JPH03268585A - Automatic focusing device - Google Patents

Abstract

PURPOSE:To quickly detect a change in an object even at the time of photographing a wall or the like having no pattern and to restart focusing operation by detecting the brightness difference of a brightness signal in an area set up in a picture, at the time of generating a prescribed change in the brightness difference after ending focusing operation, restarting the focusing operation. CONSTITUTION:Digital integration values V1, V1' corresponding to one field of a high band component more than 200kHz which is the cut-off frequency of an HPF out of brightness signals in the 1st and 2nd areas A, B are respectively outputted from integration circuits 59, 50g as the 1st and 2nd focus evaluating values Va, Vb. The evaluation value V1 and data W1 stored in the 4th memory 82 are inputted to a subtractor 84 by a variation detection circuit 81. A subtraction value (V1-W1) is outputted to the succeeding selection circuit 86 as a variation W1 of the value V1 from time immediately after the end of the focusing operation. The selection circuit 86 selects the variation Wk in accordance with an area switching signal Sa outputted from a focus motor control circuit 100 and outputs the selected value to the circuit 100.

Description

DETAILED DESCRIPTION OF THE INVENTION (A) Field of Industrial Application The present invention relates to an auto7 orcus device used in a video camera.

(b) Conventional technology In the autofocus device of a video camera, the method of using the video signal itself from the image sensor to evaluate the focus control state is essentially free of variation and has a limited depth of field. It has many advantages, such as being able to focus accurately even when shooting shallow or distant subjects. Moreover, there is no need for a special sensor for autofocus, and the mechanism is extremely simple.

Japanese Patent Application Laid-open No. 63-215268 (HO4N 5/2
32) discloses an example of the above-mentioned auto7 orcus device. The main points of this conventional technology are explained in the second section below.
This will be explained with reference to FIGS. FIG. 2 is an overall circuit block diagram of an autofocus circuit related to the prior art. The image formed by the lens (1) is turned into a video signal by an imaging circuit (4) including an image sensor, and a luminance signal therein is input to a focus evaluation value generation circuit (5).

This focus evaluation value generation circuit (5) is configured as shown in FIG. 3, for example.

The luminance signal is passed through a high pass filter (HP F) (5c)
The high-frequency components are separated and sent to the next stage detection circuit (
5d), the amplitude is detected. This detection output is converted into a digital value by the A/D conversion circuit (5e), only the signal of the focus area set at the center of the screen is extracted by the gate circuit (5f), and then the signal of the focus area set at the center of the screen is extracted, and the signal is sent to the integration circuit (5g). It is integrated for each field, and the focus evaluation value of the current field is changed. At this time, the vertical and horizontal synchronization signals separated from the luminance signal by the synchronization separation circuit (5a) are input to the gate control circuit (5b) in order to set seven orcus areas. The gate control circuit (5b) sets a rectangular focus area in the center of the screen based on vertical and horizontal synchronization signals and a fixed oscillator output, and generates a gate opening/closing signal that allows the luminance signal to pass only within this area. is supplied to the gate circuit (5C).

The focus evaluation value generation circuit (5) configured as described above outputs focus evaluation values for one field at a time.

Immediately after the start of the focusing operation, the first focus evaluation value is held in the maximum value memory (6) and the initial value memory (7).

After that, the focus motor control circuit (10) rotates the focus motor (3), which is a stepping motor, in a predetermined direction to rotate the focus ring (2) that supports the light receiving lens (1). Light receiving lens (1
) in the optical axis direction and monitor the output of the second comparator (9). The second comparator (9) compares the focus evaluation value after 7 orcasmoke driving with the initial evaluation value held in the initial value memory (7), and outputs the magnitude thereof.

The focus motor control circuit (10) includes a second comparator (9).
) outputs a large or small output, the focus motor drive circuit (31) is controlled to rotate the focus motor (3) in the initial direction until the current focus evaluation value is larger than the initial evaluation value. If this output is made, the rotation direction is maintained as it is, and if it is determined that the current evaluation value is smaller than the initial evaluation value, the rotation direction of the focus motor is reversed, and from then on, the first comparison is performed. Vessel (8)
Monitor output.

The first comparator (8) compares the current maximum focus evaluation value held in the maximum value memory (6) with the current evaluation value, and determines whether the current focus evaluation value is the maximum value memory (6). There are two types of comparison signals (51): large compared to the content (first mode), and reduced to more than a preset first threshold (second mode).
(S2) is output. Here, the maximum value memory (6) is the first
Based on the output of the comparator (8), if the current evaluation value is larger than the content of the maximum value memory (6), the value is updated and the maximum value of the focus evaluation value up to now is always held. Ru.

(13) is the focus ring () that supports the lens (]).
2) is a position memory that receives a focus ring position signal detected by a motor position detection circuit (30) and stores the 7-focus ring position as a lens position, and a maximum value memory (6). Similarly, the first comparator (8
) is updated so as to always hold the 7-occasion ring position when the maximum evaluation value is reached based on the output.

Note that the motor position detection circuit (30) specifically includes:
It consists of an up/down counter that is reset at the start of the focusing operation and counts up or down by setting the step/knob amount of the focusing motor (3), which is a stepping motor, as positive toward the near point and negative toward the far point. , the focus ring position signal becomes this count value.

The focus motor control circuit (10) includes a second comparator (9).
) focus motor in the direction determined based on the output (
3) While rotating, monitor the output of the first comparator (8), and compare the focus evaluation value with the maximum evaluation value to determine the preset point]
The focus motor (3) is reversed at the same time that the second mode, which is smaller than the threshold (ffi (λ4)), is instructed.

By reversing the focus motor (3), the moving direction of the light-receiving lens ( ) changes, for example, from a direction approaching the image sensor to a direction away from it, or vice versa, from a direction away from it to a direction toward it.

After this reversal, the contents of the position memory (13) and the current non-focus position are compared in the third comparator (14), and if they do not match, that is, the focus ring (2), that is, the lens (1) has a focus evaluation value. The focus motor control circuit (10) functions to stop the focus motor (3) when the focus motor (3) returns to the position where it is at its maximum. At the same time, the focus motor control circuit (10) outputs a lens stop signal (LS).

The change in lens position in the so-called mountain climbing focusing operation described above is the fourth
As shown in the figure.

(11) is the fourth point where the focus evaluation value at that point is held at the same time that the focusing operation by the focus motor control circuit (10) is completed and the lens stop signal (LS) is issued.
The content held in this fourth memory (11) is compared with the current focus evaluation value in the fourth comparator (12) at the subsequent stage, and the current focus evaluation value is compared with the content of the fourth memory (11). , becomes smaller than a preset second threshold value, it is determined that the subject has changed, and a subject change signal is output. When the focus motor control circuit (10) receives this signal, it restarts the hill-climbing focusing operation to follow changes in the subject.

This autofocus system has excellent focusing accuracy and compatibility with a wide range of subjects, but it is weak when shooting subjects where high-frequency components do not easily occur in the brightness signal even when in focus, such as walls without patterns. .

In other words, during the focusing operation of such a subject, the evaluation value does not fall from the maximum value of the obtained evaluation value to more than the first darkness value, and the lens position that takes this maximum value remains in focus. The position cannot be determined, and the lens is at infinity ~
The motor can no longer be stopped as it continues to displace the entire area between the periapsis points.

In order to prevent such a situation from occurring, even if the evaluation value does not fall from the maximum value to more than the first threshold, if the lens scans the entire area between the infinity point and the near point once, it will be possible to eliminate the problem. The motor can be stopped by returning the lens to the position where it takes the maximum value, or by returning it to the initial position where the focusing operation was started. In this case, stopping the motor was more important than the reliability of the focusing operation itself.
The probability that the final lens position is the focal point is significantly lower than that of a normal subject with sufficiently high contrast, but this is unavoidable.

(c) Problems to be Solved by the Invention As mentioned above, when photographing a subject in which high-frequency components are unlikely to occur in the luminance signal even when in focus, the lens may be photographed at a distance far away from the lens stop position after the completion of the focusing operation. Even if a subject with sufficiently high contrast enters the focus area, the blurring is too large and the evaluation value does not change enough to exceed the second threshold, and the focusing operation for the new subject is not restarted. A problem arises.

(d) Means for Solving the Problems The present invention extracts the amount of high-frequency components of the luminance signal within a region set within the screen as a focus evaluation value, and performs imaging so that the focus evaluation value becomes maximum. An autofocus device that performs a focusing operation by driving a system, detects a brightness difference within the area, and restarts the focusing operation when a predetermined change occurs in the brightness difference after the end of the focusing operation. It is characterized by Furthermore, it is characterized in that fuzzy inference is used to determine the change in brightness difference.

(E) Function Since the present invention is configured as described above, not only the focus evaluation value but also the luminance difference is used to determine whether to restart the focusing operation, so that even when photographing a wall without a pattern, the subject This makes it possible to quickly detect changes in the focus and restart the focusing operation.

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

FIG. 1 is a circuit block diagram of this embodiment.

Incidentally, in FIG. 1, the same parts as in FIGS. 2 and 3 of the prior art are given the same reference numerals, and explanations thereof will be omitted.

The luminance signal in the video signal output from the imaging circuit (4) is input to the focus evaluation value generation circuit (50).

The focus evaluation value generation circuit (50), as shown in FIG.
(P F (50c) A switching circuit (5h) that selects and outputs the output for each field, a detection circuit (5d) that detects the amplitude of the output of the switching circuit (5h), and an A/D that converts this detection output into a digital value. A converter (5e), a synchronization separation circuit (5a) that separates the synchronization signal from the luminance signal, and a comparison signal set in the center of the screen as shown in Figure 6 based on the separated synchronization signal and a fixed oscillator output. A gate circuit (5f
), and a second gate opening/closing signal that only allows passage of the A/D converted value of the luminance signal portion corresponding to the second area (B) which includes the first area (A) and has a larger area than this area. Gate control circuit (50f) supplied to the gate circuit (50f)
50b), the digital value of the high frequency component of the luminance signal in each area output from the gate circuit (5f) (50f) is set to 1.
Addition over the fields results in digital integration, and the integrated values are output as the first and second focus evaluation values (Va) and (Vb) for each field, respectively, and the integration is reset for each field. Circuit (5g) (50g
).

Here, the switching circuit (5h) selects HP F (5c) (50c) for each field at the output of the synchronous separation circuit (5a).
In order to select one of the outputs alternately, I-I P F (
In the field where 5c) is selected, the integration circuit (5c) is selected.
g) of the luminance signal in the first area (A).
5c) Digital integral value (vl) of one field of high-frequency components above 200K Hz, which is the cutoff frequency
is output as the first focus evaluation value (Va), and at the same time, the integrating circuit (50g) outputs the digital integrated value for one field of the same <200K Hz or higher high frequency component of the luminance signal in the second area (B). (Vl') is the second focus evaluation value (V
b) is output. In the next field, the first focus evaluation value (Va) is 600, which is the cutoff frequency of the HPF (50c) of the luminance signal in the first area (A).
The digital integral value (v2) and second focus evaluation value (vb) for one field of high-frequency components of KHz or higher are the same < 6
Digital integral value of high frequency components above 00KHz (V2
'), and the same process is repeated thereafter.

First and second focus evaluation values (Va) obtained in this way
(Vb) are input to separation circuits (51) and (52), respectively.

The separation circuits (51) and (52) both perform switching control generated based on a vertical synchronization signal generated from the synchronization separation circuit (5a) in order to switch the switching operation in the switching circuit (5h) for each field. The separation timing is controlled by the signal (SC), and the separation circuit (51) uses the first focus evaluation value (Va).
is separated into integral values m) and m') using each HPF, and each is output as an evaluation value (Vl) (Vl'). Similarly, the separation circuit (52) separates the second focus evaluation value (vb) into integral values (V2) (V2') using each HPF, and outputs each as an evaluation value (V2) (V2'). . Therefore, the evaluation values m)m')(V2)(V2') are all updated every two fields.

Next, the evaluation values (Vl) (V2) are respectively stored in the initial value memory (7
) (57), subtraction circuit (70) (80), memory (24
(58), relative ratio calculation circuits (25) (59), and switching circuit (20). On the other hand, the evaluation value (Vl'
)(V2') are the relative ratio calculation circuits (25) and (59), respectively.
) is entered.

The initial value memory (7) (57) stores an evaluation value (Vl) at the time when the focus motor (3) first starts rotating in a predetermined initial direction in order to start the focusing operation.
(V2), and while the motor (3) rotates in the initial direction and the lens (1) is displaced, the evaluation value m) is obtained when two fields have passed after the motor is started.
(V2) is updated, the subsequent subtraction circuit (70) (
80), the initial value memory (7) (
The value obtained by subtracting the contents of 57) is evaluated as vi(■1
) (v2) as the amount of change (△VI ) (△V2),
7 Output to the one-custom motor control circuit (100').

The memory (24) (58) holds the evaluation values (Vl) (V2) obtained after starting the motor (3) for two fields.
This is input to the relative ratio calculation circuits (25) and (59) at the subsequent stage.

The relative ratio calculation circuit (25) (59) is composed of a divider (61), a memory (62), and a subtracter (63) as shown in FIG. (Vl')
Each time (V2') is updated, memory (24) (58
) The ratio with the latest evaluation value m) (V2) held in
Vl'/Vl, V2'/V2 as relative ratio (rl) (r2)
Calculated as

Here, the relative ratio (rl) is the integral value (vl) of the yield (7') when using HP F (5c),
This is the ratio to the integral value (Vl') when using HP F (50c), and the relationship between the area integral value and the focus ring position (lens position) when the subject is the same is as shown in Figure 7. Become. That is, HP F (50c
) has a steep peak, and the integral value at HP F (5c), which has a low cutoff frequency, has a gentle peak.

Therefore, if the relationship between this relative ratio and the degree of blurring of the object (the amount of movement or deviation from the lens position at the time of focusing) is plotted in a graph, it will become a monotonically decreasing characteristic curve as shown in FIG.

This is because the relative ratio state quantity is a function value that can express the in-focus state (degree of blur) of the subject in the same way as the focus evaluation value, and because it is expressed as a ratio, it is a kind of normalized state quantity. , and has the property of being less affected by the environment in which the subject is placed. For example, when the illuminance of the subject changes, the absolute value of the focus evaluation value changes, but the relative ratio does not change significantly. Normally, the above properties are independent of the type of subject, so this relative ratio can be used as a parameter for the degree of blur. When this monotonically decreasing characteristic curve in Fig. 8 is made to correspond to the lens position, that is, the focus ring position, it changes almost linearly toward the near point and point ■ with the in-focus position as the apex, as shown by the chain line in Fig. 9. A characteristic diagram is obtained.

The relative ratio (rl) b2) obtained by the divider (61) is
Memory (62) and subtraction! (63) is input. The memory (62) functions to hold and delay the input relative ratio for two fields and supply it to the subtracter (63). The previous relative ratio held in the memory (62), that is, two fields ago, is subtracted from the latest relative ratio held in the memory (62), and this subtracted value is calculated as the change in relative ratio (rl) (r2) (Δr1) (Δr2). It is output to the seven-order smoke control circuit (100) as a signal. still,
It goes without saying that the amount of change (Δr1) (Δr2) can be a negative value if the initial rotational direction of the motor is opposite to the focusing direction.

The switching circuit (20) is a focus motor control circuit (10).
By the area switching signal (Sa) output from 0),
Select the area to be used for focusing operation and enter the bifocal evaluation value (V
l) Select one of (V2).

The focus evaluation value selected by the switching circuit (20) is supplied to the maximum value memory (6) and the first comparator (8), and similarly to the conventional example, a signal for executing hill climbing control is sent to the focus lens. The position memory (13) and third comparator (14) supplied to the control circuit (100) operate exactly the same as in the conventional example shown in FIG.

However, the focus evaluation values obtained from the switching circuit (20) are updated every two fields, and the comparison operations of the first comparator (8) for hill climbing operations are all performed every two fields. Ru.

The zoom position detection unit (64) generates a signal indicating the current focal length (Z) of the lens according to the wide-angle to telephoto zoom range of the zoom mechanism having a well-known zoom lens attached to this video camera. Output to the control circuit (100).

The change amount detection circuit (81) functions to detect the amount of change in the evaluation value m) (V2) after the completion of the focusing operation, and specifically, as shown in FIG. memory(
82) (83), subtracters (84) (85), and a selection circuit (86). The fourth memory (82) receives a lens stop signal (
The evaluation value (vl) when LS) is issued, that is, immediately after the completion of the focusing operation, is stored as (Wl). Similarly, the fifth memory (82) stores the evaluation value (V2) when the lens stop signal (LS) is issued as (W2).

The subtracter (84) outputs the evaluation value (■1) and the fourth memory (82
) is input, and the subtracted value of Vl - Wl is calculated as the amount of change in the evaluation value (■1) from immediately after the end of the focusing operation (
ΔWl) is output to the subsequent selection circuit (86).

Similarly, the subtracter (85) inputs the evaluation value (■2) and the data (W2) held in the fourth memory (83), and receives V2-W
The subtracted value of 2 is output to the subsequent selection circuit (86) as the amount of change (Δ1V2) in the evaluation value (V2) from immediately after the end of the focusing operation. Note that the evaluation value m) (V2) is updated every two fields even after the motor stops, so the amount of change (△W
1) (ΔW2) also changes every two fields. The selection circuit (86) receives an area switching signal (Sa) issued from the focus motor control circuit (100) to the switching circuit (20).
According to, the amount of change (△Wk) (k= 1 or2)
Select one of the focus motor control circuits (
100). Therefore, if the first area (A) was selected as the focus area during the focusing operation, the amount of change (ΔWl) would be the same, and if the second area (B) was selected, the amount of change (Δh "2) is selected.

As shown in FIG. 11, the average brightness detection circuit (65) is composed of a detection circuit (65a), an A/D conversion circuit (65b), and an integration circuit (65c), and detects a brightness signal corresponding to the entire screen. The detection circuit (65a) performs amplitude detection, and the detected output is sequentially A/D converted by the A/D conversion circuit (65b), and then the integration circuit (65c) performs A/D conversion for one field.
All the converted data is integrated, digitally integrated, and the control circuit (
100).

The contrast detection circuit (66) is configured as shown in FIG. 13 and detects the contrast in the horizontal direction within the first area (A), that is, the brightness difference. Here, the operation of this contrast detection circuit (66) will be explained. First, in order to detect contrast, the first area (A) is divided into 16 small areas (Nij) (i,
j = 1 to 4), and a separation circuit (67) separates the luminance signal for each small area, and 16 digital integrators (Kij) prepared for each small area (i, j = 1 ~4
).

Each digital integrator has exactly the same configuration as the average brightness detection circuit (65) shown in FIG. It is derived as These products ff
1 (Fij) is a Max-Min calculation circuit (Li) (i=1 to 4) with four small areas arranged in the horizontal direction as one set.
is input. That is, the integral value (Fij) (j=1 to 4)
is input to the Max-Min calculation circuit (Ll), and similarly, the integral values (F 2j ) (F 3j ) (F 4j) are input to the Max-Min calculation circuits (L2), (L3), and (L4), respectively. .

The Max-Min calculation circuit (Li) selects the maximum value and minimum value among the four integral values of the small areas arranged in the horizontal direction, and (
Maximum value - minimum value) is subtracted, and this subtracted value (G i
) is output to the subsequent maximum value detection circuit (68).

The maximum value detection circuit (68) functions to extract the maximum value among the subtracted values (Gi) for the four lines, and outputs this maximum value as the contrast (ΔEl). Therefore, in the end, the contrast (8El) corresponds to the digital value of the luminance difference itself in the row having the largest horizontal luminance difference in the first area (A) for one field.

The contrast thus obtained (8El) is supplied to a change amount detection circuit (87) to calculate the amount of change in contrast from the contrast immediately after the completion of the focusing operation. That is, as shown in Fig. 23, the value of contrast (△El) when the lens stop signal (LS) is issued is stored in the memory (88) as (EE), and the value obtained for each field thereafter is stored. contrast (△E1) and Δ(△E1)
; 8El, -EE is calculated, and this Δ(ΔEl) is output to the control circuit (100) as the amount of change in the brightness contrast in the first area (A) from the end of the focusing operation.

The focus motor control circuit (100) includes a first comparator (
8) and third comparison! Based on the (1, 4) output, a mountain climbing focusing operation is performed as in the conventional example. In addition, the output from the subtraction circuits (70) and (80), that is, the evaluation value (Vl) (
v2) from the initial value (△v1) (△V2), the output from the relative ratio calculation circuits (25) (59), that is, the evaluation for the first area (A) (if (V 1 ), each The change from the initial value of the relative ratio of areas (rl) (r2) (
The direction of the focused point is determined by fuzzy inference based on five types of data: Δr1) (Δr2), and the output from the contrast detection circuit (66), that is, the brightness contrast (ΔEl) in the first area (A). In addition to determining, the separation circuit (51)
The evaluation value output from (■1), the average brightness detection circuit (6
5), i.e. the average brightness of the entire screen (IR3),
Area selection is performed by fuzzy inference based on four types of data output from the zoom position detection circuit (64), ie, current focal length (Z) and contrast (ΔEl). Furthermore, the amount of change in contrast (△(△E]))
, focal length (Z), and amount of change in focus evaluation value after focusing of the focus area (Wk) (however, k = 1 or 2)
Based on these three types of data, confirmation of subject change and accompanying decision to restart the focus morph are performed using fuzzy inference.

Next, the processing for direction control, area selection, and restart determination using the above-mentioned fuzzy inference will be described in detail.

First, the direction determination process is executed according to the flowchart shown in FIG.
rl) and contrast (8El) as input variables, take a value from O to 1 as the conclusion part, and set the parameter (di ).

Here, the rule is [Rule (1)] rif△v1 is large and △r1 is large then
dl=1.0” [Rule (2) Corif Δv1 is large and Δr1 is not large t)+
en d2=0.7" [Rule (3)] "100v1 is small and Δ■2 is small and ΔE1 is small then d3=0.2" is set.

Next, each of the above rules will be explained.

[Rule (1)] is defined by membership functions as shown in FIGS. 11(a) and (b). Figure 14(a) shows [
Indicates the membership value of condition (1) of rule (1) that △■1 is large J, and the amount of change (△Vl
), and the amount of change (△V
1. ) becomes larger, the membership value (al
It is a function that includes a monotonically increasing straight line where l) increases, and from this function the membership value (
all) can be found.

Figure 14(b) shows the rule (1
It is a membership function for the input variable change (Δrl), and is a monotonous function in which the membership value (u 12) increases as the change (Δr1.) increases. This is a function that includes an increasing straight line, and the membership value (u 12) according to the change (Δrl) is determined from this function.

Rule (1) takes into consideration the case where both the focus evaluation value and the relative ratio in the [th] area (A) increase, and in this case, the in-focus point is in the current movement direction of the lens (1). Since the possibility is high, select the first area (A,).
The conclusion part (dl) is set to d1=1.

[Rule (2)] is defined by membership functions as shown in FIGS. 11(a) and 11(b). Figure 15(a)
is condition (1) of rule (2) that “△■1 is large”
represents the membership value of the input variable (Δ
Vl), from which the membership value (u21) can be found.

Figure 15(b) shows the membership value of condition (2) of rule (2) that “△r1 is not large”, and as the change (△rl) increases, the membership value (u
22) is a function that includes a monotonically decreasing straight line in which the value decreases, and from this function, a membership value (u22) corresponding to the amount of change (Δrl) can be found.

Rule (2) is for the case where the fluctuation trends of the focus evaluation value and relative ratio are different, and the in-focus point may be in the opposite direction to the current lens movement direction, so the conclusion part (d2) is slightly smaller than dl. It is set to d2=0.7.

[Rule (3)] is shown in Figure 11 (a), (b), and (c).
) is defined by a membership function such as 16th
Figure (a) shows the membership value of condition (1) of rule (3) that "rVl is small", and is a function that includes a monotonically decreasing straight line in which the membership value (u31) decreases as the evaluation value (vl) increases. From this function, the membership value (u 31) corresponding to the evaluation value (■1) can be found.

Figure 16(b) shows the rule (3
) is the membership value of condition (2), and the amount of change (Δ
As the membership value (u32) increases, the membership value (u32
) is a function that includes a monotonically decreasing straight line that decreases, and from this function, the membership value (u
32) can be found.

Figure 16 (C) shows the rule (3) that [△E] is small.
) shows the membership value of condition (3), and as the contrast (△El) increases, the membership value (
u33) is a function that includes a monotonically decreasing straight line that decreases,
A membership value (u33) corresponding to the contrast (ΔEl) is determined from this function.

Rule (3) takes into consideration the case where the focus evaluation amount is low in the first area (A) and no change can be extracted, and in this case, if the contrast of the first area (A) is small, the first area (A)
Assuming that there is no subject to be focused on, the direction is determined by the change in the focus evaluation value in the second area (B), and the amount of change in the focus evaluation value in the second area (B) is negative. In this case, the conclusion part (d3) is set as low as d3 = 0.2 in order to make it easier to reverse the lens movement direction, assuming that the in-focus point is likely to be in the opposite direction to the current lens movement direction. has been done.

A direction determination process in which a direction determination parameter (D) is calculated from each of the above-mentioned rules and a direction is finally determined from this parameter will be described using the flowchart of FIG. 20.

Each membership value (
u 1j) (i, j: integer) is determined according to each input variable, then in S T E P (101), the minimum membership value for each rule is determined as the probability of each rule being met. It is calculated as (Ul). For example, the rule (
Regarding 1), according to Fig. 14, u 11 > u 12
Therefore, the degree of establishment (Ul) is U 1 = u 12, and based on rule (2), from Fig. 15, u 21>
1122, so the degree of establishment (U2) is U2=u22, and furthermore, by twisting rule (3), from Fig. 15, u
31< u 33 < u 32, so the degree of establishment (U3
) becomes U3=u31.

Based on the degree of establishment (Ul) of each rule obtained in this way, S
At T E P (102), a parameter (D) for direction determination is calculated using the following formula % formula % (1). This formula (1) means that each conclusion part is weighted and averaged based on the degree of establishment of each rule.

In S T E P (103), the obtained parameter (
D) determines the direction, and specifically, if D≧0.5, the focus position is in the current lens movement direction, so a control command is issued to control the drive of the motor (3) to maintain the current direction. If D<0.5, the focus position is in the opposite direction to the current direction, so a control command to immediately reverse the motor (3) is issued to the focus motor drive circuit (31).

In this way, the direction of the in-focus point is determined with high precision by considering the five factors, and the motor (3)
The aforementioned hill-climbing focusing operation is performed while rotating the lens and moving the lens.

Next, the area selection process is executed according to the flowchart shown in FIG. 21, and the fuzzy inference used at this time is
The focus evaluation value (■1), contrast (△El), focal length (Z), and average brightness (IR5) are input variables, and the conclusion part takes a value between O and 1, and when it is small, the first area (
A), membership Vi (ai) is set, which means that the second area (B) is used when the area is large.

Here, the rule is: [Rule (4)] rif (Vl) is large then a4-0. Oj [Rule (5)] zf (Vl) is medium and (IR3) is small th
en a5=0.8J [Rule (6)] rif (Vl) is small and (ΔEl) is large and (Z) is not small then a6-0.3” Next, each of the above rules will be explained.

[Rule (4)] is defined by a membership function as shown in FIG. 17, and this FIG. 17 shows the membership value of condition (1) of rule (4) that "rVl is large",
This is a membership function for the focus evaluation value (■1), and it is a function that includes a monotonically increasing straight line in which the membership value (U41) increases as the evaluation value (Vl) increases, and from this function, the evaluation value (vl) The corresponding membership value (U41) is determined.

Rule (4) is that when the focus evaluation value (■1) is large, the conclusion part (a4) is a4
=o, set to o.

[Rule (5)] is defined by membership functions as shown in FIGS. 18(a) and (b). Figure 18(a)
The condition of rule (5) that “r(Vl) is medium” (
1) is a mountain-shaped function in which the membership value is the maximum when the evaluation value (vl) is a predetermined value,
From this function, a maintenance value (u51) corresponding to the evaluation value (vl) is determined.

FIG. 18(b) shows the membership value of condition (2) of rule (5) that "r(IR5) is small", and as the average luminance (IR3) increases, the membership value (
It is a function that includes a monotonically decreasing straight line in which u52) becomes smaller,
A membership value (u52) corresponding to the average brightness (IR5) is determined from this function.

Rule (5) is the focus evaluation value (v
l) is not very large and the average brightness of the screen is dark.In this case, the S/N ratio of the video signal is poor and the reliability of the focus evaluation value is low, so in order to capture more information, focus The conclusion part (a5) is set to a5=0.8 so that the second area (B) is likely to be given priority as the area.

[Rule (6)] is shown in Figure 11 (a), (b), and (c).
) is defined by a membership function such as
Condition (1) of rule (6) that “rm) is small”
It is a function that includes a monotonically decreasing straight line in which the membership value decreases as the evaluation value (vl) becomes thicker, and the membership value (u61) according to the evaluation value (vl) is calculated from this function. Seek.

Figure 19(b) shows the membership value of condition (2) of rule (6) that "(ΔEl) is large", and the membership value (u62) increases monotonically as the contrast (8El) increases. This is a function including an increasing straight line, and a membership value (u62) corresponding to the contrast (ΔEl) is determined from this function.

FIG. 19(c) shows the membership value of condition (3) of rule (6) that "(Z) is not small", and as the focal length (Z) increases, the membership value (u6
3) is a function that includes a monotonically increasing straight line that increases, and from this function the membership value (u
63) can be found.

Rule (6) considers the case where the focus evaluation value of the first area (A) is small but the contrast within the first area (A) is high, and the case where the focal length is short and the depth of field is deep. The conclusion part (a6) is set to a6=0.3 so that the first area (A) is relatively likely to be given priority as it is considered to be blurry or has some kind of subject.

The area selection process of calculating area selection parameters (Y) from each of the above-mentioned rules and finally making a two-norm selection based on these parameters will be described using the flowchart of FIG.

At S T E P (200), condition (j) of rule (i) (i = 4 ~
6. Once the membership value (uij) of j=1 or 2 or 3) is determined, then S T E
In P (201), the minimum membership value for each rule is calculated as the degree of establishment (Ui) of each rule.

For example, for rule (4), the degree of establishment (U4) is U4
= u41, and for rule (5), as shown in Figure 18, u 51 < u 52, so the degree of establishment (U5) is U5 = u51, and further, for rule (6), the first
From FIG. 9, since u 61 < u 62 < u 63, the degree of establishment (U6) is U6=u61.

Based on the degree of establishment of each rule (U i) obtained in this way,
In S T E P (202), a parameter (Y) for area selection is calculated using the following formula % (2). This formula means that each conclusion is weighted and averaged based on the degree of establishment of each rule. STE″P (
203), the area is selected according to the obtained parameter (Y). Specifically, if Y2O, 5, the second area (B) is selected as the focus area, and if Y<0,5, the first area (A ) is selected.

In this way, a highly accurate area selection is made taking into account the four factors, and focus evaluation is performed by the switching circuit (20) in order to execute the hill climbing focusing operation with the corresponding focus evaluation value. A value selection is made.

Note that this area switching process is also executed during a focusing operation and an operation of monitoring changes in the subject after reaching the in-focus point. In addition, immediately after the area is actually switched, the switching circuit (20)
Since there is a temporary large fluctuation in the focus evaluation value obtained from the 1st comparator (8), in order to suppress malfunctions due to this fluctuation, the comparison operation in the first comparator (8) is performed so that the same comparison result is not repeated during the mountain climbing focusing operation. It is configured to output this comparison result as valid only when it is obtained three times.

The restart determination process is executed according to the flowchart shown in FIG. 24, and the fuzzy inference used at this time is as follows:
The amount of change (Δ(△E1)), the amount of change (△Wk), and the focal length (Z) are used as input variables, and the conclusion part is a value between 0 and 1. Parameter (hi) that starts up and maintains the stopped state when it is small
It is said that

Here, the rules are: [Rule (7) rifΔ(ΔEl) is large in the positive direction and Z is not small then h7=o, 8J [Rule (8)] rifΔ(ΔEl) is not large in the positive direction and7 Orcus area △Il'k is not small then hs=
o, s” [Rule (9)] The absolute value of rifΔ(ΔEl) is not large and the 8Wk of the 7 orcus area is small then h9=Q, 0” [Rule (10) CorifΔ(8El) is in the positive direction is large and △Wk of focus area is not small then hlO=1.0
, 1.

Next, each of the above rules will be explained.

[Rule (7)] is defined by membership functions as shown in FIGS. 22(a) and (b). Figure 25(a)
is based on the rule that “△(ΔEl) is large in the positive direction” (7
), it is a membership function for the input variable amount of change (△(ΔE1)), and as the amount of change (△(△E1)) increases in the positive direction, the membership value increases. This is a function that includes a monotonically increasing straight line where the value (u71) increases, and from this function the amount of change (△(
The membership value (u 71) corresponding to 8E1)) is determined.

Figure 25(b) shows the rule “Z is not small” (7
), it is a membership function for the input variable focal length (Z), and as the focal length (Z) becomes larger (in the direction of the force), the membership value becomes ) is a function that includes a monotonically increasing straight line that increases, and the membership value (u72) according to the focal length (Z) is determined from this function.

Rule (7) states that if the brightness contrast in the first area (A) is greater than when the camera is in focus, and the focal length is long and the depth of focus is shallow, there is a possibility that the subject has entered the first area (A). The conclusion part (h7) is set to h7=0.8 so that restarting is easy because it is extremely high.

(Rule (August is defined by membership functions as shown in Figures 26(a) and (b). Figure 26(a) is "Δ
(ΔEl) is not large in the positive direction” rule (8)
It shows the membership value of condition (1) and is a membership function for the input variable amount of change (Δ(ΔE1)), from which the membership value (u81) can be found.

FIG. 26(b) shows the membership value of condition (2) of rule (8) that "8Wk of the focus area is not small", and as the focus area of the first or second area which is an input variable. It is a valley-shaped function in which the membership value (u82) increases as the absolute value of the selected amount of change (△Wk) (k4or2) increases, and from this function The membership value (u82) is determined.

Rule (8) considers the case where the brightness contrast of the first area has not increased, but there has been a change in the focus evaluation value of the focus area, and in this case, focusing on the change in the evaluation value, , the conclusion part (h8) is set to h8 = 0.8, which is slightly higher, to make restarting easier since there is a high probability that a change in the subject has occurred.

[Rule (9)] is defined by membership functions as shown in FIGS. 22(a, 22b). Figure 27(a) shows the rule that “the absolute value of △(ΔEl) is not large” (
9) shows the membership value of condition (1), and the amount of change (
It is a mountain-shaped function in which the membership value (u91) increases as △(ΔE1)) approaches zero, and the membership value (u91) according to the amount of change (△(△E1)) can be found from this function. .

Figure 27(b) shows the membership value of condition (2) of rule (9) that "8Wk of the focus area is small" and is selected as the focus area from the first or second area which is an input variable. The amount of change (
It is a mountain-shaped function in which the membership value (u92) increases as the absolute value of △Wk) (k = 1 or 2) becomes smaller, and from this function, the membership value (u92) according to the amount of change (8Wk) is Seek.

A membership value (u92) corresponding to the amount of change (ΔWk) is determined from this function.

Rule (9) takes into consideration the case where both the brightness contrast of the first area (A) and the focus evaluation value of the focus area have not changed much since the time of focusing. In this case, there is a change in the subject. Since there is a small possibility that this has occurred, the conclusion part (h9) is written as
It is set.

(Rule (], O)) is shown in Figure 28 (a) (b)
) is defined by a membership function such as 28th
Figure (a) shows the membership value of condition (1) of rule (10) that “△(△El) is large in the positive direction”,
It is a membership function for the input variable amount of change (△(△E1)), and from this the membership value (ul
ol) can be found.

FIG. 28(b) shows the membership value of condition (2) of rule (10) that "ΔWk of the focus area is not small", and it is assumed that the focus area of the first or second area is an input variable. As the absolute value of the selected amount of change (ΔWk) (k=1or2) increases, the membership value (ul) increases.

2) is a valley-shaped function that increases, and from this function, the membership value (u 102) according to the amount of change (△Wk)
is found.

Rule (10) takes into account the case where the brightness contrast of the first area increases and the focus evaluation value of the focus area changes. In this case, rule (8)
), the probability that the subject has changed is higher by the amount that the contrast has increased, and the conclusion part (
hlO) is set to a slightly higher value of hlO=1.0.

The process of calculating the restart decision parameter (H) from each of the above-mentioned rules and finally determining whether or not to restart based on this parameter will be described using the flowchart of FIG. 24.

Each membership value (
u1j) (i, j: integer) is determined according to each input variable, then in S T E P (301), the minimum membership value for each rule is determined as the probability of each rule being met. (Ui). For example, the rule (
Regarding 7), according to Fig. 25, u 71 < u 72
Therefore, the degree of establishment (U7) is U7=u72, and as for rule (8), as shown in FIG. From Figure 27, u 91 < u 92, so the degree of validity (U9) is U9=u91, and furthermore,
Regarding rule (10), from Figure 28, u 101>
Since u 102, the degree of establishment (U 10) is [11
0=u102.

Based on the degree of establishment of each rule (U i) obtained in this way,
At S T E P (302), a parameter (H) for determining restart is calculated using the following formula (3). This equation (3) means that each conclusion part is weighted and averaged based on the degree of establishment of each rule.

In S T E P (303), the obtained parameter (
H) determines whether to restart or not, specifically H2
If O, 5, it is assumed that there is a change in the screen, and the initial value memory (7) (57), maximum value memory (6), and position memory (1
3) etc., and the control circuit (100) issues various control commands to restart the series of mountain climbing focusing operations from the beginning. If H<0.5, it is assumed that there is no change in the subject and the motor (3) maintain the stopped state. In this manner, a highly accurate restart decision is made after considering the three factors, and a focusing operation that quickly follows changes in the subject becomes possible.

Further, the area division, rules, etc. are not limited to those in this embodiment. Further, it goes without saying that the operation of this embodiment can be executed by software using a microcomputer.

(G) Effects of the Invention As described above, according to the present invention, changes in the subject after the end of the focusing operation can be detected regardless of the focusing state, and it is possible to restart the focusing operation with high precision and appropriately. It becomes possible. Furthermore, by using fuzzy inference for this determination, it is possible to comprehensively determine whether to restart the focusing operation, taking into account the brightness contrast and other factors.

[Brief explanation of drawings]

1, 4 to 27 relate to one embodiment of the present invention, in which FIG. 1 is an overall circuit block diagram, FIG. 4 is an explanatory diagram of mountain climbing focusing operation, and FIGS. 5 and 10. , Fig. 11, Fig. 13, Fig. 22, and Fig. 23 are main circuit block diagrams, Fig. 6 is an explanatory diagram of area setting, Fig. 7 is a diagram of the relationship between lens position and focus evaluation value, and Fig. 8 The figure is a diagram of the relationship between relative ratio and blur degree, No. 9.
The figure shows the relationship between lens position, focus evaluation value, and relative ratio. Figure 12 is an explanatory diagram of area division. Figure 25, 2nd
6, FIG. 27, and FIG. 28 are diagrams showing membership functions, and FIG. 20, FIG. 21, and FIG. 24 are diagrams showing flowcharts. Further, FIGS. 2 and 3 are circuit block diagrams of conventional examples. (1)... Lens, (50)... Focus evaluation value generation circuit, (81) (87)... Change amount detection circuit, (100
)... Focus motor control circuit, (4)... Imaging circuit.

Claims (2)

[Claims] (1) The amount of high-frequency components of the luminance signal within the area set in the screen is extracted as a focus evaluation value, and the focusing operation is performed by driving the imaging system so that the focus evaluation value becomes maximum. An autofocus device, characterized in that the autofocus device detects a brightness difference within the area, and restarts the focusing operation when a predetermined change occurs in the brightness difference after the end of the focusing operation. (2) The autofocus device according to item 1, wherein fuzzy inference is used to determine a change in brightness difference.