, the lens moves at a low speed, so that it does not return to JP after a large shift from JP.

In the case of the normal mode, since the direction is determined by WOB, it is determined that the JP position is close at the start of the lens operation. In this case, it is better to converge to JP at a low speed while avoiding the peak from going too far, so that the image looks better. Therefore, when the focus is in the vicinity of the JP or in a small blur state, the focus advances to the JP at a low speed.
A smooth focus drive is obtained. In addition, the case where the focus moving speed is too fast and the peak goes too far is reduced. The above value 12, 36 fields are numerical values determined experimentally.

In step S2005, low-speed switching determination is performed. In this AF, after starting as described above,
Except for the initial low-speed period, it is advisable to advance the lens at a high speed near the peak position, and switch to a low-speed when the lens approaches the peak position. In other words, the lens is moved at high speed from the large blur state to the vicinity of the peak position, and when it is near the peak position, the lens is suddenly decelerated to a low speed, and converges to the peak position smoothly and without going too far. .

Therefore, the timing of switching the speed of the lens from the high speed to the low speed becomes a problem. In this AF, the following discriminants (1) and (2) determine whether or not to switch to low speed.
If Expression (3) is satisfied, the speed is switched to low speed;

[0198]

(Equation 1)

[0199]

(Equation 2)

The first term of the equation (1) includes two evaluation values ID0 (IIR1-W1-HPeak) and ID in the current field.
2 (IIR4-W1-HP Peak) and the ratio of the difference to the sum (e0
[0] -e0 [2]) / (e0 [0] + e0 [2]) is quadrupled. The second term indicates the amount obtained by adding the ratio of the difference between these two evaluation values and the sum for the past four fields. This equation is for detecting the rising point of the evaluation value ID [0]. The difference-to-sum ratio (ie, ID
[0] normalized by ID [2]. ) Is an index of how close ID [0] is to ID [2] having a low cutoff frequency (that is, good sensitivity). Since the ratio is taken, even if the illuminance increases, the value e [i] of each evaluation value is multiplied by a predetermined value, so that the ratio of the difference between the evaluation values and the sum is determined independently of the illuminance. Equation (1) is 4
As can be seen by dividing by the field, the difference obtained by subtracting the index of the moving average of the past four fields from the index of the current field, that is, whether or not the present is closer is shown.

Equation (2) is defined for noise removal,
This is the value of rf (1) one field before. rf (0),
The switching timing is determined based on the two indices of rf (1) based on the predetermined threshold value of Expression (3).

[0202]

If rf (0) <− 25 and rf (1) <− 10, switch to low speed. However, ej [0] is the value of ID0 before j field, ej [2] is j ID2 value before the field

When the lens is moved from the blurred position as shown in FIG. 21, rf (0), rf
(1) Both values are larger than 0. When approaching the peak position, rf (0) and rf (1) suddenly show values of 0 or less.
This discriminant uses such a property to determine that it has approached the peak position. Although this discriminant was obtained by repeating experiments and performing data analysis, various trials and errors were repeated until this discriminant was determined.

This discriminant merely selects the best characteristic among the trial and error, and is not perfect. However, for a subject having a large evaluation value at the peak position, that is, a subject having large details, this discriminant works effectively. On the other hand, the evaluation value ID0 (IIR1-W
1-H Peak) is small (for example, several hundreds or less)
In the case of a subject, the above discriminant may not be satisfied even near the peak position. However, there is a salvation that a peak position having little detail is hardly noticeable even if the peak position is too large. Therefore, regarding the determination of the threshold value, it is determined as a first target that braking is properly performed from a high speed to a low speed on a subject having large details.

In AF, detection of a false mountain is one point. Therefore, in the present AF, the following false evaluation is performed by paying attention to characteristics of evaluation values such as that the occurrence of a false mountain is caused by a high luminance edge around the evaluation frame and that the saturation luminance number ID8 is increased due to the high luminance edge. False mountains are detected by the mountain judgment (S1011) and the saturation luminance judgment (S1012), and the frame size change processing (S1013) is performed.

3.7 False Mountain Determination In step S1011, false mountain determination is performed. In the present AF, a false mountain determination method uniquely devised is performed, and the malfunction of the AF caused by the generation of the false mountain is reduced.

First, a false mountain will be described. The false mountain is defined as a phenomenon in which the evaluation value decreases when the focus approaches the peak position with respect to the subject within the AF frame.

FIG. 22C shows an evaluation value curve in which the ordinate represents the evaluation value and the abscissa represents the lens position (focus). The AF operates in the direction of increasing the evaluation value → (hill climbing) to find the JP. However, in reality, the evaluation value curve sometimes shows the characteristic as shown in the figure. Such a rising portion of the evaluation value irrelevant to JP is called a false mountain. In this case, the AF advances the focus in the direction in which the evaluation value increases, so that the lens sometimes moves in the direction opposite to the direction of JP → blurring, and stops at a largely blurred position. In this way, when a false mountain occurs, it causes a malfunction (hereinafter, referred to as a “false mountain trap”) such that the AF converges at the blurred position or the focus advances in the blurred direction. The existence of false mountains makes the development of the AF function difficult.

[0209] The false mountain generation mechanism is that when the luminance is saturated, limited lines become saturated luminance in the JP state, and the energy is wasted. When this becomes blurred, the wasted energy spreads and effectively affects the peripheral lines, resulting in a brightness state of a certain level or more. This is attributable to the characteristic of the CCD that a portion extremely high in energy is saturated. This is because a phenomenon in which the luminance increases as the blurring phenomenon progresses occurs.

The state of occurrence on the false mountain screen will be described with reference to FIG. As shown in FIG. 22A, a scene is assumed in which a target subject (target) 2201 is inside an AF frame (evaluation frame 1), and a high-luminance edge 2202 having more detail than the subject is outside the evaluation frame. Target 22
01, assuming that both high-luminance edges 2202 are at the same distance, the focus is
In the case of, JP is also obtained for a high luminance edge.

FIG. 22A shows the JP state. This J
FIG. 22B shows how the focus is shifted from the P state. The subject gradually expands as the blur progresses, and similarly, the edge of the high luminance difference swells like a ripple so that the ripple spreads, and then penetrates into the evaluation frame (actually, on the opposite side around the center). There is a corresponding point of, and swells to stop, but the other side is omitted because it has nothing to do with the evaluation frame.) FIG. 22C shows a change in the evaluation value at that time.

FIGS. 22B and 22C show the state of JP. When the focus is moved from this JP state, both the target 2201 and the edge 2202 are blurred and the boundary with the surroundings disappears, and the ripples become larger so as to spread. The evaluation value decreases from the JP state to a certain blur state (range from FIG. 22C). Then, the ripple of the high-luminance edge 2202 spreads and covers the evaluation frame (the state of FIG. 22B). When the focus is further moved, ripples of high-luminance edges begin to enter the evaluation frame (FIG. 22).
B state).

The high brightness edge 2202 is the target 2
Since the detail is larger than 201, the line of the portion where the high-luminance edge is within the frame is the subject 22
The ripples of the high-luminance edge 2202 than 01 greatly contribute to the evaluation value. As the penetration of the edge progresses, the penetration range in the V direction also increases, and the evaluation value increases (→ in FIG. 22C).
Range). In other words, a range in which the evaluation value increases as the focus advances in the blurred direction occurs. Is a ripple (shadow) of the high-luminance edge, not the high-luminance edge itself. Even if the lens is moved and followed in the direction of the higher evaluation value, the focus cannot be achieved.

Next, a method for judging a false mountain and a method for avoiding a false mountain trap, which are performed numerically in the AF of this embodiment, will be described. The false mountain judgment is based on the evaluation value ID0 (IIR1-W1-H
Peak) and the ratio ID0 / ID1 of the evaluation value ID1 (IIR1-W2-HPeak) are monitored, and judgment is performed using this ratio.

It is determined that a false mountain has occurred when hill-climbing has started, the ratio ID0 / ID1 is 1.6 or more, and the ratio is continuous for 10 fields. ej [0] / ej [1] ≧ 1.6 (j = 1 to 10)

Evaluation value ID0 (IIR1-W1-HPeak)
And the evaluation value ID1 (IIR1-W2-HPeak) are shown in FIG.
As shown in A, the other characteristics are the same except that the evaluation frame W1 of ID0 is slightly wider in the H direction than the evaluation frame W2 of ID1. If only the details in the center of the evaluation frame W1 in the H direction (that is, the overlapping portion of the evaluation frames W1 and W2) contribute to the evaluation value ID0, the evaluation value ID0 and the evaluation value ID1 become the same value. . Actually, in the case of a normal scene or a subject where no false mountain occurs, as shown in FIG. 23B, substantially the same value is shown from large blur to JP. Conversely, when the evaluation value ID0 and the evaluation value ID1 are different and the evaluation value ID0 ≧ the evaluation value ID1, the portion contributing to the evaluation value ID0 is H
It means that it is in the peripheral part of the direction (the part of the difference between the evaluation frames W2 and W1). This false mountain determination method uses the frame size W1
(Evaluation value ID0) and two evaluation frames having slightly different sizes of the frame size W2 (Evaluation value ID1) are used.
It is called "double frame method".

For example, if the ratio of the evaluation value ID0 ≧ the evaluation value ID1 becomes equal to or more than a certain value when the camera is moved, there is a very high possibility that a false mountain has occurred. The above ratio ID0 / I
When D1 greatly exceeds 1, it means that the maximum detail in the evaluation frame W1 is concentrated on the right end or the left end of the evaluation frame W1 as shown in FIG. 22B. An example of this is a state where the above-described high-luminance edge 2202 is located at the end of the evaluation frame W1. In this false mountain determination method, it is determined that a false mountain occurs when the maximum detail in the frame is concentrated at the end of the evaluation frame W1.

Here, the reason that “hill climbing starts and 10 fields are continued” is as follows. In the case of shaking, even in a scene where a false mountain does not occur, when the edge of the subject passes through the end of the evaluation frame W2, the above ratio may change greatly and exceed 1.6. In addition to shaking, it is also to remove intermittent things such as neon light. For scenes where false mountains occur,
It greatly exceeds the above ratio 1.6 over a long field. From the above, if the ratio exceeds a predetermined ratio over a predetermined plurality of fields, it is determined that a false mountain has occurred.

Next, a method of avoiding a false mountain trap when it is determined that a false mountain has occurred will be described. If it is determined that a false mountain has occurred, the process proceeds to change the frame size in step S1013.

[0220] In step S1013, a frame size changing process is performed. That is, the focus is temporarily stopped, and the evaluation frames W1 and W2 of the evaluation values ID0 and ID1 are enlarged and changed from their basic sizes to the sizes of the evaluation frames W3 and W4, respectively. This is due to the operation of making the evaluation values ID0 and ID1 the same as the evaluation values ID9 and ID10 in the AF CPU (reference numeral 203 in FIG. 2). The reason that the evaluation values ID0 and ID1 themselves were not replaced with the evaluation values ID9 and ID10, respectively, is that the evaluation value ID0 is a basic evaluation value and is used thereafter, so the frame size is changed without changing the evaluation value ID0. I have to do that. Then, WOB (S1005-100
7) and mountain climbing (S1008 to 1021) are started, and a false mountain determination is performed again in the evaluation frame of the enlarged size. In the above description of the evaluation frame, the evaluation frame W2 slightly narrower in the horizontal direction with respect to the evaluation frame W1 and the evaluation frame W4 slightly narrower in the horizontal direction with respect to the evaluation frame W3 are similarly prepared. That's why.

Note that the evaluation frames W1 for IDO and ID1 have already been set.
In a state where the size of W2 has been changed to the size of the evaluation frames W3 and W4, a false mountain determination is performed, and if it is determined that a false mountain has occurred,
Only the evaluation frame W3 of the evaluation value ID0 is enlarged and changed to the size of the evaluation frame W5 (maximum frame size) (similarly, the evaluation value ID0 is made the same as the evaluation value ID11 in the CPU). By increasing the size of the evaluation frame, a high-luminance edge around the frame is moved to a position near the center of the frame. Even if an edge has a high luminance, if the edge is positioned near the center of the frame, the edge is not a ripple (virtual image) of the edge but a real image, and the evaluation value increases in a direction approaching JP so that the edge is in focus. .

As shown in FIG. 23A, in the “double evaluation frame method”, two evaluation values ID0 (evaluation frame W1) and ID1 (evaluation frame W2) obtained by changing the frame size in the horizontal direction are compared. However, in the vertical direction, the frame size remains constant, and the size is not changed for comparison.
This is because, as shown in FIG. 23C, when the vertical edge 2301 extending in the vertical direction enters from the right or left, many lines are simultaneously affected, and the evaluation value of the H peak method is greatly affected. On the other hand, even if the horizontal edge 2302 enters from above or below as shown in FIG. 23D, only a limited number of entered lines is affected.
As described above, the influence of the disturbance entering in the vertical direction is considered to be relatively small as compared with that in the horizontal direction, and is ignored.

3.8 Saturation Luminance Determination In step S1012, saturation luminance determination is performed. For a normal subject, the evaluation value decreases when the focus is out of focus. However, as described above, the evaluation value of a subject having a saturation luminance may increase as the blur phenomenon progresses. In order to solve this, when the saturation luminance number is equal to or more than a predetermined number, there is a method of moving the focus in a direction to decrease the saturation luminance number. However, since the saturated subject is saturated, even if the subject is slightly blurred, the brightness is not reduced, and the ripple spreads. On the other hand, if the subject is blurred, the subject is blurred and the silhouette is expanded. Thus, the saturation luminance number is shown in FIG.
In general, such a method is not effective because it generally exhibits bimodal characteristics as shown in FIG.

Therefore, in the saturation luminance determination, the presence / absence of an object having the saturation luminance is determined by using the saturation luminance number ID8 (Y-W1-Satul) in the evaluation frame W1. This is because if a pixel having a saturation luminance exists in the evaluation frame, a false mountain easily occurs and a false mountain trap is generated. If the saturated luminance number exceeds a threshold value (for example, 600) for a predetermined period (for example, 5 consecutive fields), it is determined that a false mountain has occurred, and the process proceeds to step S1013, in which the evaluation frame W1 is used in each evaluation value. Is enlarged to the frame size W5. The five consecutive fields are used to avoid a change in frame size due to shaking or momentary noise.

The mechanism by which a false mountain occurs when a saturated luminance pixel is within the evaluation frame will be described with reference to FIG. An area A of the target object (target) 2401 shows the shape of the target having the saturation luminance when the focus is JP. When the focus becomes out of focus, the target starts to blur, and the edge position of the target expands away from the center so that ripples spread from the center of the target. Even if the subject having the saturation luminance is blurred, the original luminance is high, so that the ripples are greatly spread compared to the subject having no saturation luminance (the sharpness of the edge does not decrease even if it becomes large). The area B is in a state where the ripples of the target are spread until they overlap with the evaluation frame 2302, and the area C is in a state where the ripples of the target are spread and the evaluation frame 230
2 shows a state in which it has reached 2.

In the state of the area D, the ripple of the target spreads, so that the ratio (ID0 / ID1) of the evaluation values between the evaluation frames W1 and W2 does not increase so much and does not exceed 1.6. In this case, a false mountain cannot be determined in the false mountain determination even though the false mountain described above has occurred, and a false mountain trap occurs.

FIG. 24B shows a state in which it is determined from the result of the saturation luminance determination that there is saturation luminance, and the size of the evaluation frame W1 is expanded to the evaluation frame W5. In this case, since the entirety of the saturated luminance object 2401 itself falls within the frame, the occurrence of a false mountain can be suppressed. As for the false mountain, a problem arises when only the ripples of the target are within the evaluation frame.

By the above-described false mountain determination (S1011) and saturation luminance determination (S1012), generation of false mountain is suppressed.

3.9 Shaking Judgment (During Lens Operation) In step S1014, a swing judgment during lens operation is performed.
In the shake determination (S1001) performed when the lens is still, the fluctuation of the evaluation value is small even if the lens is shaken in a state of large blur, and the normalized difference value gives a determination result of “no shake”. In this shaking determination, when the lens approaches the movement JP, the evaluation value increases, and the shaking can be determined,
It is determined again whether or not it is shaking. Here, only the luminance addition value is used, and the evaluation value ID7 (luminance addition value) of 6 is used.
If the maximum normalized difference value for the field exceeds 2%, it is determined that there is a swing (sway mode 1).

Ndiff e [7] ≦ 2% → stationary mode ndiff e [7]> 2% → sway mode 1

The reason why the normalized differences for six fields are used is to remove intermittent ones such as neon light.
Judgment of the definition of the normalized difference (when the lens is stationary)
Since it has already been described, the description is omitted here. In the case of the swing mode, the low-speed switching is not performed in the lens speed determination (S1010) in the next hill-climbing loop, and the high-speed is maintained. If it is determined that "there is shaking" in this determination, the process proceeds to the shaking mode setting in step S1015, and then proceeds to the direction determination (S1016). If you determine that there is no shaking,
The process directly proceeds to the direction determination (S1016). Step S10
The setting of the 15 shaking modes is the same as the setting of the shaking mode in step S1002, and a description thereof will be omitted.

3.10 Direction Judgment (Climbing) In step S1016, a direction judgment in hill climbing is performed. On the hill-climbing stage, the direction is determined once in one field. In this direction determination, the evaluation value increases, decreases, or is in a flat state (Up, Dow) based on the increase or decrease of each evaluation value.
n, Flat) is determined, and the result of the determination is stored in the memory. In the direction judgment, individual judgment and
Two judgment calculations are performed. As for the relationship between the two, the weighted majority decision of the individual decision is a comprehensive decision.

The individual judgment is a judgment result of each evaluation value ID. FIG. 25 shows the calculation method. Where e0
[I] represents an evaluation value in the current field, and efs [i] represents an evaluation value one movement unit (depth of focus Fs) before. i is the evaluation value ID #. The predetermined reference value αfe is 0.5 in this embodiment.
And

[0234] Regarding one movement unit, when the depth of focus is deep, the depth of focus may not exceed the depth of focus. Therefore, when the depth of focus is exceeded, it is regarded as one movement unit and the direction is determined. Also, the focus lens moves at a speed of at least 2 Fs / field as described in the setting of the lens speed, so that the focus lens usually exceeds the unit of movement for each field. There is. In such a case, the direction determination is not performed until it exceeds one movement unit. This is because it is meaningless to determine the direction even if the lens has not moved by a predetermined amount.

As shown in FIG. 25, in the UP determination of the individual determination, e0 [i] -efs [i]
Is zero until the first threshold th1 [i], and the second threshold th
If it is 2 [i] or more, it is set to 100%, and the interval is a proportional relationship. The comparison circuit 2503 compares this value with a reference value αfe (for example, 50%).
p) If the latter is large, it is determined that the output is 0 (Not Up). Similarly, in the DOWN determination, efs [i] -e0
It is determined by [i]. If it is neither Up nor Down, it is Flat. UP judgment individual judgment result (Up, D
(own, Flat) is stored in the memory (see FIG. 14), and is used in the reverse transmission determination (S1017) and the CHECK DOWN determination of the hill descent determination (S1022) described later.

In the overall judgment, a fuzzy judgment is made based on a plurality of evaluation values as shown in FIG. The configuration and the processing method of FIG.
It has the same synapse configuration as the ear judgment and Far judgment,
The thresholds th1 [i] and th2 [i] are different. This is a process of weighting and deciding the individual determination results of a plurality of evaluation values. However, at this time, the weight of the evaluation value is W [] = {10,0,0,0,0,0,0,0,0,0,0,0,0,0}, and only the evaluation value ID0 is determined. And the result is the same as the result of the individual determination of the evaluation value ID0. At present, since there is only one input, it is meaningless as fuzzy processing, but this configuration is left as it is for future technical development.

As shown in FIG. 27A, when two consecutive Ups are continued in the overall judgment, the mode is changed from “flat mode”, which means large blur, to “normal mode”, which means that the user has started climbing a mountain and approached JP. .

The difference in handling between the flat mode and the normal mode is that it is not determined that JP has been found in the flat mode in the hill descent determination (S1022) described later. This is in order to avoid a situation in which the evaluation value has noise and Down is detected in a large blur state in which the mountain has not yet started to climb the mountain, and to avoid erroneous determination as the evaluation value peak. On the other hand, in the normal mode, JP is found only when Down is detected after Up is detected.

In the individual judgment, each evaluation value itself is
EVAL TH [i] x γ [i] (threshold x luminance addition correction coefficient)
In the following cases, the flat mode is forcibly set. This is because it is necessary to set so as not to easily react to the stationary noise of the evaluation value caused by the optical noise of the lens, the screen shake caused by the minute vibration, and the like. Further, as described below, the luminance addition correction coefficient γ [i] is used as a variable. This is because when the evaluation value is low, the direction determination is often erroneous due to stationary noise, and the luminance is high.
That is, in a scene with a high ID7 (brightness added value), the entire evaluation value curve rises, and the absolute value of the steady noise level also rises, leading to the Up or Down determination result.
It is based on an empirical rule obtained through experiments that it is necessary to increase this threshold value in correspondence with the luminance addition value. Therefore, EVAL TH [i] × γ [i] (threshold value × luminance addition correction coefficient) is defined as follows.

Here, EVAL TH [i] = {250,250,400,800,0,0,250,0,0,250,250,250,0,0} γ [i] = 1: Proportional distribution when ID7 is 300 or less = 1 to 2 300-1500 = 2: ID7 is 1500 or more, however, ID7 is Y-W1-HIntg.

When determining the noise level in this way,
By correcting the threshold value in accordance with the luminance addition value and raising the level, even if the illumination illuminating the subject becomes brighter, the malfunction of the AF operation caused by a large change in the evaluation value in a state where the focus is largely blurred is reduced. be able to.

3.11 Reverse Transmission Determination In step S1017, reverse transmission determination is performed. If Near or Far can be determined in the first-stage direction determination in WOB, the focus lens can start moving toward JP. However, in the case of the Flat mode in which the WOB proceeds to the second-stage direction determination, the direction in which the motion starts is provisionally determined according to the criteria described in the second-stage determination. In this case, the moving direction is wrong and the probability of starting to move away from the JP is not low.

Further, even in the hill climbing stage, the player does not always go to the just pin, and sometimes moves in the direction away from him. The reason is that the evaluation value is low due to blur or the detail of the subject is low, and if the direction in which the evaluation value increases is unknown, the focus is set to either the Far direction or the Near direction and the change in the evaluation value is determined. I have no choice but to wait. When a false mountain occurs in the evaluation value, if the focus is advanced so that the evaluation value increases, as a result, the focus moves away from the just pin. In general, in an evaluation value obtained by extracting a high-frequency component with an HPF having a high cutoff frequency, reverse running due to the above is likely to occur. Conversely, in an evaluation value obtained by extracting a high-frequency component with an HPF having a low cut-off frequency, the cause is reduced, but reverse running is likely to occur.

Therefore, in this reverse feed determination, if a predetermined decrease in the evaluation value after the lens starts moving exceeds a predetermined reference, it is determined that the vehicle is moving in the reverse direction away from the JP. When all of the following conditions are satisfied after the lens starts moving, it is determined that the vehicle is moving backward from the JP. E0 [3] / e0 [2] ≤ 8 e0 [7] ≤ 1100 e0 [12] and change of e5 [12] must be within 6% Flat mode The individual direction determination result of ID0 is down five consecutive times or the individual direction determination result of ID3 is down five consecutive times.

Here, if it is determined that the reverse transmission is performed, the process proceeds to step S10.
Proceeding to 18, a reverse feed process is performed to reverse the traveling direction of the lens. If it is determined that it is not a reverse transmission, step S1019
To make an end reach determination.

[0246] Each criterion will be described. Figure 2
This is to prevent an infinite loop of reverse reverse transmission as shown in FIG. 7B. Figure 2 shows the evaluation value ID3 (IIR4-W3-HPeak).
Assume a scene with a false mountain as shown in FIG. 7B. Assume that the AF hill-climbing start focus position is A and the focus starts moving toward Far. Evaluation value ID0 (IIR1
-W1-HPeak) hardly changes, but it is assumed that the evaluation value ID3 decreases at the position B and exceeds the reference for the reverse transmission determination.
Then, the focus reverses the moving direction and returns to the position A. It is assumed that the evaluation value ID3 starts decreasing when the position exceeds the position A, and the evaluation value ID3 exceeds the criterion of the reverse transmission determination at the position C. If the direction of movement of the lens is reversed here, the lens moves to position B
And C endlessly go back and forth. In order to avoid this infinite loop, a condition is added to limit the reverse feed reversal to one time.

The condition is that ID3 / ID2 is smaller than a predetermined value. ID3 (IIR4-W3-H
Peak) and the ratio of ID2 (IIR4-W1-HPeak), having the same HPR of IIR4, and the evaluation value ID of the frame W1.
This is for performing the false mountain determination by comparing the evaluation value ID3 of W2 with the evaluation value ID3 of W3. From data analysis of experiments, it has been found that when this ratio exceeds a predetermined value (for example, 8), there is a high-luminance edge around the frame 1 and a false mountain is generated in the evaluation value, which often causes a wrong reverse transmission determination. Therefore, this condition is provided.

Since it is highly probable that the reverse transmission determination is erroneous when the screen is bright, the luminance added value ID7 is a predetermined value (for example,
00) The following conditions are provided.

The condition of the change of the luminance addition value of the frame W3 is shown below. When the change of ID12 (IIR4-W5-HPeak) is large, ID3 (IIR4-W3-HPeak) of
Is likely to be caused by shaking or disturbance intrusion. In this case, it is erroneous to determine that the vehicle is running backwards with ID3.

In the normal mode, reverse transmission rarely occurs, so this condition is added. The flat mode is determined by WOB (S1006) and direction determination (S1016).

This reverse transmission determination is made substantially by using ID3. ID3 (IIR4-W3-HPe
In ak), since the sensitivity is high, a significant increase or decrease in the evaluation value can be obtained with a slight blur. However, false mountains are likely to occur, so if the determination is made as it is, the wrong reverse transmission will be determined. Therefore, the erroneous determination is prevented by adding the additional condition of (1).

3.12 Reverse feed processing In step S1018, reverse feed processing, that is, reverse processing is performed. This is for reversing the moving direction of the lens if it is determined in the reverse feed determination (S1017) that the feed is reverse. At this time, the variables to be initialized at the time of starting mountain climbing, such as the start point position, the start evaluation value, the maximum value of the evaluation value, and the maximum value position, are initialized again (see FIG. 14). When reverse feeding is reversed, the same state as that at the start of hill-climbing is newly set from that position.

The difference is that a flag indicating that the reverse feed reversal has been performed once is set. That is, when the reverse feed reversal is performed while the focus is progressing toward the Far end, the “Far end reaching flag” is set,
When the reverse feed reversal is performed while traveling toward the near end, the “near end reaching flag” is set. These two end arrival flags will be described in the next end arrival judgment (S1019).

3.13 Edge arrival determination In step S1019, edge arrival determination is performed. In the AF hill climbing, the search for the evaluation value peak is continued without changing the direction in which the focus advances until the significant evaluation value peak is found or the direction is reversed by the reverse running determination or the search is stopped. However, the lens has a physical near end and Fa.
There are r ends, and it is not possible to jump over these ends to search, so processing when the ends are reached is required.

When the end is reached, if it is the first end, since the entire area has not been searched yet, the direction of focus advance is reversed and the search is continued. In this case, the hill-climbing parameter is initialized and the parameter indicating the direction is changed in "edge inversion processing" (S1021) described below.

If the focus has already been searched for from the near end to the far end, it means that all the areas to be searched have been searched. In this case, even if JP cannot be detected, further search is meaningless, so that AF convergence must be made somewhere within the searched range. This convergence operation will be referred to as “insurance mode”. In addition,
The whole area search has the following problems. For example, when the scene to be photographed is dark, a significant evaluation value peak may not be easily generated. In this case, AF convergence is performed in the insurance mode after searching for the determined range.
If the entire area is from the end to the Far end, the entire area is searched every time, and it takes too much time to converge.

Therefore, a provisional near end is provided between the physical near end and the far end (these are referred to as “true near ends” and “true far ends”) to reduce the search time. I'm trying. That is, when the focus searches from the temporary Near end to the Far end, the JP within the range is searched.
Is once determined. Since the determination is made in the middle without searching the entire area, the determination is called a halfway determination, the temporary Near end position is called a “halfway determination position”, and the determination method here is called a “halfway determination method”. The midway determination position is set substantially at the center between the true Near end and the true Far end. If the JP exists, AF convergence can be performed in a relatively short time as compared with the case where the entire area is searched.

In the halfway judgment method, J is determined by using an acute angle ratio (described later) at which a local peak (maximum value) can be detected.
It is determined whether P exists. As shown in FIG. 28A, the lens is moved from the lens movement start position, a flag is set at a temporary Near end, then a flag is set at a true Far end, and once the flag is set twice, an intermediate decision is made. Then, it is determined whether JP exists using the acute angle ratio. JP
If it is determined that is present, the focus is moved to that position. As described above, the midway determination position is provided between the Far end and the Near end, and when the focus reaches that position, it is determined whether or not JP exists during the progress of the focus up to that point. , It is determined whether or not to continue the subsequent search. By adopting such an intermediate judgment, a considerable subject such as a subject with little detail and a dark subject which has been focused on JP after performing the whole area search can be converged without performing the whole area search. Become. As a result, the average focusing time of AF can be shortened. Note that the midway determination position is not limited to one. A plurality can be provided as desired.

If it is determined in the midway determination that no JP exists, as shown in FIG. 28B (in edge reversal processing S1021)
Invert and continue the search to the true Near end.
When the focus is advanced to the true Near end, it means that the entire area has been searched, and a position for converging the focus must be determined.

FIG. 29 shows the processing flow of the above-mentioned edge arrival determination. The flow will be described below. Step S2901
Is used to determine whether the current position of the lens is at the end. If the end has not been reached, the process proceeds to the hill descent determination (S102 in FIG. 10).
2) If the end is reached, proceed to the next. Step S2902
Then, set the end reaching flag. That is, if the focus position is at the far end, the far end arrival flag is set. If the focus position is closer to the near end than the predetermined midway determination position, near is set.
Set the r-end arrival flag. In step S2903,
It is determined whether both ends of Far and Near have already passed. If only one of the two end arrival flags is set, it is regarded as one end arrival, the lens is stopped,
The process proceeds to the edge inversion process (S1021 in FIG. 10). The above Fa
If the r and Near end arrival flags are set,
Next.

In step S2904, it is determined whether the current focus position has already reached the true Near end.
When reaching the true Near end (Q in FIG. 28B)
Point), the process proceeds to a peak position selection process (S1020 in FIG. 10), and a peak position is selected according to the rules of the process. When the true near end has not yet been reached (FIG. 28)
(P point of A). In step S2905,
It is determined whether a significant peak exists in the searched range. If there is a peak, a peak position selection process (S in FIG. 10)
Proceed to 1020). If not, go to the next. Step S2
In step 906, for the sake of processing, a true N
substituting the ear end, eliminating the midway judgment there, and
The r-end reaching flag is reset to continue the search operation. In step S2907, it is determined whether or not the current position is at the far end.
0 to S1021). If it is not the far end, the process proceeds to the hill descent determination (S1022). If it is a midway determination position, the traveling direction is left as it is.

3.14 Intermediate Judgment Method The interim judgment method for judging whether or not a JP exists within a range from a tentative Near end to a Far end when searching is performed using "acute angle ratio". . The acute angle ratio will be described below.

Particularly in a dark scene where a high-luminance edge from the periphery of the evaluation frame invades and a false mountain occurs, the evaluation value curve for focus does not become a single-peak shape which is maximum in JP. Not a few. An example is shown in FIG.
0 is shown. FIG. 30 is an evaluation value curve in which the horizontal axis represents time and the vertical axis represents evaluation value. Since the focus is moving the lens at a constant speed, the time on the horizontal axis can be regarded as the focus position. The acute angle ratio is an index of how much the image is sharpened (maximum value) within a short field.
JP can be found in the evaluation value curve including the false mountain as shown in FIG.

The acute angle ratio is calculated for all evaluation values ID [i], (i
= 1 to 13) using the current evaluation value e0 [i], the evaluation value e3 [i] three fields before and the evaluation value e6 [i] six fields before, and is expressed by equation (4). .

[0265]

(Equation 4)

The concept of the acute angle ratio is as shown in FIG.
The ratio of e3 [i] to the difference between e6 [i] and e3 [i] (e3
[I] -e6 [i]) / e3 [i]. However, e3
Since it may increase monotonically beyond [i], both sides are taken and {(e3 [i] -e6 [i]) / e3 [i]} ×
{(E3 [i] -e0 [i]) / e3 [i]}. Note that (e3 [i] -e0 [i]) and (e3 [i]
If both -e6 [i]) take a negative value, a positive value will appear as the acute angle ratio, so (e3 [i]-e0 [i])
If <0 or (e3 [i] -e6 [i]) <0, the acute angle ratio is set to zero.

Here, the difference from the equation (4) actually used is as follows.
The denominator of +1 is to avoid infinity when e3 [i] is zero (preventing zero divide).

The calculation of the acute angle ratio is performed every field, and the maximum value of the acute angle ratio and the lens position of the maximum value are also updated every field (see FIG. 14B). In the above-mentioned halfway judgment, the maximum value of the acute angle rate of each evaluation value recorded so far is checked, and if any of the maximum values of the acute angle rate exceeds the threshold value (for example, 20), JP exists in the searched range. If so, the process proceeds to the peak position selection process in step S1020. If none of the maximum acute angle ratios exceeds 20, JP
Is determined not to exist.

The method of adopting the concept of the acute angle ratio in each field and determining that a peak value is found when the value exceeds a predetermined value is also used in a check-down method for hill descent determination described later. However, the threshold differs between the acute angle ratio used in the midway determination and the check-down method, and the threshold is higher in the check-down method. This is because the halfway judgment uses the acute angle rate over the search area, whereas the checkdown method basically deals with only a few fields (up to 10 fields) of information.
This is to prevent the result that P exists.

3.15 Edge Inversion Processing In step S1021, edge inversion processing is performed. The edge inversion processing performs the following processing. Reverses the direction of focus. Initialize the hill climbing parameters. However, the end arrival flag is not reset.

The initialization of the hill-climbing parameter is the same processing as the “initialization of the hill-climbing parameter” performed immediately before entering the hill-climbing loop. However, if the flag reaching one end is reset, the end inversion is repeated indefinitely. Therefore, only the end arrival flag remains set.

3.16 Peak Selection Peak selection processing is performed in step S1020. That is, when the peak is not detected in the hill descent determination and both ends of Far and Near are reached, or when it is determined that there is a peak in the midway determination, a position where the focus is converged within the searched range is selected. .

In the peak selection, the maximum acute angle ratio is selected from the maximum acute angle ratio of each evaluation value. If it exceeds a predetermined value, the focus position corresponding to the maximum acute angle ratio is determined as the peak position. I do. If not, as a second best measure, for a specific evaluation value, find the difference between the maximum value and the minimum value and the ratio of the maximum value to the minimum value.
The lens position corresponding to the evaluation value is determined as the peak position.

FIG. 31 shows a flow of peak selection. In step S3101, the maximum acute angle ratio and the corresponding lens position are selected from the maximum acute angle ratios for the evaluation values IDO to ID13. If it is determined in step S3102 that the selected maximum acute angle ratio exceeds a predetermined value (for example, 20), the process advances to step S3103. If it is equal to or less than the predetermined value, the process proceeds to step S3104. Step S3103
Then, the lens position corresponding to the maximum acute angle ratio is determined as the peak position.

In step S3104, the case where the maximum acute angle ratio is equal to or less than the predetermined value is used.
Cannot be determined. Therefore, as a last resort, a portion where the difference between the maximum and the minimum of the evaluation value is the maximum is determined as JP. That is, the evaluation values ID0, ID6, I
For each evaluation value of D4, ID9, and ID11, in this priority order, the difference between the maximum value and the minimum value is equal to or more than a predetermined amount, and the ratio of the maximum value to the minimum value is equal to or more than a predetermined amount. Check in order. If there is an evaluation value that satisfies these conditions, the subsequent search is stopped, and the lens position corresponding to the maximum value of the satisfied evaluation value is determined as the peak position. Note that this peak position is stored in the memory. If there is still no evaluation value that satisfies and, as a last resort, the frame size is
The lens position corresponding to the maximum value of the evaluation value ID11, which is 5, is the peak.

3.17 Downhill Decision (Peak Detection) In step S1022, a downhill decision is made. In the AF of the image processing method, the JP can be detected only after passing through the peak of the evaluation value. Therefore, the hill-down determination means the peak detection determination of the evaluation value. Therefore, a method of detecting a hill descent, that is, a decrease in the evaluation value is very important. In the downhill judgment, the difference judgment,
The peak detection determination is performed by two methods, that is, check-down determination.

The difference between the difference judgment and the checkdown judgment is as follows.
It is assumed that the former monitors the difference between the evaluation value of the current field and the evaluation value before the predetermined field, and detects a peak when the difference exceeds a certain threshold. On the other hand, in the latter, it is assumed that the former monitors the ratio between the evaluation value of the current field and the evaluation value before the predetermined field, and detects a peak when this ratio exceeds a certain threshold. In the difference judgment, the contrast is low, and therefore JP
Also, in the case of a subject having a low evaluation value, the change in the evaluation value is small, and JP detection may not be performed in some cases. The check-down determination has a disadvantage that peak detection is delayed for a subject having a high contrast and thus a high evaluation value even in JP, and as a result, overruns exceeding the peak frequently occur. In this AF peak detection, both the difference determination (difference) and the check-down determination (ratio) are used, and the peak is detected without increasing excessively for a subject having a higher contrast by the former, and a subject having a lower contrast is detected by the latter. Has the characteristic that the sensitivity of peak detection can be increased.

In the difference determination, as described below, a hill descent is determined based on a decrease in the evaluation value of one field or two fields, so that the amount of overshoot is smaller than that of the check-down method described below. However, as a practical matter, the evaluation value is also low in JP (for example, 50 for ID0).
(0 or less), the direction determination result is twice consecutive Do.
wn may not be the case. A check-down method is prepared for the purpose of JP detection of a subject having such a low evaluation value. First, a difference determination is performed, and if a peak cannot be detected, a check-down determination is performed.

Difference Determination The difference determination is a basic determination method in the peak detection of the main AF. Evaluation value ID in direction determination (S1016)
0 and ID1 are successively determined as “upn
If “g th” (field) Down continues, it is determined that a peak has been detected, and the flow advances to step S1023 to calculate a peak position. This threshold “upng th” is changed according to the depth of focus. That is, when the depth of focus is long, one down is performed. When the depth of focus is small, according to an empirical rule that a small false mountain easily occurs in the evaluation value curve,
The false mountain trap is prevented by the rule of twice consecutive Down.

[0280] Here, "the depth of focus is short": This refers to a case where the focus amount advanced in one field at the maximum lens speed is larger than the depth of focus. “Long depth of focus”: refers to a case where the focus amount to be advanced in one field at the maximum lens speed is smaller than the depth of focus.

Note that ID7 (luminance added value) described above is used.
Normalized difference value (50 × | e0 [7] -e2 [7] | /
e0 [7]) is a predetermined value (for example, 10
%), The difference judgment result of peak detection judgment is invalidated (reset down count value in the memory) as shaking or disturbance intrusion in order to prevent easy stop due to malfunction due to shaking. ing.

50 × │e0 [7] -e2 [7] │ / e0
[7]) ≧ 10% → Invalid difference determination result

As a result, it is possible to reduce erroneous determination of peak detection caused by shaking or disturbance of the subject. As a result, it is possible to reduce a malfunction in which the AF operation ends with the focus in the blurred state.

[0284] Since the difference judgment uses the individual judgment of the direction judgment described above, Do difference is made regardless of the magnitude of each evaluation value.
The difference threshold for determining wn is the same. It seems that if the evaluation value is low, the threshold value may be small, and if the evaluation value is high, the threshold value may be high. However, a scene with a high evaluation value generally has a high detail, and a slight blur is conspicuous. If the evaluation value is high but the threshold value is also high, the downhill at the top of the JP is overlooked, and as a result, the overshoot becomes conspicuous. On the other hand, in a scene with a low evaluation value, the details are generally low, so that even if the focus slightly exceeds the JP, it is not so noticeable. Therefore,
Scenes with high evaluation values must be determined strictly, while scenes with low evaluation positions need to be determined loosely. As described above, a constant threshold value is adopted regardless of the magnitude of each evaluation value.

Checkdown Determination A checkdown determination method will be described with reference to FIG. 32B. In the checkdown determination method, e0
Peak judgment is performed using three evaluation values of [i], ej [i], and e2j [i]. I in [i] indicates the ID of the evaluation value. e0 [i] is the evaluation value of the current field, ej [i] is the evaluation value of the jth previous field, and e2j [i] is the evaluation value of the 2jth previous field. j is an integer from 2 to 5, starting with 2 and increasing sequentially. Equation (5)
It is shown in (6).

[0286]

Ej

[0]> a * e0 [i] and ej
[I]> a * e2j [i], where j = 2, 3, 4, 5 a = 1.2 (static mode) = 2.5 (sway mode 1)

[0287]

1 / b <ej [7] / e0 [7] <b, and 1 / b <ej [7] / e2j [7] <bb = 1.1

Equation (5) has the same idea as the acute angle ratio described in the halfway judgment method, and the center evaluation value ej [i] of the three evaluation values having different sample times is the other two e0 [i]. ,
If it is greater than e2j [i] by a predetermined factor a or more, (ej [i] /
e0 [i]> a and ej [i] / e2j [i]>
a) It is determined that a peak has been detected. Where j = 2,3
4, 5, ej [i] is sequentially expanded to 2, 3, 4, 5 fields before, and correspondingly, e2j [i] is also 4, 6,
The determination is made by sequentially expanding the field eight and ten fields before. Note that the constant a is reduced when there is no fluctuation to improve the detection sensitivity. If there is a fluctuation, the constant a is increased so as not to erroneously detect the fluctuation of the evaluation value due to the fluctuation.

The equation (6) is equivalent to the evaluation value ID7 (Y-W1-HIn
(tg, luminance added value) using a rate of change, which limits the checkdown determination. That is, when the rate of change of the evaluation value ID7 does not satisfy the expression (6), it is determined that there is fluctuation or disturbance, and the check-down determination of the peak detection is invalidated. Thereby, it is possible to reduce erroneous determination of peak detection caused by shaking of the subject or disturbance. as a result,
It is possible to reduce a malfunction in which the AF operation ends with the focus in a blurred state.

If the peak cannot be detected even in this check-down determination, the JP is set in the insurance mode described earlier.
Converges to

FIG. 32B is a table summarizing the difference judgment, the checkdown judgment, and the operation in the insurance mode. In the chart, the detection delay time indicates the time from when the evaluation value reaches the peak to when the peak is detected. The smaller the detection delay time, the smaller the overshoot amount. From the chart, it can be seen that the detection delay time satisfies the difference determination <checkdown determination <insurance mode. Therefore, the difference determination is desirable as long as the peak can be detected. However, when the evaluation value is low, the peak detection sensitivity of the difference determination is small, so that the other determination is required.

The evaluation values used for the check-down determination are as follows. The order of determination is numerical, and if a peak is detected, the remaining determination is not performed. (1) When there is no shaking ID0 (IIR1-W1-HPeak) ID6 (IIR1-W1-HIntg) ID3 (IIR4-W1-HPeak) ID4 (IIR0-W1-VIntg)

When ID4 is used here,
There are the following restrictions. (a) ID0 <600 (b) ID6 <600 (c) ID3 <1600 This is a horizontal line having no horizontal component and only a vertical component by using the vertical evaluation value ID4 using the VINgt method. It became possible even for scenes like the scene of. However, in the case of ID4, if there is a horizontal component, a pixel shift occurs due to a change in the angle of view caused by moving the focus, an intrusion of an edge due to a change in blur, and the like.
The change in the horizontal direction affects the calculation of the average of 64 pixels in the horizontal direction when calculating ID4, and a false mountain is generated in the evaluation value ID4 in some scenes. To avoid this, H
Evaluation values ID0 and ID using Peak method and HIntg method
According to No. 6 and ID3, when the horizontal direction component is more than a certain value, the vertical direction evaluation value ID4 is used to restrict the peak detection from being performed.

(2) In case of shaking ID6 (IIR1-W1-HIntg)

3.18 Peak Position Calculation (Calculation of Center of Gravity) In step S1023, a peak position is calculated. That is,
If a peak is detected in the downhill determination, the peak position is calculated.

FIG. 33 shows that the focus is advanced at a constant speed,
This shows a state where the evaluation value and the focus position are sampled for each field. If the lens speed is within one focal depth, which advances in one field, select the focus position where the evaluation value is the maximum, and return the focus to that position.
The driving accuracy can be reduced to a half depth of focus.

However, "lens speed setting" (S101)
As described in (0), the image may pass through the JP at a high speed, and generally the focus amount advanced in one field is larger than the depth of focus. In this case, the sample location does not coincide with the evaluation value peak and often jumps. Therefore, it is necessary to calculate an accurate JP position by performing interpolation calculation instead of simply selecting the peak value from the sampled data.

In this AF, the center of gravity calculation formula (7) is used as an interpolation calculation for calculating an accurate JP position. Even if there is some noise, the JP calculation result has little effect,
Unlike the least square method, it is not necessary to determine the shape of f (x), so the center of gravity calculation method was adopted. This is because the evaluation value curve cannot be modeled by a mathematical formula such as the least squares method because the shapes of the subject are innumerable.

[0299]

(Equation 7)

In equation (7), x is a lens position, x1 and x2 are integration ranges, and f (x) is an evaluation value at the lens position x. A method of setting the integration range will be described later. By performing the calculation of the center of gravity, as shown in FIG. 33, the JP position can be calculated even if the sample point does not overlap the evaluation value peak, that is, the JP position.

In order to increase the accuracy of calculating the center of gravity, it is necessary to appropriately select the setting of the integration range. For example, as shown in FIG. 34A, if an integration range is set to imbalance before and after the evaluation value peak, the deviation between the calculation result and JP becomes large.

As shown in FIG. 34B, the integration range is defined as x2 where the lens position at the time of detecting the hill descent is f (2) on the opposite side of the hill in the evaluation value curve formed by the lens position and the evaluation value that have passed. It is optimal to find the lens position x1 corresponding to the evaluation value equal to x2) and use it.

Here, since the sample data is actually discrete (one sample in one field), continuous integration cannot be performed as in equation (7). Therefore, discrete integral calculation is performed as shown in equation (8).

[0304]

(Equation 8)

In this case, regarding the integration range, if the found x1 matches the data stored in the buffer, it is used as it is as the start data of the discrete integration calculation. But usually they do not match. If they do not match,
The lens position xp that is f (xp) ≦ f (x1) and is closest to x1 is retrieved from the data stored in the buffer and used as the integration start position.

In this case, although the integration range is slightly unbalanced, it is not so much affected for the following reason. The reason is that when the moving speed of the lens is high before and after the evaluation value peak, f (xp) is sufficiently smaller than the normal evaluation value peak, and the contribution of the integral calculation is small.
The contribution does not greatly affect the calculation accuracy, and the contribution increases when the lens movement speed is slow before and after the peak of the evaluation value. On the contrary, since the interval between the evaluation value data is narrow, the JP calculation accuracy is increased. Offset as a result,
This is because a decrease in accuracy is offset.

Even in the above-described peak position calculation method, there is no guarantee that the deviation between the calculation result and JP is within the depth of focus in any scene. This is because the shape of the evaluation value curve cannot be modeled, so that the accuracy of the above-mentioned calculation of the center of gravity cannot be calculated. However, the newly developed AF
Under the condition that a 15x lens and a focal length of 120 mm, an iris 1.7 and an extender 2x are inserted,
It has been confirmed that when AF is performed on a normal scene, it is possible to satisfy the required specifications and converge to just focus with sufficient accuracy.

3.19 Peak Position Movement In step S1024, the focus lens is moved to perform peak position movement processing. This is performed by giving a focus lens control command from the AF block 137 to the lens block side CPU 114 (see FIG. 1).
That is, when the peak position is calculated by the peak position calculation, the focus is moved to the peak position by the peak position moving process. In this process, the difference between the focus position at the time of calculation and the focus position to be returned is calculated, and if the difference is greater than or equal to a predetermined distance, the lens speed is calculated so that the distance is returned in 25 fields. This is because when moving away from the peak position, it moves at a relatively high speed, and when it is near, it moves at a relatively slow speed. After calculating the speed, the speed is transmitted to the lens as a speed command. When the lens approaches the target, the position is switched to a position command to accurately reach the target position.

As described above, the reason why the lens is returned to the peak position over 25 fields when the distance is equal to or more than the predetermined distance is that if the speed is returned too high, the screen movement will feel unnaturally jerky. .

Next, a long filter judgment (S1003) and a long filter process (S10) which have not been explained yet.
04) will be described. In the case where a subject having a mirror surface is reflecting light and sunlight while shaking, or when a light emitting body such as a mirror ball is moving violently, each evaluation value,
Brightness addition value fluctuates drastically. In such a scene,
Even if the user is away from the JP, the above-described condition of the hill descent determination (peak detection) is easily satisfied, and the lens stops at a blurred position. In the hill descent determination, the normalized difference (that is, change) of ID7 (luminance added value) is checked, and if the change is large, the peak detection is invalidated, but not all cases can be completely overlooked.

Therefore, this long filter mode is provided for the purpose of performing AF even on a subject whose evaluation values such as a luminance addition value vary greatly.

3.19 Long Filter Determination First, in step S1003, long filter determination is performed. In the AF flowchart shown in FIG. 9, the evaluation value peak is normally detected while going around the hill-climbing loop several times. However, when the vibration is abnormally strong, there is a risk of malfunctioning in a hill climbing loop. In order to avoid such a malfunction, it is determined in advance whether or not the shaking is abnormally severe by performing a long filter determination. If it is determined that the shaking is severe, a long filter process (S1004) is performed to perform hill climbing. The AF operation ends without proceeding to the stage.
Whether to activate the long filter processing or to perform the normal AF processing without activation is determined under the following conditions.

In the long filter determination, it is determined whether or not the shaking is severe.
01) is performed in the same procedure. However, in the shake determination when the lens is stationary, the determination is made using the representative normalized difference value that is the maximum value of the normalized difference values of the evaluation values ID0 and ID7 (brightness added value). For these evaluation values, judgment is made by adding the evaluation values ID9 and ID12 (brightness added value) obtained by expanding the evaluation frame to W3.

If the average value of the evaluation values ID0 and ID9 (the average value of the 12 fields in the shake determination period) is 200 or more, ndiff e [0] ≧ 30% and ndiff e [9] ≧ 30% → shake mode 2 .

If the average value of the evaluation value ID0 is less than 200, ndiff e [7] ≧ 12.5% and ndiff e [12] ≧ 12.5% → swing mode 2

In the long filter determination, if the swing is large (swing mode 2), the long filter processing (S1
004). If the vibration is not large, normal AF
Proceed to processing. This condition was determined experimentally.

3.20 Long Filter Mode In the long filter mode, unlike the hill-climbing stage described above, the lens is moved at a constant speed while taking a long moving average, and the lens is returned to the lens position where the moving average is maximized. The method is adopted. An advantage of this is that the probability of a malfunction is small even for a subject whose evaluation value varies greatly. The disadvantages are that the detection of peaks is relatively slow because a large number of moving averages are taken, and the overshoot of the peaks is large.

FIG. 36 is a flowchart of the long filter processing. Hereinafter, each process will be described. Start long filter processing. In step S3601, a lens speed to be moved at a constant speed corresponding to the depth of focus is set under the following conditions. FS ≦ total stroke / 1000 → v = total stroke / 180 / field FS ≧ total stroke / 200 → v = total stroke / 90 / field all stroke / 1000 ≦ Fs ≦ total stroke / 200 → Proportionally distributed speed in the above range , Fs: Depth of focus Full stroke: Focus lens from physical Far end to
Expressed by the number of pulses of the pulse motor when moved between Near ends. The actual moving distance in the optical axis direction is about tens of mm, depending on the lens of the camera, and this corresponds to 0 to 20,000 pulses in this embodiment.

In step 3603, the moving direction of the lens is set. If it is an initial setting, it follows FIG. In FIG. 35, the horizontal axis represents the lens position FPOS from the far end to the near end.
(Focus Position) is specified by the number of pulses 0 to 20000, and FPOS has a relationship between the rotation angle of the focus ring and a linear function.

Lens position moving direction 0 (Far end) to 5000 → Far end direction 5000 to 10000 → Near end direction 10000 to 15000 → Far end direction 15000 to 20000 → Near end direction

Step S3604, 3605, 3607
Then, the long filter downhill determination is performed. First, S360
In step 4, a moving average (11 steps) of the luminance normalization evaluation value for each field is calculated and its maximum value is updated. When the number of inversions is 1, the luminance normalization evaluation value is half (50) of the maximum value.
%) Or the end, the flow advances to step S3608 to set the number of inversions to 1, and returns to step 3603.
Otherwise, the process proceeds to the lens position calculation in step S3608. .

The luminance normalized evaluation value is the evaluation value ID0
Is multiplied by a moving average of 32 ID7s (additional luminance values) to obtain I
D7 divided by three moving averages.

E [0] * ｛Σ (e0 [7] +... + E31
[7]) / 32｝ ÷ ｛Σ (e0 [7] + ... + e2
[7]) / 3｝

Here, (moving average of 32 luminance added values /
(Moving average of three luminance addition values) is used to reduce the influence of strong reflected light from the mirror.

FIG. 37 shows that the evaluation value ID0 having a large fluctuation (32 moving average of luminance addition value / 3 moving average of luminance addition)
, And taking a moving average of 11 steps, the variation is small, that is, the noise is reduced, and it is easy to find the JP position. By taking 11 steps of moving average, the evaluation value peak position is delayed by 5 fields (strictly speaking,
As described above, since the evaluation value itself is a moving average of three fields, it is delayed by one field, and furthermore, since the CCD stores one field, it is eventually delayed by seven fields, but is omitted because it is not directly related to the present invention. . ).

If the luminance normalized evaluation value reaches half (50%) or the end of the maximum value when the number of inversions is 0 in step S3605, the flow advances to step S3606 to set the number of inversions to 1, and the flow advances to step S3603. Then, the moving direction is set. Otherwise, the process proceeds to step S3607.

In step S3607, the next movement position is specified for the lens, and the flow returns to step S3604.

[0328] In step S3608, the JP position is calculated. If the luminance normalized evaluation value is reduced to half of the maximum value in the long-filter hill-down determination and has already been inverted by that time, it is determined that a JP has been found, and the process enters long filter processing (S1004). As described above, since the luminance normalized evaluation value takes a moving average of 11 steps, the lens position where the peak appears is delayed by 5 fields (actually, it is delayed by 7 fields). Therefore, the lens speed × 7
It is necessary to return the lens position by the amount of the field. In the JP position calculation processing, the JP position is calculated in consideration of these delays.

[0329] In step S3609, the player moves to the JP position. The lens is moved to the position calculated in the JP position calculation processing.

[0330] In step S3610, a procedure for ending the long filter processing is performed, and the current AF processing ends. The AF of this embodiment is a one-shot type, and is in an idling state until the AF switch is pushed next time.

FIG. 38 shows an example of the movement of the lens by the long filter processing (see also FIG. 36). In the figure,
The long filter processing is activated, the moving direction is determined by the moving direction setting processing described above, and the moving starts at a constant speed as shown in FIG. In this case, since the luminance normalized evaluation value decreases, the maximum value and the maximum value position remain at the starting point. The lens advances and the luminance normalized evaluation value becomes half of the maximum value, and since it has not been inverted, it is inverted from that position and restarted.

After the reversal, the lens advances at a constant speed, and further advances through the normalized evaluation value peak as shown in FIG. When the luminance normalized evaluation value becomes half of the maximum value, it has been inverted by this time, so J
Calculate the P position. After that, as shown in (2), the process returns to the JP position, and the long filter processing ends.

The description of this embodiment is completed.

[Effects of Embodiment] The effects of this embodiment will be described below item by item. (1) In the shake determination when the lens is stationary (S1001), the evaluation value data from when the AF switch is pressed until a predetermined period elapses does not contribute to the calculation of the shake determination. That is, by not contributing to the calculation of the swing determination during a predetermined period in which the influence of the pan and the switch ON occurs, the swing determination excluding these effects can be performed, and erroneous determination can be avoided. If this erroneous determination occurs, it is determined that the subject is shaking even if the subject is not shaking, and the lens is driven at a high speed, and there is a risk that the amount of overshoot of the lens after passing through the just pin becomes large. In the present embodiment, it is possible to eliminate the inconvenience of erroneously determining that the subject is shaking even if the subject is not shaking.

(2) In this embodiment, the lens movement speed can be determined in advance by performing the shake determination (S1001) in advance when the lens is stationary before the lens movement starts.
Further, even after the movement of the lens is started, the swing is periodically determined (S101).
By performing 4), it is possible to accurately perform a shake determination in a large blur state that cannot be detected by the shake determination in advance when the lens is stationary.

(3) Setting the lens speed in this embodiment (S101
In the case of 0), when there is shaking, the lens is operated at high speed, moved at high speed, and low-speed switching is not performed. Since the evaluation value fluctuates even if the lens is stationary when there is shaking, when AF is started, the evaluation value includes a contribution due to shaking and a contribution due to a change in evaluation value due to a change in focus caused by moving the lens. Are synthesized. If the contribution from the shake is large, a malfunction such as “AF converges at a blurred focus position” occurs. In order to avoid this malfunction, in this embodiment, by increasing the speed of the lens, the contribution of the fluctuation is relatively reduced, and the contribution of the evaluation value change due to the change of the focus is relatively increased. , The evaluation value characteristics which do not affect the shaking are obtained.

(4) In the AF of the image processing method, JP can be detected only after passing the peak of the evaluation value. Therefore, in the present embodiment, the hill descent determination (peak detection) (S1022) is performed. A method of detecting a mountain descent, that is, a decrease in the evaluation value is very important. In the hill descent determination, the peak detection determination is performed by two methods, a difference determination and a check-down determination. The former monitors the difference between the evaluation value of the current field and the evaluation value before the predetermined field, and detects a peak when the difference exceeds a certain threshold. In this method, the change in the evaluation value is small for a subject having a low contrast, and therefore a low evaluation value even in JP, and JP detection may not be performed in some cases. On the other hand, in the latter, it is assumed that the former monitors the ratio between the evaluation value of the current field and the evaluation value before the predetermined field, and detects a peak when this ratio exceeds a certain threshold. This method
Peak detection is delayed for a subject having a high contrast and thus a high evaluation value even in JP, and as a result, excessive detection (JP
Has a drawback that uranium frequently occurs. In the AF peak detection, both the difference determination and the check-down determination are used, and the sensitivity of the peak detection is increased even for a subject having a low contrast by the latter without excessively increasing the subject having a high contrast by the former. be able to.

(5) In the hill descent determination (S1022), peak detection determination is performed by two methods, a difference determination and a checkdown determination. The former monitors the difference between the evaluation value of the current field and the evaluation value before the predetermined field, and detects a peak when the difference exceeds a certain threshold. In the latter case, it is assumed that the former monitors the ratio between the evaluation value of the current field and the evaluation value of a predetermined field, and detects a peak when the ratio exceeds a certain threshold. However, the evaluation value ID7 described earlier
If the normalized difference value of (brightness added value) exceeds a predetermined value at the time of peak detection, the difference determination result of the peak detection determination as shake or disturbance intrusion is used to prevent the stop due to a malfunction due to the shake. Is disabled (reset the down count value in the memory). Similarly, evaluation value ID7
If the rate of change of (brightness added value) does not satisfy Expression (6), it is determined that there is fluctuation or disturbance, and the check-down determination result of peak detection is invalidated. With these, it is possible to propose an erroneous determination of peak detection caused by shaking or disturbance of the subject. As a result, it is possible to reduce a malfunction in which the AF operation ends with the focus in the blurred state.

(6) In the direction determination (S1016), when each evaluation value itself is equal to or smaller than a predetermined threshold, the Fla is forcibly forced.
t. Thus, it is possible to prevent the stationary noise of the evaluation value from easily reacting. In addition, when the evaluation value is low, the direction determination is often erroneous due to stationary noise. In a scene with a high ID7 (luminance added value), the entire evaluation value curve increases, and the absolute value of the stationary noise level also increases. It rises together and leads to the judgment result of Up or Down. Therefore, the predetermined threshold is set to EVAL TH [i] × γ
[I] (threshold value × luminance addition correction coefficient), and this threshold value is sequentially increased in accordance with the luminance addition value. Therefore, when the illumination for illuminating the subject is further brightened, it is possible to reduce the malfunction of the AF that occurs due to a large fluctuation in the evaluation value when the focus is largely out of focus.

(7) End arrival determination of this embodiment (S1019)
In this document, a provisional near end is provided between the true near end and the true far end to shorten the search time. That is, when the focus is searched from the temporary Near end to the Far end, it is determined whether or not the JP exists within the range. If the JP exists, it means that the AF convergence has been completed in a shorter time than in the case where the entire area is searched. In the midway determination method, it is determined whether or not JP exists using the acute angle ratio at which a local peak can be detected. Thus, Fa
A midway determination position is provided between the r end and the near end, and when the focus reaches that position, it is determined whether or not JP exists during the progress of the focus up to that point. It is determined whether or not to continue the search. By employing the midway determination, a considerable amount of subject such as a subject with little detail or a dark subject that has been focused on the JP after performing the whole area search can be focused without performing the whole area search. Average convergence time can be reduced.

(8) Determination of Saturation Luminance in this Embodiment
By adopting 2), when the saturated luminance number exceeds a predetermined threshold, the frame size of the evaluation frame is forcibly changed from the basic frame size W1 to the maximum frame size W5. By doing so, as shown in FIG. 24, the subject having the saturation luminance is contained in the frame, thereby suppressing the occurrence of the false mountain and preventing the occurrence of the malfunction.

(9) In the WOB (S1006) of the present embodiment, the first stage judgment and the second stage judgment are used in combination. The first-stage determination uses the evaluation value ID0 of the usage data IIR1 having a relatively high cutoff frequency, so that the probability of an erroneous determination is low, but the sensitivity of the evaluation value is relatively low and a significant value is obtained even in a slight blur state. I can't get it. In the second stage determination, since the evaluation value ID2 of the IIR4 of the use data having a relatively low cutoff frequency and ID3 having a further different frame size are used, the sensitivity is good, but false peaks are likely to occur and erroneous determination is likely to occur. Therefore, in this embodiment, the first-stage determination is given priority first, and when the direction cannot be determined here, backup is performed in the second-stage determination, thereby increasing the sensitivity of the WOB while keeping the probability of WOB erroneous determination low. I have.

(10) Setting the lens speed in this embodiment (S101
In (0), except for the initial low speed period, the lens is advanced at a high speed near the peak position, and is switched to a low speed when the lens reaches near the peak position. That is, by moving the lens at high speed from the large blur state to the vicinity of the peak position, and rapidly reducing the speed to a low speed near the peak position, it is possible to quickly and smoothly converge to the peak position without going too far.

(11) In the HP Peak method of this embodiment, high frequency components are extracted from the luminance data Y. In this case, for example, when a color bar is photographed, the change of the evaluation value described later becomes the same as that of a uniform wall in the HP Peak method, and the peak of the evaluation value may not be detected in some cases. On the contrary,
Rather than obtaining luminance data by adding R, G, and B image signals, horizontal direction evaluation value calculation filters are provided independently for the color signals R, G, and B, and each output is added to obtain an evaluation value. As a result, a significant evaluation value can be obtained even for a subject having a different color but a uniform luminance such as a color bar.

[0345]

According to the present invention, it is possible to provide a very accurate AF mechanism suitable for a professional or professional video camera. Further, according to the present invention, it is possible to provide a video camera provided with an extremely accurate AF mechanism suitable for a professional or professional video camera. Further, according to the present invention, it is possible to provide a highly accurate AF method suitable for a professional or professional video camera.

Further, according to the present invention, when determining the moving speed of the focus lens in the AF mechanism suitable for a professional or professional video camera, the focus can be quickly and without exceeding the evaluation value peak in accordance with various conditions. Means for determining the moving speed of the lens can be provided.

Further, according to the present invention, even when the video camera and the subject shake, the rate of change of the evaluation value is relatively reduced so that the change of the evaluation value due to the movement of the focus lens can be grasped. Means can be provided.

Further, according to the present invention, it is possible to provide a shake judging means which takes into account camera shake caused when the AF switch is manually operated when the shake judgment is introduced.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating an overall configuration of an imaging device including a video camera.

FIG. 2 is a diagram showing a configuration of an AF block of the video camera in FIG.

FIG. 3 is a diagram illustrating a circuit configuration of a horizontal direction evaluation value calculation filter. Here, FIG. 3A illustrates a circuit of a horizontal direction evaluation value calculation filter, and FIG. 3B illustrates a difference in cutoff frequency of the HPF.

FIG. 4 is a diagram illustrating a frame control signal used in the evaluation value generation circuit.

FIG. 5 is a diagram illustrating a circuit configuration of a total integration type horizontal evaluation value calculation filter.

FIG. 6A is a diagram illustrating a circuit configuration of a vertical direction evaluation value calculation filter, and FIG. 6B is a diagram illustrating a configuration of a saturation luminance number calculation circuit.

FIG. 7 is a diagram illustrating a frame size of an evaluation frame of each evaluation value.

FIG. 8 is a diagram showing a tendency of evaluation values ID0 and ID2.

FIG. 9 is a diagram illustrating the relationship between evaluation values.

FIG. 10 is a diagram illustrating an entire flowchart of AF executed in the present embodiment.

11 is a diagram illustrating a shake determination period of a shake determination process (when the lens is stationary) in the overall AF flowchart of FIG. 10;

12 is a diagram illustrating a flowchart of a wobbling process in the overall flowchart of the AF in FIG. 10;

13 is a diagram showing a time chart of a wobbling operation in the wobbling process of FIG.

FIG. 14 is a diagram illustrating data used when executing the overall flowchart of AF in FIG. 10;

15 is a diagram showing wobbling and a change in an evaluation value together with FIG. 16, and FIGS. 15A, 15B and 15C show a case where a just focus determination is made.

FIG. 16 is a diagram showing wobbling and a change in an evaluation value together with FIG. 15, and FIGS. 16C, 16D, and 16E show cases where it is determined that the subject is not just in focus.

FIG. 17 is a diagram illustrating a condition when it is determined that the pin is a just pin, with reference to FIG. 15C;

FIG. 18 is a diagram illustrating a determination method for determining a just pin, with reference to FIGS. 15A, 15B, and 15C;

FIG. 19 is a diagram illustrating a method of determining a direction in the wobbling process of FIG. Here, FIG. 19A shows the Near direction determination, and FIG. 19B shows the Far direction determination.

20 is a diagram illustrating a flowchart of a lens speed determination process in the overall AF flowchart of FIG. 10;

21 is a diagram illustrating a discriminant used in the low-speed switching determination in FIG.

FIG. 22 is a diagram for explaining a false mountain generation process in relation to the false mountain determination in the overall flowchart of AF in FIG. 10; Here, FIG. 22A shows a just pin state, and FIG.
2B is a diagram showing a state in which the degree of blur gradually advances, and FIG. 22C is a diagram showing a transition of the evaluation value at this time.

FIG. 23 is an explanatory diagram regarding a swing determination performed using evaluation values ID0 and ID1. Here, FIG. 23A illustrates the double frame method, FIG. 23B illustrates the behavior of the evaluation value of a scene where no false mountain occurs, FIG. 23C illustrates the effect of the vertical edge on the false mountain generation, and FIG. Explains the effect of horizontal edges on false mountain occurrence.

24 is a diagram illustrating a process of generating a false mountain in relation to the saturation luminance determination in the overall AF flowchart of FIG. 10; Here, FIG. 24A describes the occurrence of a false mountain, and FIG. 24B describes the expansion of the evaluation frame when it is determined that the false mountain has occurred.

FIG. 25 is an explanatory diagram regarding individual determination of direction determination in the overall flowchart of AF in FIG. 10;

FIG. 26 is an explanatory diagram relating to comprehensive determination of direction determination in the overall AF flowchart of FIG. 10;

27A is an explanatory diagram of a determination method of direction determination in FIG. 26, and FIG. 27B is an explanatory diagram of a determination method of reverse feed determination in the overall flowchart of AF in FIG.

28 is an explanatory diagram related to edge arrival determination in the overall flowchart of AF in FIG. 10; Here, FIG. 28A shows a case where the midway determination method works effectively, and FIG. 28B shows a case where the entire area is searched.

FIG. 29 is a diagram showing a flowchart of end reach determination in FIG. 28.

FIG. 30A is a diagram illustrating an acute angle ratio, and FIG. 30B is a diagram illustrating a saturation luminance determination.

FIG. 31 is a flowchart of peak position selection in the overall AF flowchart of FIG. 10;

FIG. 32A is a chart showing a comparison between a difference determination, a checkdown determination, and an operation in an insurance mode. FIG.
B is a diagram for explaining the check-down determination of the hill-down determination in the overall AF flowchart of FIG. 10.

FIG. 33 is a diagram showing a state in which the focus is advanced at a constant speed.
FIG. 9 is a diagram illustrating a state where an evaluation value and a focus position are sampled for each field.

FIG. 34 is a diagram illustrating an integration range of a barycenter calculation for peak position calculation in the overall flowchart of AF in FIG. 10; Here, FIG. 34A shows a case where the integration range is not appropriate, and FIG. 34B shows a method of setting the integration range.

FIG. 35 is a diagram illustrating a method of setting an initial lens moving direction for long filter determination in the overall AF flowchart of FIG. 10;

36 is a diagram illustrating a flowchart of a long filter process in the overall flowchart of the AF in FIG. 10;

FIG. 37 is a diagram for explaining a luminance normalized evaluation value in relation to FIG. 36;

FIG. 38 is a diagram illustrating the long filter mode in relation to FIG. 36;

FIG. 39 is a diagram illustrating a depth of field and a depth of focus. Here, FIG. 39A shows the depth of field, and FIG.
B indicates the depth of focus.

FIG. 40 is a diagram illustrating a relationship between an allowable circle of confusion, a distance between CCD pixels, and an allowable circle of confusion.

[Explanation of symbols]

111 focus lens, 111a focus lens position detection sensor, 111b focus lens drive motor, 111c focus lens drive circuit, 112 wobbling lens, 112a wobbling lens position detection sensor, 112b wobbling lens drive motor, 112c wobbling lens drive circuit, 113
Iris mechanism, 113a Iris position detection sensor, 1
13b Iris mechanism drive motor, 113c Iris drive circuit, 114 Lens block side CPU, 115
Auto focus switch, 121 color separation prism, 122R, 122G, 122B imaging device, 123
R, 123G, 123B preamplifier, 124 CCD
Drive circuit, 125 timing signal generation circuit, 131
R, 131G, 131B A / D conversion circuit, 132R,
132G, 132B gain control circuit, 133R, 13
3G, 133B signal processing circuit, 134 encoder,
135 Iris control circuit, 136 White balance control circuit, 137 Auto focus block, 138
Auto focus CPU, 139 Art focus dedicated integrated circuit (AF-IC), 141 main CPU,
142 ROM, 143 RAM, 145 operation unit, 2
01 luminance signal generation circuit, 202 evaluation value generation circuit, 20
3 CPU for auto focus, 204 ROM, 20
5 RAM

Claims (16)

[Claims] 1. An input means for inputting an image signal of a video camera, an evaluation value generating means for extracting a high frequency component of a specific area of the image signal to generate an evaluation value, and the video camera according to the evaluation value An autofocus apparatus comprising: signal processing means for calculating a command value to be given to a focus drive unit of the lens of the above; and output means for sending the command value to a drive unit of the lens of the video camera. A circuit for calculating a value obtained by adding the luminance of the image signal; examining an evaluation value during a predetermined period and a rate of change of the luminance addition value; determining whether or not the video camera and the subject have shake; An auto-focus device, wherein the speed of a focus lens at the time of auto-focus is controlled in accordance with (1). 2. The auto-focusing device according to claim 1, wherein when it is determined that there is further shaking, the speed of the focus lens is set to a relatively high speed. 3. The auto-focusing device according to claim 1, wherein when it is determined that there is no further shaking, the camera moves from the blurred state to the vicinity of the just pin at a first lens speed, and moves from the vicinity to the just focus. Moving at a second lens speed relatively lower than the first lens speed. 4. The auto-focusing device according to claim 3, wherein the switching from the first lens speed to the second lens speed is performed in accordance with the rate of change of the evaluation value. . 5. The autofocus apparatus according to claim 4, wherein the switching from the first lens speed to the second lens speed is performed when the rate of change of the evaluation value becomes relatively large. There is an autofocus device. 6. The auto-focusing apparatus according to claim 3, wherein the switching from the first lens speed to the second lens speed is performed at least as the first evaluation value.
And a second evaluation value that is more sensitive than the first evaluation value, and the ratio of the sum of the second evaluation value difference and the sum of the first evaluation value to the first evaluation value over the past predetermined period is obtained. An autofocus device that switches when the moving average of the ratio of the difference between the evaluation value and the second evaluation value and the sum approaches the moving average. 7. The auto-focusing device according to claim 6, wherein the ratio of the sum of the difference between the first evaluation value and the second evaluation value is equal to the first evaluation value and the second evaluation value over a predetermined period in the past. An autofocus apparatus in which the moving average of the ratio of the sum of the differences between the evaluation values of two is compared at least a plurality of times before the current field and a predetermined field before, and switching is performed based on these results. 8. The autofocus apparatus according to claim 3, further comprising an initial low speed period when moving from the blurred state to the vicinity of the just pin at a first lens speed. An autofocus device that moves at a very low speed and moves at the first lens speed after the initial low speed period has elapsed. 9. An input means for inputting an image signal of a video camera, an evaluation value generating means for extracting a high-frequency component of a specific area of the image signal to generate an evaluation value, and the video camera according to the evaluation value An auto-focusing device comprising: signal processing means for calculating a command value to be given to a focus drive unit of the lens of the above; and output means for sending the command value to a drive unit of the lens of the video camera. Providing a circuit for calculating a value obtained by adding the luminance of the image signal, examining the rate of change of the evaluation value and the luminance addition value in a predetermined period, when determining the presence or absence of shaking between the video camera and the subject, The data of the evaluation value and the luminance addition value during a first predetermined period after the trigger of the autofocus start is given does not contribute to the determination of the presence / absence of shaking. Over autofocus system. 10. The auto-focusing device according to claim 9, wherein the shaking determination is performed once when the lens is stationary before the start of the auto-focusing operation, and is periodically performed during the AF operation thereafter, so as to determine whether or not the shaking has occurred before the auto-focusing operation starts. An auto-focusing device that accurately determines the blurring of a blurred state that cannot be correctly determined by the shaking determination. 11. A video camera provided with the autofocus device according to claim 1. 12. An image signal of a video camera is inputted, a high-frequency component of a specific region of the image signal is extracted to generate an evaluation value, and a value obtained by adding luminance in the specific region of the image signal is calculated. A change in the evaluation value and the change in the brightness addition value during a predetermined period, and determine whether or not there is a fluctuation from the change rate of the evaluation value and / or the change rate of the change in the brightness addition value. Auto-focusing method of a video camera, including various steps. 13. The video camera autofocus method according to claim 12, wherein the speed of the focus lens is varied depending on the presence or absence of shaking. 14. The autofocus method for a video camera according to claim 12, wherein when it is determined that there is further shaking, the speed of the focus lens is set relatively high. 15. The video camera auto-focusing method according to claim 12, wherein when it is determined that there is no further shaking, the first from the blurred state to the vicinity of the just pin is determined.
An autofocus method for a video camera, wherein the video camera moves at a second lens speed from the vicinity thereof to the just focus at a second lens speed relatively lower than the first lens speed. 16. A recording medium storing the video camera autofocus method according to claim 12. Description: