JPH0477289B2 - - Google Patents

Description

[Detailed description of the invention]

(Technical Field of the Invention) The present invention relates to a focus detection device for an optical device such as a camera, which detects the focus adjustment state of an imaging optical system that forms an optical image of a target object. (Background of the Invention) A conventional focus detection device for a single-lens reflex camera is composed of a large number of photoelectric conversion elements arranged to generate two images of almost the same part of a subject by a light beam passing through different areas on the pupil of a photographic lens. The relative displacement is guided onto a pair of photoelectric conversion element arrays, and the relative displacement of the light image on the photoelectric conversion element array is detected by calculating the photoelectric output of the photoelectric conversion element array, and focus detection is performed from this detection result. Detection methods are known. Projecting a subject image onto a photoelectric conversion element array and processing the photoelectric output in this manner is nothing but processing a frequency component related to a certain spatial frequency band in the subject image. If the spatial frequency components to be processed are only relatively low-order ones, even if the photographing lens is far away from the in-focus position and the subject image is significantly blurred, the low-order spatial frequency components will be included in the subject image. In other words, since it remains in the photoelectric output, even for a large amount of defocus (the amount of defocus is the amount of deviation in the optical axis direction between the intended focal plane of the imaging optical system and the image plane of the subject), it is difficult to focus on the front or rear focus or It has the advantage of being able to determine the back focus, but low-order spatial frequency components are susceptible to various electrical or optical error factors, making it difficult to achieve high detection accuracy when determining the focus position itself. There is a drawback that it cannot be done. On the other hand, when the spatial frequency components to be processed are only relatively high-order ones, it is less susceptible to the effects of the various error factors mentioned above, and the amount of defocus can be accurately determined near the in-focus position. In other words, high-precision focus detection is possible. However, if the photographing lens moves far away from the focus position, the high-order spatial frequency components in the subject image will drastically decrease, so the accuracy of focus detection will drop significantly, making it impossible to even determine front and rear focus. There is a drawback to this. In addition, when processing a mixture of high-order spatial frequency components and low-order spatial frequency components without separating them, depending on the nature of the subject, one of the high-order spatial frequency components and low-order spatial frequency components may be The flaws become obvious and undermine the strengths of the other. In other words, the focus detection ability can be roughly divided into two parts: determining the amount of defocus accurately in the vicinity of the in-focus position, and determining the approximate amount of defocus or the front and back focus when far away from the in-focus position. Ideal focus detection can be performed by handling the former mainly using high-order spatial frequency information and the latter mainly handling low-order spatial frequency information. As relative displacement detection methods, a method using correlation calculation and a method of detecting a phase shift after Fourier transformation are known. In the case of the Fourier transform phase detection method, Fourier transform necessarily extracts specific spatial frequency components, and filter processing is performed.
In this case, when the subject contains many of the specific spatial frequency components, highly accurate focus detection is performed, but when the subject does not include the specific spatial frequency components, focus detection becomes impossible. , the present applicant has proposed a focus detection device that detects a plurality of spatial frequency components in Japanese Patent Laid-Open No. 124112/1983. On the other hand, in the conventional relative displacement detection method using correlation calculation, calculation processing is performed using all spatial frequency components below the Nyquist frequency determined by the photoelectric conversion element pitch, and high-order spatial frequency components and low-order spatial frequencies Because the components are not separated and are processed in a mixed manner, depending on the subject, the drawbacks of one of the high-order spatial frequency components and the low-order spatial frequency components become apparent, and the advantages of the other are impaired. Furthermore, a focus detection device that saves correlation calculation time and memory capacity has been proposed in Japanese Patent Laid-Open No. 107312/1983, which was filed before the present application. In this device, when the amount of defocus is large, the correlation calculation time and memory capacity are reduced by coarsening the sampling pitch by selecting every other source data or by taking the sum (or average) of the data of two adjacent pixels. In addition, when the amount of differential focus near focus is small, the original data is used as is. The characteristics of the data processing, that is, the filter processing performed by this device, are as shown in Figure 6a.The second filter uses the original data as it is, and the sampling pitch is made coarser and the average of two pixels is taken. corresponds to the first filter. However, in this case, even the second filter, which extracts relatively high-order spatial frequency components, contains a large amount of DC components (components with zero spatial frequency), so electrical or optical A problem arises in that focus detection accuracy deteriorates due to the influence of various error factors. Therefore, the purpose of the present invention is to eliminate the influence of various electrical or optical error factors caused by DC components and to improve the accuracy of focus detection near the in-focus area. It is an object of the present invention to provide a focus detection device capable of detecting focus even when the focus detection device does not include the focus detection device. (Summary of the Invention) In order to achieve this object, the present invention provides first and second filter means for inputting and filtering the photoelectric outputs of a pair of photoelectric conversion element arrays; A correlation calculation is performed based on the filtered output obtained by the means, and focus adjustment is performed based on the calculation result. In other words, the first filter means is configured to mainly extract low-order spatial frequency components so that focus detection can be performed even when the amount of defocus is large or when the subject does not include high-order spatial frequency components, In addition, the second filter means mainly extracts high-order spatial frequency components while containing almost no direct current component (component with zero spatial frequency) of the subject image, in order to increase focus detection accuracy near the in-focus area. is configured to do so. (Embodiment of the Invention) An embodiment of the focus detection device of the present invention will be described below with reference to the drawings. In FIG. 1 showing the focus detection optical system, a large number of small lenses 2 are arranged in a straight line on the planned imaging plane of the imaging optical system 1, such as a photographing lens whose focus is to be detected.
A small lens array 2 consisting of lenses a, 2b, . . . 2n is arranged. This small lens array 2 is formed on the plane side of the plano-convex lens 3. This small lens array 2 receives first and second beams of light that have passed through areas 1A and 1B on the pupil plane of the imaging optical system 1, respectively.
A subject image is formed. The amount of deviation of both subject images in the direction perpendicular to the optical axis corresponds to the amount of defocus of the imaging optical system 1. A photoelectric conversion device 4 for photoelectrically detecting the amount of deviation between the two images is placed immediately after the small lens array 2, and this device is connected to each of the small lenses 2a to 2.
A pair of photoelectric conversion element groups ( ₁ , ₁ ) facing n
( ₂ , ₂ )... has ( _o , _o ). The first one located at the top of these pair of photoelectric conversion elements
Photoelectric conversion element arrays ₁ , ₂ ... _o transfer the first subject image to the second photoelectric conversion element array located at the bottom.
_b1 , ₂ ... _o photoelectrically convert the second subject image, respectively. These photoelectric conversion elements ₁ ... _o , ₁ ... _o
Let the photoelectric output of a ₁ ,... correspond to the sign of each element.
Assuming that a _o , b ₁ , ... b _o , the photoelectric conversion device 4 converts these photoelectric outputs in time series as a ₁ , b ₁ , a ₂ , b ₂ ...
Outputs a _o and b _o in order. Such a focus detection optical system is disclosed in detail in USP4185191. Note that the photoelectric output referred to herein includes an output obtained by linearly amplifying or logarithmically amplifying the output of a photoelectric conversion element. Also,
Each small lens 2a...2n of the above small lens array 2
As shown in Figure 2, the MTF characteristics determined by the shape of
The pitch of the lens array is reduced from low-order frequencies close to zero.
It exhibits a high value over the Nyquist frequency 1/2p ₀ given as a function of p ₀ . Assuming that the first and second optical images projected onto the small lens array 2 each contain sufficient amounts of low-order spatial frequency components to high-order spatial frequency components near the Nyquist frequency, the above MTF characteristic Therefore, the photoelectric output also includes all the spatial frequency components from the low order to the high order. In FIG. 3, input terminals of a first filter means 5 and a second filter means 6 are connected to the output terminal of the photoelectric device 4. These first and second filter means 5, 6 respectively output the photoelectric outputs a ₁ , b ₁ , a ₂ , b ₂ . . . outputted in time series from the photoelectric device 4 as described above.
...a _o and b _o are filtered and chronologically
A ₁ , B ₁ , A ₂ , B ₂ ... Outputs A _o and B _o . The MTF characteristics of these first and second filter means 5 and 6 are shown in FIGS. 4a and 5a, respectively. As is clear from each figure, the first filter means 5
The MTF is sufficiently large at low-order frequencies, decreases continuously from there, and reaches almost zero at the Nyquist frequency 1/ _2p0 , which is the second filter means 6.
The MTF is sufficiently large near the frequency 1/4p ₀ , which is half the Nyquist frequency, and then decreases continuously to the lower and higher order sides, and becomes almost zero at the low-order frequency and the Nyquist frequency 1/2p _0. . Therefore, the first filter means 5 controls higher-order spatial frequency components than lower-order spatial frequency components and mainly passes the lower-order spatial frequency components, whereas the second filter means 6 conversely controls higher-order spatial frequency components. In comparison, lower-order spatial frequency components are suppressed and higher-order spatial frequency components are mainly passed. The filter selection means 7 selectively selects the output of the first filter means 5 and the output of the second filter means 6 and sends it to the memory means 9 via the A/D converter 8. This memory means 9 stores the photoelectric outputs a ₁ , b ₁ ,
a ₂ , b ₂ . . . has a storage capacity capable of storing all outputs of the A/D converter 8 corresponding to a _o , b _o at one time. The calculation means 10 converts the filtered outputs A _{1 .}
……A(i)…A(L), and the other filtered output B ₁ ……B _o is similarly divided into multiple areas B(1),
B(2)...Divided into B(j) and B(L), the consistency is calculated for multiple pairs of A(i) and B(j) in these areas, and the correlation degree with the best consistency is calculated. By knowing the high combination (i, j), the focus detection signal Zi representing the relative displacement amount of the first and second optical images on the first and second photoelectric conversion element arrays, that is, the amount of defocus
and an information amount signal Di representing the certainty of the information used in the calculation. The sign of this signal Zi indicates the distinction between front focus and rear focus, and its absolute value indicates the absolute value of the amount of defocus, and the signal Di indicates that it is a predetermined value.
When it is greater than or equal to Do, it guarantees that the focus detection signal is sufficiently reliable. A specific example of this calculation means is disclosed in Japanese Patent Laid-Open No. 57-45510.
This focus detection signal Zi and the information amount signal Pi
45510 (expressed as Dm) is input to the control means 11. This control means 11 compares whether the absolute value of the focus detection signal Zi is larger than a predetermined value Zo, and when it is larger, the filter selection means 7 selects the first filter means 5.
, and a second filter means selection signal which selects the second filter means 6 when the second filter means 6 is selected. Next, this effect will be explained. When the imaging optical system 1 forms first and second subject images on the small lens arrays 2, the first photoelectric conversion element array ₁ ... _o converts the illuminance distribution pattern of the first subject image into a photoelectric output _a1 ... …Convert to a _o and do the same with the second
The photoelectric conversion element arrays b ₁ and b _o convert the second optical image into photoelectric outputs.
b ₁ ...Convert to b _o . The photoelectric device 4 converts the photoelectric outputs of the first and second photoelectric conversion element arrays into a ₁ , b ₁ , a ₂ , b ₂
...Alternately output in chronological order like a _o and b _o . This series of photoelectric outputs is repeatedly output at certain time intervals. If the selection means 7 is the second filter means 6
is selected, the above series of photoelectric outputs
a ₁ , b ₁ ...a _o , b _o are subjected to filtering processing by the second filter means 6, and the filtered outputs A ₁ ,
B ₁ ... After being converted into A _o and B _o , the A/D converter 8
The data is stored in the memory means 9 via. Arithmetic means 1
0 calculates a focus detection signal Zi and an information amount signal Di representing the accuracy of information used in the calculation based on this stored content. From now on, when selecting the first filter means and when selecting the second filter means,
When selecting the filter means, the signals Zi and Di are respectively Zi(1) and
They are expressed as Di(1), Zi(2), and Di(2). The control means 11 first determines whether Di(2) has reached a predetermined level Do or not and whether the absolute value |Zi(2)| of the focus detection signal is greater than a predetermined value Zo. If Di(2)〓Do and |Zi(2)Zo, the information is valid, so control means 1
1 controls the display device 12 and the drive device 13 based on this signal Zi(2) and continues to output the second filter means selection signal. As a result, the display device 12
and drive device 13 each display the focus adjustment state and drive the imaging optical system 1 toward the in-focus position. On the other hand Zi
(2), Di(2) other than the above, the accuracy of the information is low, and the first and second subject images are largely blurred, with few high-order spatial frequency components and relatively low-order spatial frequency components. Since the number of filters has increased, the control means 11 sends the first filter selection signal to the selection means 7 without displaying or driving the current information. Due to the selection of the first filter means 5 by the selection means ₇ , the next generated series of photoelectric outputs a ₁ , _{b 1 .} The calculation means 10 generates the focus detection signal Zi(1),
An information amount signal Di(1) is calculated. If Di(1)=Do, the control means 11 controls the display device 12 and the drive device 13 based on the signal Zi(1). When the first filter means 5 is selected, which filter means to select next can be determined in various ways depending on how the imaging optical system is driven. For example, the imaging optical system 1 is stopped, an image shift detection calculation is performed, the imaging optical system is moved by the amount of def focus calculated based on the result, and then stopped, the image shift detection calculation is performed again, and the result is also Let us consider the case where intermittent driving of the imaging optical system is performed, in which the imaging optical system is driven again after a short time. In this case, if the first filter means 5 is selected and the signal Zi(1) is effectively determined, the imaging optical system is driven based on this result to ensure that the image is close to being in focus the next time the image shift is detected. Therefore, it is appropriate to select the second filter means 6 next time. Therefore, in the case of such intermittent driving of the imaging optical system, if the first filter means is selected at a certain time, the first filter means will always be selected next time, regardless of the contents of the calculation results Zi(1) and Di(1) at that time. It is appropriate to select two filters. On the other hand, if the imaging optical system drive and focus detection are performed in parallel, Di(1)〓Do and |Zi(1)|＞
In the case of Zo, select the first filter next time,
In the case of other Di and Zi, it is better to select the second filter means next time. In other words, in the case of Di(1)〓Do and |Zi(1)|Zo, it is close to convergence, so it is better to select a second filter centered on high-order spatial frequency components, and Even when Di(1)<Do, depending on the subject, there may be only high-order components without low-order spatial frequency components, so the result of using the first filter is insufficient information (Di(1)<Do) Even if it is, it can be detected by using the second filter (Di(2)〓Do)
This is because in some cases, it is appropriate to switch to the second filter. As described above, the control means 11 causes the selection means 7 to select either the first filter means 5 or the second filter means 6. As described above, when the imaging optical system is located near the focusing position, that is, the amount of defocus is small and there are many high-order spatial frequency components in the first and second optical images, the low-order spatial frequency components are suppressed. Since the second filter means 6 is automatically selected, near the in-focus position,
High-precision focus detection based on high-order spatial frequency components is possible without being affected by low-order spatial frequency components, and conversely, since the imaging optical system 1 is far away from the focus position, When the spatial frequency component decreases and the low-order spatial frequency component relatively increases, the first filter 5 that passes the low-order frequency component is automatically selected, and focus detection based on the low-order spatial frequency component is performed. Note that the filter selection means 7 is not limited to the illustrated position, and may be arranged between the output terminal of the photoelectric device 4 and the input terminals of the first and second filter means 5 and 6, and may also be arranged between the output terminal of the photoelectric device 4 and the input terminals of the first and second filter means 5, 6. It may be placed later.
In this latter case, the series of photoelectric outputs are processed by first and second filter means 5, 6, and arithmetic means 1
0 is a focus detection signal Zi(1) obtained by processing the outputs of the first and second filter means 5 and 6, respectively, and based on the former.
and the focus detection signal Zi(2) based on the latter, and then the selection means 7 may alternatively select both signals Zi(1) and Zi(2). In this way, in the present invention, selectively selecting the first filter means and the second filter means means that the output of the first filter means and the output of the second filter means to be input to the calculation means 10 are selected as shown in the example shown in the figure. This includes a case in which the output of the calculation means is alternatively selected, and a case in which the output of the calculation means is selected based on the output of the first filter means and the output of the calculation means based on the output of the second filter means. Specific configuration examples of the first and second filter means 5 and 6 are shown in FIGS. 4b and 5b, respectively. In FIG. 4b, the delay circuit D ₁ for one pixel,
D ₂ and D ₃ are connected in series, and an adder circuit T ₁ adds the outputs of delay circuits D ₁ and D ₃ . As a result, time-series photoelectric output a ₁ , b ₁ , a ₂ , b ₂ , a ₃ , b ₃ from the photoelectric device 4
... are sequentially input to delay circuits D ₁ to _{D 3} , adder circuit T ₁ sequentially inputs (a ₁ + a ₂ ), (b ₁ + b ₂ ), (a ₂ + a ₃ )
Outputs... The configuration shown in Figure 5b is a series-connected one-pixel delay circuit.
Difference between the outputs of the first and fifth delay circuits D ₅ and D 1 of D ₁ , D ₂ , D ₃ , D _{4 ,} D ₅ (a ₁ −a ₃ ), (b ₁ −b ₃ )...
... is determined by the subtraction circuit _S1 . In addition, if the first
When the output of the second photoelectric conversion element array is read out after all the outputs of the photoelectric conversion element array are read out, in other words, the time-series photoelectric output of the photoelectric device 4 is
When a ₁ , a ₂ ... _a o , b ₁ , b ₂ ... b _o , the specific configuration example of the first and second filter means 5 and 6 is the fourth
It will look like Figure c and Figure 5c. Of course, instead of constructing the filter means 5 and 6 with hardware as described above, they can be constructed using a microcomputer or the _like .
a ₂ ), (a ₁ −a ₃ ), etc. may be performed. Further, the first filter means 5 of this embodiment is
As can be seen from the comparison between Figure 4a, which shows the MTF characteristics, and Figure 2, which shows the MTF characteristics of the original signal, filtering processing is performed to suppress the high-order spatial frequency components of the photoelectric output. Since the function of the filter 5 is to pass at least low-order spatial frequency components, a filter with an infinite frequency band is used as the filter 5, for example. in particular,
The photoelectric output may be directly connected to the A/D converter 8 via the selection means 7. In such a case, the first filter means 5 is equivalent to not substantially performing filtering processing, but in this specification, it is said that the first filter means 5 has a frequency band different from the MTF frequency band of the second filter means 6. From this point of view, such virtual filter means that does not substantially perform filtering processing is also included in the concept of filter means. Next, an example will be described in which the above-mentioned virtual filter means is used as the first filter means and the filtering process of the second filter means is executed within the calculation means 10. In this case, the photoelectric output is directly input to the A/D converter 8 as described above. Therefore the memory means 9
Data having the band shown in FIG. 2 is stored in .
The calculation means 10 performs a correlation calculation on the data stored in the memory means 9 and generates a focus detection signal Zi(1) and an information signal.
Calculate Di(1). Here, if Zi(1)<Zo, the calculation means 10 calculates every other difference data A _i =a _i −a _i+2 ……B _j =b _j −b _j+2 from the data already stored in the memory.
...and calculates this filtered output A ₁ , A ₂ ...
A _o-1 , B ₁ , B ₂ ...B _o-1 is obtained, and the correlation calculation is performed again on this filtered output to obtain the focus detection signal Zi(2) and the information amount signal Di(2). . At this time, the control means 11 controls the focus detection signal Zi when Zi(1)=Zo.
Based on (1), and when Zi(1)<Zo, the focus detection signal
The display device 12 and the drive device 13 are each controlled based on Zi(2). Regarding the next series of photoelectric outputs, in exactly the same way as above, the calculation means 10 initially receives the signals Zi(1),
Di(1) is calculated, and when this signal Zi(1) is smaller than a predetermined value Zo, filtering processing is performed to extract high-order spatial frequency components, and then signals Zi(2) and Di(2) are calculated. As is clear from the above explanation, the method of selecting the filter means is as shown in the embodiment of FIG. The means may be selected alternatively, or the second filter means may be selected exclusively after the first filter means, or, as in the last example above, the first filter means (the upper In the example, this first filter means was a virtual filter means), and based on the calculation result Zi(1) at this time, it is determined whether or not to select the second filter means, etc., and settings are made as appropriate. However, regardless of the selection method, if the imaging optical system is near the focus position and there are sufficient high-order spatial frequency components, the filtered output of the second filter means must be used. Focus detection signal Zi(2)
It is necessary to determine that the display and imaging optical system is driven using the following. Next, the characteristics of the plurality of filter means will be explained. As a combination of multiple filter means,
As in the first embodiment, the characteristics shown in Fig. 4 a are mainly extracted from the DC component, and the characteristics shown in Fig. 5 a are that the DC components are removed and mainly the high order spatial frequency components are extracted. Although the combination with those having MTF characteristics has been described, various combinations are possible as shown in FIGS. 6a to 6c. MTF of the filter means shown in solid line in Figure 6a
The characteristics are such that high-order spatial frequency components are suppressed and relatively low-order spatial frequency components are transmitted.
The MTF characteristics of the filter means shown by the dashed dotted lines are determined to pass spatial frequency components from low order to high order. Both filters contain everything from DC components to low-order spatial frequency components,
When the subject has few high-order spatial frequency components and many low-order spatial frequencies, the correlation calculation through both filters is performed based on the low-order spatial frequency components, so in reality, the influence of the accompanying error factors is unavoidable. As a result, the focus detection error increases. In the present invention, as described in the first embodiment, the second filter removes the DC component and mainly extracts the high-order spatial frequency components, so that the high-order spatial frequency components are small and the low-order spatial frequency components are large. Highly accurate focus detection is possible even for objects. The first and second filter means shown in FIG. 6b
The MTF characteristics are similar to those in Figures 4a and 5a, respectively, and the difference is that the first filter means in Figure 6b suppresses higher spatial frequency components more than in Figure 4a. That's true. This makes it possible to prevent false focusing signals from occurring as in FIG. 6a. That is, in this case, in the vicinity of the focus, the second filter means shown in FIG. 6b is used to remove low-order spatial frequency components, and focus is determined using a signal mainly consisting of high-order spatial frequency components. Therefore, focus detection accuracy is greatly improved, and in areas with large defocus, by using the first filter shown in Figure 6b, high-order spatial frequency components that can cause false focusing are removed, and low-order spatial frequency components are removed. Since the amount of defocus is determined based only on the component, this is an ideal combination of filters that allows front and rear focus to be determined without causing false focusing even when the lens is far from the in-focus position. The MTF characteristic of the first filter means in FIG. The MTF has a peak on the higher frequency side than that of the first filter means. In this way, the first and second
Setting the filter means to suppress the DC component has the following advantages. That is, there may be a difference between the sensitivity and DC component of the photoelectric output of the first photoelectric conversion element array ₁ ... _o and the sensitivity and DC component of the output power of the second photoelectric conversion element array ₁ ... _o . This results in a decrease in focus detection accuracy. However, suppression of the DC component described above can eliminate the effect of such a difference. Note that the above-mentioned difference in sensitivity and DC component is due to, for example, a difference in sensitivity between each array that occurs when the first and second photoelectric conversion element arrays are formed on separate chips, a difference in temperature drift between each array, etc. arising based on. In the previous explanations, we have discussed cases in which the MTF band mainly consists of high-order spatial frequency components and cases in which it mainly consists of low-order spatial frequency components.
Essentially, the upper limit of the effective band is the Nyquist frequency 1/2P ₀ determined by the pitch P ₀ of the small lens array 2, so the above discussion of higher and lower orders is also within the frequency band below this Nyquist frequency. It is. A second embodiment of the invention will be described below. In FIG. 7, a field lens 15 is placed near the intended focal plane of the imaging optical system 1, such as a photographic lens.
The field lens 15 has a rectangular light transmitting area 15a at its center, and the area other than the area 15a is a light shielding area. The substantially rectangular parallelepiped transparent block 16 is made of a high refractive index material such as glass or plastic, and the field lens 15 is attached to one end surface 16a. On the other end surface 16b opposite to this one end surface 16a, there is a pair of concave mirrors 17, 1 slightly inclined in opposite directions.
8 is provided. Both end surfaces 16a, 16b
A pair of mirrors 19 and 20 are diagonally disposed at an angle of approximately 45° in the block 16 between the two, with a predetermined gap therebetween. Photoelectric conversion devices 21 are arranged below the transparent blocks 16, respectively. In this photoelectric conversion device 21, photoelectric conversion element arrays 22 and 23 are formed below the mirrors 19 and 20, respectively. The light beam that has passed through the imaging optical system 1 passes through the light transmission area 15a of the field lens 15 and enters the block 16.
The light passes through the gap between the mirrors 19 and 20 and enters the pair of concave mirrors 17 and 18. One concave mirror 17 reflects the incident light toward the mirror 19, and the other concave mirror 18 reflects the incident light toward the mirror 20.
Each reflected light reaches photoelectric conversion element arrays 22 and 23 via mirrors 19 and 20, respectively. In this way, a pair of subject images of substantially the same subject are formed on the arrays 22 and 23. A circuit system for processing the photoelectric output from this photoelectric device 21 will be explained with reference to FIG. In FIG. 8, the photoelectric device 21 is a CCD image sensor, and the first photoelectric conversion element array 22
It includes at least a second photoelectric conversion element array 23, a transfer gate 24, and transfer sections 25 and 26, and may also include an amplifier for linearly or logarithmically amplifying the photoelectric output. Of course, the photoelectric device may be a MOS image sensor or other structure. _The photoelectric conversion _{elements 1} . . This photoelectric device 21 includes a first array 2
Similarly to the first embodiment, the photoelectric outputs a ₁ ...a _o of the second array 23 and the photoelectric outputs b ₁ ... _bo of the second array ₂₃ are
b ₁ , a ₂ , b ₂ ...a _o , b _o are output alternately, and this series of photoelectric outputs a ₁ , b ₁ ... a _o , b _o
is output repeatedly at predetermined time intervals. Such photoelectric conversion element arrays 22 and 23 each have MTF characteristics as shown in FIG. 9a. The input terminals of the first and second filter means 27 and 28 are connected to the output terminal of the photoelectric device 21. As shown in FIG. 9b, this first filter means 27 has an MTF characteristic that allows low-order spatial frequency components to pass through, but sufficiently suppresses high-order spatial frequency components with a frequency of around 1/8P _{0 or} higher. The filter means 28, as shown in FIG. 9c, has an MTF characteristic that allows low-order spatial frequency components to pass through, but sufficiently suppresses high-order spatial frequency components having a frequency around 1/4P ₀ or higher. In this way, the second filter means 28 is designed to pass even higher-order spatial frequency components than the first filter means 27. Furthermore, the 9th
As can be seen by comparing Figures b and c with Figure 4a or Figure 5a, the first and second filter means 2 of this embodiment
The edges of the MTF frequency band of 7 and 28 are 1/8P ₀ and 1/4P ₀
As shown, the frequency band edge 1/2P ₀ of the first embodiment is shifted to the lower frequency side. The selection means 29 is the first
Same as in the embodiment, the outputs of the first and second filter means 27 and 28 are alternatively selected and sent to the sample hold means 30. This sample and hold means 30 includes a first stage sample and hold circuit 30A and a second stage sample and hold circuit 30B which are connected in series.
It is composed of. A/D converter 31 A/D converts the output of sample hold circuit 30B. The memory means 32 are the same as in the first embodiment. An example of the calculation contents of the calculation means 33 will be briefly described below. The filtered output for one photoelectric conversion element array is divided into multiple areas A(1), A(2), A(3)...A(i).
...Divided into A(L), and the filtered output regarding the other photoelectric conversion element array is also correspondingly divided into B(1) and B.
(2), B(3)...B(j)...B(L), calculate the consistency of multiple pairs of A(i), B(j) in these areas, and find the most consistent one. The amount of relative displacement of the optical images on both arrays is calculated by using the values of the combination (i, j) with good quality and the combinations in its vicinity, and the amount of defocus Zi is calculated. For example, if the number of components in each area is equal (M+1), then A(i) = {Ai, A _i+1 , A _i+2 , ... A _i+M } B(j) = { Bj, B _j+1 , B _j+2 , ...B _j+M }. The correlation amount representing the degree of consistency is l=i for the case where the image is shifted by l data positions.
−j, and let [x] represent the largest integer not exceeding x: C(l)= _M 〓 〓 ⁿ⁼⁰ |A _i+M −B _j+n |; (i=[L+l+1/2] ,j=
i-l) is given by Let this correlation amount C(l) be l=-(L-1),...-
Calculations are performed for each shift amount of 1, 0, 1, . . . (L-1) to find the position of maximum correlation, that is, the shift amount l=l ₀ at which C(l) is the smallest value. When l ₀ is not equal to the value at both ends (L-1 or -L+1), the fraction Δl ₀ of the amount is calculated using the following formula, Δl ₀ = 1/2 (C(l ₀ +1) - C(l ₀ - 1))/(2×C(l ₀ )−C(l ₀ +1)−C(l ₀ −1)) can be extrapolated. The deviation amount including the fraction obtained in this way l ₀ +
Focus detection signal Zi representing the amount of defocus from Δl ₀
is required. In the second embodiment, as will be described later, the sampling pitch is different when using the first filter means and when using the second filter means, and the above deviation amount l ₀ +
The proportionality constant when calculating the defocus amount from Δl ₀ is different. Furthermore, since the number L of the plurality of divided regions is not necessarily the same in the case of selecting the first filter means and in the case of selecting the second filter means, the calculation means 3
The contents of the calculation in step 3 will differ somewhat depending on the filter means selected. This can be seen in Fig. 8.
4 to the filter means selection signal 34a to the calculating means 3.
3, it will be identified and some different calculations will be performed. Further, as the information amount signal Di representing the amount of information being output from the first and second filter means, for example, the difference between the maximum and minimum output data of the filter means may be used, but the correlation amount C (l)
You can also take the difference between the maximum value C(l) _MAX and the minimum value C(l) _MIN . Also, considering the case where the amount of image shift does not fall within the range of image shift determination area - (L-1) l (L-1), in this case also C(l)
The value of is confusing because it is minimum at a certain value of l within the above range. However, in such a case, C(l) _MIN /(C(l) _MAX - C(l) _MIN ) is not as small as when the amount of image shift is within the judgment area, so an appropriate threshold value C _TH must be set. can be set and excluded. That is, C(l) _MIN /(C(l) _MAX −C
When (l) _MIN )>C _TH , the information amount signal Di, which normally takes a positive value, is given a zero or negative value to be excluded as uncorrelated. The determining means 34 inputs the focus detection signal Zi and the information amount signal Di, and
Information amount signal Di when selecting the first and second filter means
(1) and Di(2) are smaller than the respective predetermined values Do(1) and Do(2), the selection means 29
If currently the first filter means 27 is selected, the second filter selection signal, specifically the H level output, which causes the second filter means 28 to be selected, conversely, the second filter means 28 is currently selected. If so, a first filter selection signal, specifically an L level output, for selecting the first filter means 27 is generated at the output terminal 34a, and on the other hand, an information amount signal Di(1) is generated.
Or, if Di(2) is greater than the corresponding predetermined amount Do(1), Do(2), the absolute value of the focus detection signal Zi(1) or Zi(2) is equal to the corresponding predetermined value Zo(1), Zo (2) When the first filter selection signal is greater than the predetermined value Zo(1), Zo(2) or less,
The second filter selection signal is outputted to the respective output terminals 34a.
occurs in Further, this determining means 34 generates a memory update signal at the output terminal 34b when the information amount signals Di(1), Di(2) are equal to or greater than the predetermined amounts Do(1), Do(2).
In response to this memory update signal, the memory circuit 35 stores the focus detection signal Zi at that time. According to the focus detection signal Zi stored in the memory circuit 35, the display device 36 displays the focus adjustment state, and the drive device 3
7 drives the imaging optical system 1 toward the in-focus position. The sample pulse generation circuit 38 is connected to the output terminal 34a of the discrimination means 34, and the sample hold circuit 3
A sample pulse to start sample hold is supplied to 0A and 30B. The period of this sample pulse changes according to the output of the discrimination means 34, and the period when it is the first filter means selection signal is larger than when it is the second filter means selection signal, and in this embodiment, the period is greater than the period when it is the second filter means selection signal. It has been selected twice. The sample pulse generation circuit 38 has a first counter 39.
When a start signal, specifically an H level signal, is received from the output terminal 39a of the second counter 40, the generation of the sample pulse is started, and the second counter 40 outputs the output terminal 40 of the second counter 40.
Upon receiving the termination signal from a, specifically the H level signal, the generation of the sample pulses is stopped.
This first counter 39 is a presettable counter that presets a preset value sent from the setting section 41 via the gate means 42, and
The pulse output from the AND gate 43 is counted down, and when the content reaches zero, an H level start signal is output. The second counter 40 is also a presettable counter, and is preset to a preset value from the setting section 41 via the gate means 44, and also counts down the sample pulse to the subsequent sample and hold circuit 30B until the content becomes zero. At this time, an H level end signal is generated. The setting section 41 sets a first preset value for the first counter, a first preset value for the second counter, which is used when selecting the first filter means 27, and a second preset value for the first counter, which is used when selecting the second filter means. The value and the second preset value for the second counter are stored in advance, and the first preset value for the first counter and the first preset value for the second counter when the first filter means is selected are output to the output terminals 41a and 41c, respectively. The second preset value for the first counter and the second preset value for the second counter when the second filter means is selected are output to the output terminals 41b and 41d, respectively. In this example, the first preset value of output terminal 41a is less than the second preset value of output terminal 41b, and the first preset value of output terminal 41c is set equal to the second preset value of output terminal 41d. An H level signal is input to the input terminal 45 from a sequence control unit (not shown) in synchronization with the start of transfer of a series of photoelectric outputs a ₁ , b ₁ . . . a _o , b _o from the photoelectric device 21 . This signal is reset to the L level at an appropriate time from the end of sampling and holding all data until the next preset value is set in the preset counters 39 and 40. A clock synchronized with the transfer clock for transferring the series of photoelectric outputs is input to the input terminal 46. This effect will be explained below. Assume that the determining means 34 generates an L level output, which is the first filter means selection signal, at the output terminal 34a. In response to this selection signal, the selection means 29 selects the first filter means 27, and the gate means 42 and 44 select the first counter preset value and the second counter preset value from the output terminals 41a and 41c of the setting section 41. input to the first counter 39 and the second counter 40, respectively, and preset each counter to its preset value. After that, a series of photoelectric outputs a ₁ , b ₁ , a ₂ , b ₂ ...a _o ,
b _o is read. This series of photoelectric outputs a ₁ , b ₁ ...
. . a _o , b _o , the photoelectric outputs a ₁ , a ₂ . . . a _o from the first photoelectric conversion element array 22 are shown in FIG. 10a. The series of photoelectric outputs a ₁ , b ₁ ...a _o , b _o are filtered by the first filter means 27, and the first
Filtered outputs A ₁ , B ₁ ...A _o , B _o shown in Figure 1 a
is converted to These filtered outputs A ₁ , B ₁ ...
Among A _o and B _o , those related to the first photoelectric conversion element array A ₁ , . . . A _o are shown in FIG. 10b. Comparing FIG. 10b and a, the first filter means 2
7 is clearly effective in suppressing high-order spatial frequency components. On the other hand, since an H level signal is input to the input terminal 45 in synchronization with the reading from the photoelectric device 21, the AND gate 43 outputs the transfer clock from the input terminal 46. The first counter 39 subtracts the number of transfer clocks from the preset value and generates an H level output as a start signal when the number of input clocks becomes equal to the preset value. This start signal is input to the sample pulse generation circuit 38, and is inverted and input to the AND gate 43, which closes the gate. The sample pulse generation circuit 38 supplies sample pulses SP1 and SP2 for the front and rear stages shown in FIGS. 11b and 11c to the front and rear sample hold circuits 30A and 30B, respectively, in response to the start signal. The pre-stage sample and hold circuit 30A outputs filtered outputs A ₁ , B ₁ . . . according to the pre-stage sample pulse SP1.
...A _o , B _o to A ₄ , B ₄ , A ₈ , B ₈ , A ₁₂ , B ₁₂ ...
to sample. This pre-stage sample and hold circuit 30A outputs outputs A _{4 , 1 and 2} related to the first photoelectric conversion element array, as shown by the arrows in FIG. 11b.
A ₈ ... is held for a short time, and outputs B ₄ , B ₈ ... related to the second photoelectric conversion element array are held for a relatively long time. In order to equalize the holding time of both, the subsequent sample and hold circuit 30B samples and holds the output of the previous sample and hold circuit 30A in response to the subsequent sample pulse SP2. The second counter 40 counts the second stage sample pulse SP2,
When it becomes equal to the first preset value, a termination signal is generated and the front and rear sample pulses SP1,
Stop the generation of SP2. In FIG. 10b, sampled filtered outputs _{A 4 ,} A ₈ , A _{12 .} . . of the filtered output A 1 . _S is attached.
As can be seen from this figure, the distribution range (hereinafter referred to as sampling area) of the sampled filtered output l ₂ occupies most of the range of the filtered output A ₁ ... A _o . The A/D converter 31 A/D converts the output of the subsequent sample and hold circuit 30B and sends it to the memory means 32. The reason for providing the latter stage sample and hold circuit 30B is as follows. If the output of the preceding sample and hold circuit 30A is directly A/D converted, the outputs _{A 4} , _{A 8 .} . .
Since the retention time of ... is short, the A/D conversion operation is completed within the shorter retention time.
An expensive high-speed A/D converter must be used as the converter 31. Furthermore, even if a high-speed A/D converter is used, the A/D of outputs B ₄ , B ₈ . . .
In D conversion, the advantage of high speed is not utilized. However, the subsequent sample and hold circuit 30
The use of B eliminates the above problem. As shown in FIGS. 11d and 11e, the sampling period when the second filter means is selected is smaller than that when the first filter means is selected. Since it is shorter when selecting the means, A/D converter 3
The time required for conversion 1 is determined by the holding time when the second filter means is selected. Naturally, when the first filter means is selected, the holding time of the subsequent sample and hold circuit becomes unnecessarily long compared to the time required for conversion. In order to avoid this waste, the frequency of the photoelectric output transfer clock when the first filter means is selected is made higher than when the second filter means is selected, and the holding time of the subsequent sample and hold circuit 30B when both are selected is made equal. . The calculation means 33 calculates the filtered output stored in the memory means 32 and outputs a focus detection signal Zi(1) and an information amount output Di(1). The determining means 34 compares the signals Zi(1) and Di(1) with corresponding predetermined values Zo(1) and Do(1). (a) When Di(1) is greater than or equal to Do(1) In this case, the determining means 34 generates a memory update signal to the output terminal 34b, and stores the focus detection signal Zi(1) at this time in the memory circuit 35. be done. The display device 36 and the drive device 37 receive this stored signal Zi.
Based on (1), the focus adjustment state is displayed and the imaging optical system 1 is driven to the focusing position. Further, the discrimination means 34 continues to output the first filter means selection signal to the output terminal 34a when the signal Zi(1) is larger than the predetermined value Zo(1). Therefore, at this time, the photoelectric device 21 further outputs a series of photoelectric outputs a ₁ , b ₁ .
When a _o and b _o are output, this entire circuit performs the same operation as described above. When the signal Zi(1) is less than the predetermined value Zo(1), the determining means 34 outputs an H level output, which is the second filter means selection signal, to the terminal 34a. In response to this second filter means selection signal, the selection means 29 selects the second filter means 28, and the gate means 42 and 44 are connected to the output terminal 4 of the setting section 41.
2nd for the 1st and 2nd counters from 1b and 41d
The preset values are transferred to the first and second counters 39, 39, respectively.
Send to 40. Thereafter, the series of _{photoelectric} outputs a ₁ , b _{1 .}
A ₁ , B ₁ ...converted to A _o , B _o . The filtered outputs A ₁ , A ₂ . . . A _o regarding the first photoelectric conversion element array at this time are shown in FIG. 10c. Comparing this Figure 10c and Figure 10b, Figure 10c
It can be seen that the figure is less smooth and that the second filter means 28 passes higher-order spatial frequency components than the first filter means 27. On the other hand, the first counter 39 counts the output transfer pulses of the AND gate 43 in synchronization with the reading of the photoelectric output,
When the counted value matches the second preset value, a start signal is generated. Since the second preset value for the first counter when the second filter means is selected is set to be larger than the first preset value for the first counter when the first filter means is selected, the start signal generation point at this time is , 1st
This is later than when the start signal is generated when the filter means is selected. This start signal causes the sample pulse generation circuit 38 to
Sample pulse SP3 for the front and rear stages shown in
Generates SP4. This sample pulse SP3,
The period of SP4 is shorter than the sample pulses SP1 and SP2 when selecting the first filter means, and is set to 1/2 in this embodiment, according to the second filter means selection signal sent from the determining means 34. Therefore, as shown in FIGS. 11d and 11e, the front and rear sample and hold circuits 30A and 30B output the filtered outputs A ₁ , B _{1 .} Sample B _o and output A ₈ , B ₈ , A ₁₀ , B ₁₀ , A ₁₂ , B ₁₂ . The second counter 40 counts the second-stage sample pulse SP4, and when the counted value matches the second preset value, generates an end signal and stops the generation of the sample pulses SP3 and SP4. Since this second preset value for the second counter is determined to be equal to the first preset value for the second counter when the first filter means is selected, the filtered output A ₈ sampled when the second filter means is selected, _B8 , _A10 , _B10 ...
The number is equal to that when the first filter means is selected. Among the filtered outputs sampled in this way, those related to the first photoelectric conversion element array are indicated by a mark Ms in FIG. 10c. In this embodiment, the sampling period and the number of samples when selecting the first filter means are respectively 1/2 and the same as when selecting the second filter means, so as shown in FIG.
The sampling area l ₁ shown in FIG. 10c is twice as large as the area l ₂ in FIG. 10c. Of course, the number of samples in both cases does not necessarily have to be equal. In addition, the first
For convenience of drawing, the graphs in Figure 1 d and e are drawn at an earlier point in time when sampling begins. The sampled output is sent to the calculation means 33 via the A/D converter 31 and the memory means 32 and is calculated. Since the filtered output at this time contains more high-order spatial frequency components than when the first filter means is selected, the focus detection signal Zi(2) when the second filter means is selected has a higher frequency near the focus position. It has become accurate. Di(2)〓Do
(2) and |Zi(2)|Zo(2), the determination means 34
outputs the second filter means selection signal to the terminal 34a.
At the same time, a memory update signal is sent to the output terminal 34b, and the focus detection signal Zi(2) at this time is stored in the memory circuit 35. Display and imaging optical systems are driven according to the stored contents.
If Di(2)〓Do(2) and |Zi(2)|>Zo(2), then
The determination circuit 34 outputs a first filter means selection signal. (b) When Di(1) or Di(2) is smaller than Do(1) or Do(2). In this case, regardless of the focus detection signal Zi, the determining means 34 selects the second filter means selection signal if the first filter means is selected, and conversely, selects the first filter means if the second filter means is selected. The signals are respectively output to the output terminals 34a.
As a result, the selection means 29 switches the filter means to be selected. Furthermore, since the focus detection signal Zi when this signal Di is smaller than the predetermined value Do is extremely low in accuracy, the determining means 34 does not generate a memory update signal. Therefore, the signal Zi at this time is not used for driving the display and imaging optical system. Furthermore, the signal Zi
The switching of the filter means which is independent of the filter means is carried out for the following reasons. That is, for example, when the subject contains almost no low-order spatial frequency components and a large amount of high-order spatial frequency components, the second filter means 2
This is because necessary information can be obtained by selecting item 8. In this embodiment, when selecting the first filter means,
That is, when the amount of defocus is large and the amount of relative deviation between the subject images on the first and second photoelectric element arrays is large, the sampling area _l1 is widened as shown in FIG. When the amount of deviation is small, the sampling area _l2 is narrowly defined as shown in FIG. 10c. This is very effective for focus detection. That is, by widening the sampling area, even if the subject image shifts relatively significantly, the shift can be detected. Therefore, the amount of defocus can be detected even if the photographing lens is far away from the in-focus position. On the other hand, widening the sampling area increases the possibility that objects at different distances or objects with depth will enter the sampling area. For focus detection when the amount of defocus is large, it is sufficient to determine whether the focus is from the front or the back, or to determine the approximate amount of defocus, and accurate measurement of the absolute value of the amount of defocus is not necessarily necessary. Even if it exists in the sampling area, the effect is small. However, when the amount of defocus is small and its absolute value must be accurately measured, the presence of the object with a certain depth tends to cause a large error in the measurement. Therefore, when selecting the second filter means that requires high-precision focus detection, high-order spatial frequency component information is used to increase detection accuracy, and the sampling area is narrowed to reduce the possibility that a deep subject will enter there. There is. In general, it is not necessarily necessary to increase the sampling period just because the sampling area is widened; for example, the area of l ₁ related to the output in Fig. 10b is set to be smaller than the sampling pitch 4P ₀ shown in the figure, and the sampling period is set to P ₀ or 2P _0. Of course, you can sample at the same pitch. However, reducing the sampling pitch to P ₀ or 2P ₀ is not very preferable since the number of samples is quadrupled and doubled, respectively, resulting in a significant increase in the storage capacity of the memory means 32 and the calculation scale of the calculation means 33. Therefore, even when the sampling area is changed as in this embodiment, it is extremely effective to keep the number of samples approximately the same. In this way, when selecting the first filter means that passes only low-order spatial frequency components, the sampling period is set to 4P ₀ , and when selecting the second filter means that passes high-order spatial frequencies, the sampling period is set to 2P ₀ . This is extremely advantageous in terms of information utilization. Here, to explain in detail the relationship between the sampling pitch and the MTF characteristic of the filter, the sampling period was set to 4P ₀ when the first filter was selected, so the Nyquist frequency at this time was 1/8P ₀ . According to the sampling theorem, spatial frequency components with a frequency of 1/8 P ₀ or higher may cause malfunctions, so it is desirable to remove them. As shown in FIG. 9b, the MTF of the first filter means sufficiently suppresses the components above the Nyquist frequency near 1/8P ₀ and passes the components below it, so the passed components can be used effectively. can. However, if the sampling period is set to P ₀ when the first filter means is selected, the Nyquist frequency at this time becomes 1/2P ₀ , and spatial frequency components below this frequency can be used for focus detection. However, as shown in FIG. 9b, components with a frequency of 1/8 P ₀ or more are removed by the first filter means, so the sampling period P ₀ is 4 times smaller than the sampling period 4P ₀ . Even if the number of samples is increased by a factor of two, the amount of available spatial frequency components remains the same, and the increase in the number of samplings is completely wasted. As is clear from the above, from the viewpoint of effective use of information, it is desirable to determine the sampling period so that the Nyquist frequency determined by the period is near the end of the MTF frequency band of the filter means. Furthermore, if the preset value of the setting means 41 can be changed from the outside, the extent of the sampling area, that is, the subject area used for focus detection can be arbitrarily varied, and it is possible to make it possible for a deep subject to fall within the above area. It can be prevented. Next, a specific example of the configuration of the block shown in FIG. 8 will be explained. FIG. 12 shows an example of the filter means 27, 28, in which delay circuits D ₁ , D ₂ ...D _n for one pixel are connected in series, and delay circuits D ₁ , D ₃ , D ₅ ...D _n are connected to multipliers W ₁ . . . _WS via amplifiers A _n , respectively. These multipliers W ₁ ... _WS multiply the inputs by weights W ₁ ... _WS , respectively. This weight can be a positive or negative number. Adder circuit _T2 adds the outputs of each multiplier. When a series of photoelectric outputs from the photoelectric device 21 are sequentially input to the delay circuit _D1 , the addition circuit
Filtered output is output from T ₂ . Although there are many ways to take the weight W _{1 .} . . _WS that provides a predetermined MTF characteristic and it cannot be determined uniquely, some specific examples are shown below. To obtain the filter means with the MTF characteristic as shown in FIG. 9c, set D _n = D ₉ and W _S = W ₅ , and set W ₁ ... W ₅ as shown in FIG. 13 a. Designated to become a kimono. As a specific example, W ₁ =
0.28 W ₂ =0.76 W ₃ =1 W ₄ =0.76 W ₅ =0.28. Similarly, to obtain the filter means with the MTF characteristic shown in FIG. 9b, D _n =D ₁₇ , W _S =W ₉ and W ₁ . . .
Define W ₉ as shown in Figure 13b. A specific example is W ₁ = 0.28 W ₂ = 0.52 W ₃ = 0.76 W ₄ = 0.94 W ₅
=1 W ₆ = 0.94 W ₇ = 0.76 W ₈ = 0.52 W ₉ = 0.28
It is. The characteristics in Figure 9 d are shown in Figure 13 c or d.
For the characteristics of the dotted line e ₁ and the solid line e ₂ in Figure 9 e, the weights in Figure 13 e and f can be used, and the weights in Figure 13 g can be used for the characteristic in Figure 9 f . By appropriately combining the MTF characteristics shown in FIGS. 13a to 13g, the combination of the first filter means and the second filter means shown in FIG. 6 can be obtained. Furthermore, if a CCD transversal filter is used as the filter means, the filter means can be easily constructed. FIG. 14 shows a specific example of the configuration of the determining means 34 shown in FIG. 8. In FIG. 14a, the first memory 340 and the second memory 341 have predetermined values Do(1), Do(2), Zo(1), Zo, respectively.
(2) is sent to comparators 344 and 345 via gate means 342 and 343. This comparator 344 connects one of the outputs Do(1) and Do(2) of the memory 340 selected by the gate means 342 and the arithmetic means 3.
3 is compared with the information amount signal Di from 3. Similarly, the comparator 345 outputs the output Zo(1) of the memory 341,
Compare one of Zo(2) and the focus detection signal Zi. The gate means 346 outputs the output α of the comparator 344, the output β of the comparator 345, and the discrimination means 34.
Input the output γ. A specific configuration of this gate means 346 is shown in FIG. 14b. A D-type flip-flop 347 outputs a clock pulse generated at a timing after the outputs of α and β are determined.
The output δ of the receiving gate means 346 is inputted and stored. The updated output of the flip-flop 347 is used as the output of the determining means 34. An example of the operation of this determining means 34 is shown in the table below.

[Table] Where Di(1)<Do(1) or Di(2)<Do(2), α=L Di(1)=Do(1) or Di(2)=Do(2), α=H Suppose that |Zi(1)|<Zo(1) or |Zi(2)|<Zo(2) and β=L. Otherwise, β=H. In the explanation of the second embodiment, the discussion will proceed based on the theme of switching a plurality of filter means depending on the magnitude of the defocus amount Zi, and in a subordinate manner to this, the expansion of the sampling area will be switched in accordance with the magnitude of the defocus amount. I mentioned switching the sampling pitch. In reality, it is most preferable to have both of these as in the second embodiment, but switching the filter depending on the amount of defocus and switching the sampling area and sampling pitch depending on the amount of defocus are related to the Nyquist frequency. Although the above problem exists, it is a different matter, and it is possible to provide a reasonably effective focus detection device even if only the latter is used. For example, as a filter means, only one filter having the MTF characteristic as shown in Fig. 9c is used, and in the vicinity of the focus, as shown in Fig. 10c, calculations are performed using data sampled over an area of l ₂ at a sampling period of 2P ₀ , In areas where the defocus is large, the filter means is left as is, but front and rear pin determination calculations are performed using data sampled over an area as wide as _l1 in FIG. _10b at a sampling period of 4P0. In this case, compared to the case where the MTF characteristic of the filter means is switched to that shown in FIG. The possibility of false focusing increases due to the presence of spatial frequency components, but apart from this possibility, a certain effect can be expected, partly due to the fact that where the defocus is large, the blur of the optical image is also large. obtain. That is, the effect of changing the sampling area and sampling pitch according to the amount of defocus described above can be expected as is. Of course, the MTF characteristic of a single filter is not limited to that shown in FIG. 9c, but may be as shown by the dotted line _e2 in FIG. 9e or other characteristics. In both the first and second embodiments described above, the focus detection signal based on the filtered output of the first filtering means and the focus detection signal based on the filtered output of the second filtering means are It was an alternative choice. Next, a third embodiment of the present invention will be described in which, instead of the above-mentioned alternative selection, the respective focus detection signals are used simultaneously in a predetermined relationship. In FIG _. 15, a series of photoelectric outputs a ₁ , _{b 1} . Sent. The photoelectric device 50 has the same configuration as that of the first or second embodiment,
The first and second filter means 52 and 53 are also
The first filter means 52 is similar to that of the embodiment.
The center of the MTF frequency band is shifted to the lower spatial frequency side than that of the second filter means 53. The delay time of the delay means 51 is such that after all the filtered outputs of the second filter means 53 are sent to the sample and hold means 54 for the series of photoelectric outputs, the filtered output of the first filter means 52 is transmitted to the sample and hold means. It is set so that it is sent to 54. Of course, this delay means 51 is the second
It can also be provided on the filter means side. This sample and hold means 54, followed by an A/D converter 55, a memory means 56, and an arithmetic means 57 have the same structure as in the second embodiment. This calculation means first calculates the filtered output of the second filter means 53, calculates the signals Di(2) and Zi(2), and then calculates the signal Di(2) for the filtered output of the first filter means 52. 1), calculate Zi(1). The memory circuit 58 receives the signal Di(1) from the calculation means 57,
Memorize all Di(2), Zi(1), and Zi(2). Synthesis means 5
9 inputs the above signal from the memory circuit 58;
An output Z is calculated by combining the focus detection signals Zi(1) and Zi(2) using the following predetermined relationship. That is, Z=(1-α)
Zi(1)+αZi(2)Here, the weight α is a number between 0 and 1, and is determined according to the magnitude of the signals Zi(1), Zi(2), Di(1), and Di(2). Ru. Specifically, the determination of α is based on the signal Zi
When (1) or Zi(2) is small, that is, when the imaging optical system is located near the focusing position, the signal Zi( 2) is emphasized by setting α to 1 or a value close to it, and conversely, when the signals Zi(1) and Zi(2) are sufficiently large, the band is set to the output of the first filter means 52 on the lower spatial frequency side. α so that the signal based on Zi(1) is emphasized
Let be zero or a value close to it. Also, if the signal Di(2) is very small near the focus, the signal Zi
(2) is degraded in accuracy, so at this time the signal
If Di(1) is large, the weight of the signal Zi(1) is increased, and conversely, even if the in-focus position is far away, if the signal Di(1) is very small, the weight of the signal Zi(2) is increased. If is large, the weight of the signal Zi(2) is increased. The memory circuit 60 stores the combined output Z when at least one of the signals Di(1) and Di(2) exceeds the corresponding predetermined values Do(1) and Do(2). In response to the output of the memory circuit 60, the display and imaging optical systems are driven as in the first and second embodiments. The sample pulse generating circuit 61 is similar to that shown in FIG. (Effects of the Invention) As is clear from the above description, the present invention has the following effects:
a first filter means for outputting a first filtered output mainly containing low-order spatial frequency components from the plurality of photoelectric outputs of the pair of photoelectric conversion element arrays; Since a second filter means for outputting a second filtered output mainly containing high-order spatial frequency components is provided, in practice, focus detection near the focus is affected by inevitable error factors. This solves the conventional drawback of large errors, and has the effect of enabling focus detection even in areas where the amount of defocus (defocus amount) is large.

[Brief explanation of the drawing]

FIG. 1 is an optical diagram showing the optical system of an embodiment of the focus detection device according to the present invention, FIG. 2 is a graph showing the MTF characteristics of the small lens array shown in FIG. 1, and FIG. 3 is a graph showing the above-mentioned first embodiment. 4a is a graph showing the MTF characteristics of the first filter means, FIGS. 4b and 4c are block diagrams showing specific configuration examples of the first filter means, and FIG. 5 6a, b, and c are graphs showing the MTF characteristics of the second filter means and a block diagram of a specific example of its configuration; FIG. 6 a, b, and c are graphs showing the MTF characteristics of the first and second filter means;
FIG. 7 is an optical diagram showing the optical system of the second embodiment of the present invention, FIG. 8 is a block diagram showing the circuit system of the second embodiment, FIG. 9a is a graph of the MTF characteristics of the photoelectric conversion element array, Figure 9b to f are MTF of filter means.
Characteristic graphs, Figures 10a, b, and c are waveform diagrams showing the photoelectric output, the output of the first filter means, and the output of the second filter means, respectively, and Figures 11 a to e show the output of the filter means and the sample pulse. 12 is a block diagram showing a specific example of the configuration of the filter means, FIGS. 13a to 13g are diagrams showing the weights of the filter means, and FIGS. FIG. 15 is a block diagram showing a circuit system of a third embodiment of the present invention. 4,21,50...Photoelectric device, 5,27,52
...first filter means, 6, 28, 53...second
Filter means, 7, 29... Selection means, 10, 3
3,57...Calculating means.

Claims (1)

[Scope of Claims] 1. A focus detection device that detects the focus adjustment state of an imaging optical system that forms an optical image of an object, comprising: a pair of photoelectric conversion element arrays each including a plurality of photoelectric conversion elements; a focus detection optical system that projects two images of substantially the same part of the object onto the photoelectric conversion element array; and a focus detection optical system that mainly extracts low-order spatial frequency components from the plurality of photoelectric outputs of the pair of photoelectric conversion element arrays. a first filter means for outputting a first filtered output; and a first filter means for removing a DC component from the plurality of photoelectric outputs of the pair of photoelectric conversion element arrays to mainly generate higher-order spatial frequency components than the low-order spatial frequency components. a second filter means for extracting and outputting a second filtered output; and a second filter means for extracting and outputting a second filtered output; A focus detection device comprising: correlation calculation means for calculating displacement, and detecting a focus adjustment state of the imaging optical system based on an output of the correlation calculation means. 2. The focus detection device according to claim 1, wherein the first filtered output is a low-order spatial frequency component that has a DC component and does not include a high-order spatial frequency component. 3. The focus detection device according to claim 1, wherein the first filtered output is a low-order spatial frequency component that does not include a DC component and a high-order spatial frequency component.