JPH03164709A - Focus detector - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] A. Industrial application field The present invention relates to a focus detection device for a camera or the like.

B. Conventional technology Conventionally, there has been a re-imaging optical system that separates an optical image of a target object obtained through a photographing lens into a pair of identical optical images and re-images them, and a re-imaged pair of optical images. A focus detection device is known that includes a pair of image sensors that photoelectrically convert and output each image sensor, and a detection circuit that detects the relative position of the pair of optical images based on output signals from these image sensors.

C1 Problems to be Solved by the Invention However, in this conventional focus detection device, when the target object has a periodic pattern, highly correlated focus detection results are obtained due to false focusing, and true It was difficult to differentiate from the focus.

This will be explained in more detail.

For example, as disclosed in Japanese Unexamined Patent Application Publication No. 60-37513, in a conventional detection device, the outputs of individual pixels constituting a pair of image sensors are a,...an, b...
When expressed as bn, the mutual shift amount of both images is L
The correlation amount C(L) between both images is defined as follows, C(L) = Σ 1ai-bLl, and C(L) is calculated for the number of consecutive shifts.

Here, when a periodic pattern as shown in FIG. 14 is projected onto a pair of image sensors, the correlation 1c (L) for the pair of image outputs at that time becomes as shown in FIG. As is clear from the figure, there are a plurality of image shift amounts Q, , Q, , Q, , Q, . . . for which the correlation is determined to be good. If Q2 is selected, it corresponds to the true in-focus position, but if any other image shift amount is selected, false focusing will occur. Note that in FIG. 14, the ★ marks correspond to the same portions on the subject. Furthermore, in FIG. 12, the scale marked at the top represents the pixel shift scale, and the scale at the bottom represents the defocus amount of the corresponding photographic lens.

In general, in focus detection devices using conventional image shift detection methods, for subjects with periodic patterns like this, the correlation is good even where the images are shifted by one complete cycle, so it is possible to make incorrect focus judgments. There is sex.

In addition, in the conventional focus detection device, there is a case where the amount of defocus cannot be determined by one focus detection operation.

It was necessary to scan the photographing lens and perform the focus detection operation again, which took time to detect the focus.Furthermore, in conventional focus detection devices, in order to improve detection accuracy, the axis of the pair of re-imaging optical systems If the distance (baseline length) is increased unnecessarily, the amount of defocus that can be detected will become smaller. Therefore, the inter-axial path n of the pair of re-imaging optical systems is determined as a compromise between the detection accuracy and the amount of defocus that can be detected, and the desired detection accuracy may not be obtained.

A first object of the present invention is to provide a focus detection device that can accurately detect focus even in a periodic pattern.

A second object of the present invention is to provide a focus detection device that does not reduce the amount of defocus that can be detected and improves detection accuracy.

06 Means for Solving the Problems The present invention will be explained with reference to FIG. 1(a), which is a diagram corresponding to the claims. An imaging optical system A that separates and forms a subject image, at least three photoelectric conversion means B that photoelectrically converts and outputs the at least three formed optical images, and outputs from these photoelectric conversion means B. A detection means C detects the relative positions of at least two pairs of optical images based on signals, and the detection results regarding the relative positions of each pair obtained by this detection means C are compared and calculated to detect false coincidences due to a periodic pattern. 3. The apparatus according to claim 2, further comprising an exclusion means for eliminating focus detection results having a high correlation with focus, and a signal forming means E for forming a signal for focus adjustment based on the focus detection results after being excluded. The invention provides an inter-axial distance Ql between at least two pairs of pupils of an imaging optical system according to claim 1.

l1 and lQ2 are determined so that there is no integer ratio relationship between Q2.

Further, as shown in FIG. 1(b), the invention of claim 3 is arranged such that the optical image of the target object obtained in the primary image plane through the photographing lens is transmitted in a straight line at predetermined intervals in the primary image plane. A re-imaging optical system F separates and re-images three or more identical second optical images using three or more split pupils arranged above, and a re-imaging optical system F separates and re-images the same second optical images using three or more split pupils arranged above, and a photoelectric system Multiple photoelectric conversion means G for converting and outputting
, a detection means H for detecting the relative positions of at least two pairs of second optical images based on the output signal from the photoelectric conversion means G; A signal forming means I is provided for forming a signal for focus adjustment based on a detection result in which focus detection is possible.

Furthermore, the invention of claim 4 provides that in the focus detection device of claim 3, when a plurality of focus detection results are obtained, the focus detection result for the second optical image by the pair of split pupils having the longest axis distance is A focus adjustment signal is formed from this.

According to the fifth aspect of the invention, the detection means H of the focus detection device according to the third aspect performs focus detection calculation as follows. sequentially from the photoelectric conversion means G output corresponding to the second optical image formed by the pair of split pupils having the longest axis distance to the photoelectric conversion means G output corresponding to the second optical image formed by the pair of split pupils having the shortest distance. are used to calculate the relative positions of each pair of second light images, respectively.

The first relative position obtained is taken as the focus detection result.

E0 Effect In the invention of claim 1, the object image is separated into at least three substantially identical optical images having parallax and projected onto the corresponding photoelectric conversion means B, respectively.

Based on the output from the photoelectric conversion means B, the detection means C calculates the relative positions of the two pairs of optical images obtained by the imaging optical systems having different axis distances. Furthermore, the elimination means eliminates focus detection results that are highly correlated due to false focusing by comparing and calculating the two.

As a result, a focus detection result indicating true focus is obtained.

In this case, as in claim 2, the distance Q between the axes of the two pairs of pupils,
By determining ,Q, false focusing can be eliminated even more precisely.

In the third aspect of the invention, the primary image plane is divided into at least three secondary images by split pupils arranged at different axis distances. From at least three secondary images, a pair of secondary images having a long distance between the axes of the split pupils and a pair of secondary images having a short distance between the axes of the split pupils are extracted.

Focus detection is performed from the photoelectric conversion output regarding either one of the pair of secondary images. Therefore, even if the interval between split pupils is set to have an inter-axis distance that provides the desired detection accuracy and an inter-axis distance that allows the desired amount of defocus to be detected, focus detection can be performed using either one. Therefore, a focus detection device having high detection accuracy and a large defocus amount detection range can be obtained.

F. Example Embodiments of the present invention will be described using FIGS. 2 to 13.

In FIG. 2(a), an exchangeable photographing lens barrel 10
0 has a well-known configuration in which a coupler 101 receives a driving force from a lens driving device 201 of a body 200 and moves a movable lens 103 via a gear train 102. A storage circuit 104 provided within the photographic lens barrel 100 stores the aperture F value and exit pupil position information of the photographic lens, and performs calculations and calculations within the body as necessary via a contact point 105 with the body 200. The control unit 202 reads out the data.

A part of the light passing through the photographing lens is guided to the optical system 210 of the focus detection device via the central semi-transparent part of the quick return mirror 203 and the sub-mirror 204. Here, as shown in FIG. 2(b), the focus detection optical system 210 in this embodiment has an optical system Y arranged on the optical axis of the photographic lens as in the conventional case, and a predetermined distance apart from the optical axis. It has optical systems X and Z arranged symmetrically, respectively. Each optical system Y, X, and Z includes a field diaphragm 211 that cuts extra light outside the focus detection area, a field lens 212, an aperture plate 213 that determines the pupil of the re-imaging lens, and a re-imaging lens 21.
4 and an IC board 215 equipped with a plurality of image sensors.
It consists of

This figure 2 (b) shows the focus detection optical system 2 from the film surface side.
10 is a view of the figure.

Image outputs regarding the optical images formed on the respective image sensors of the optical systems X, Y, and Z are stored in the memory section 206 via the interface section 205 in FIG. 2(a). The calculation/control unit 202 calculates a relative image shift amount with respect to a pair of image outputs having different parallaxes using a well-known method, and based on this calculates the relative image shift amount by driving the lens driving device 201 by a predetermined amount to achieve focusing. At this time, the display device 207 is turned on.

3 and 4 show FIG. 2(b) in more detail, and FIG. 3 shows each focus detection optical system Y.

4(a) is a front view of the field stop 211, FIG. 4(b) is a front view of the aperture plate 213, and FIG. 4(c) is a front view of the IC board 215. The field stop 211 is an aperture 211Y for three optical systems y, x, and z.

It has 211X and 211Z. Similarly, the aperture plate 213 is
Apertures (pupils) ya, yb, x for the three optical systems y, x, z
a, xb, xc, za, zb.

It has zc. As you can see from the diagram, there are three pupils xa, x
b and xc are arranged so that their pupil centers are on a straight line. Similarly, the IC board 215 also has three optical systems Y and X.
, image sensor Ya for Z.

It has Yb, Xa, Xb, Xc and Za, Zb, Zc. Note that optical systems X and Z are arranged symmetrically with respect to the optical axis of the photographing lens, and optical system 2 is similar to optical system X except that it is symmetrical, so a description thereof will be omitted.

Further, as shown in FIG. 3, the re-imaging lenses of the optical system Y are a pair of lenses 214ya as before.

214yb, and the reimaging lens of optical system X has three lenses 214xa, 214xb, and 214xc. The field lens 212 of this embodiment forms a conjugate image of each aperture of the re-imaging lens at approximately the exit pupil position of 100 mm, and the spread α of the detection light flux is approximately F7.

Furthermore, the focus detection area (a) to (a) on the optical axis of the photographic lens is
The images of (b) to (c) are formed on the image sensors Ya and Yb, and the images of focus detection areas (d) to (e) to (f) are formed on the three re-imaging lenses 2L4xa, 214xb, 214xc and in front of them. The image is formed on the image sensors Xa, Xb, and Xc through three pupils xa, xb, and xc placed at the center of the image.

Reference numerals 215a and 215b in FIG. 3 are light shielding plates for preventing stray light from entering the adjacent glass.

Next, regarding the focus detection optical system X configured in this way,
The relationship between the exit pupil of the attached photographic lens and the vignetting of the detected light beam will be explained in detail with reference to FIGS. 5 and 6.

FIG. 5 is a diagram illustrating in which range of the exit pupil position the detected light flux of the optical system X is eclipsed for an F5.6 lens;
Figure 6 shows the exit pupil of the photographic lens with the field lens 21.
2 is a diagram projected onto the pupil position of the diaphragm 213 by No. 2.

The following can be seen from FIGS. 5 and 6.

■Re-imaging lenses 214xa, 214
Which detection light flux La, Lb, Lc of xb, 214xc
Therefore, all three image outputs from the image sensors Xa, Xb, and Xc can be used for focus detection.

(2) For a photographic lens whose exit pupil position is in the range of Ll from about 50+om to 80+m, the image output of the image sensor Xc cannot be used because the detection light flux Lc is vignetted. However, the detection light flux La.

Since Lb does not cause vignetting, image sensors Xa and
Focus detection is possible by detecting image shift from the image output of b.

■The exit pupil position is about 120-200m+m.

For a photographic lens within the range of , the detection light beam La is vignetted and cannot be used. However, since no vignetting occurs in the detection light beams Lb and Lc, focus detection is possible by detecting image shift from the image outputs of the image sensors Xb and Xc.

Since vignetting does not always occur in the optical system Y, focus can be detected from the image outputs of the image sensors Ya and Yb. That is, in the focus detection optical system Y along the optical axis, focus detection is always performed by the pair of image sensors Ya and Yb.

Here, in the focus detection optical systems X and Z, the calculation/control unit 202 determines which image sensor's image output is to be used as a pair of image outputs having different parallaxes, as shown in Table 1 below. Ru.

In the focus detection optical systems x and z mentioned above, Table 1
As shown in FIG.
Condition C under which the detection light flux La is eclipsed can be known in advance 6 For example, if the exit pupil position PO is between 90 and 109.9 mm and the aperture F value is 5.6 or less, then condition A is satisfied, and the exit pupil position is between 50 and 59.9 mm.
If the exit pupil position is between 110 and 129.901!1 and the aperture F value exceeds 5 and is less than or equal to 5.6, then condition C is satisfied. It is determined as follows.

And in the case of 1 condition A, how to take image pairs.

Regarding optical system X, image sensors Xa and Xb.

Image sensors xb and Xc, image sensors Xa and Xc
As for the optical system Z, any of the pairs of image sensors Za and Zb, image sensors zb and Zc, and image sensors Za and Zc can be used. In the case of condition B, optical system X is used as the method of taking the image pair.

Regarding Z, there are a pair of image sensors Xa and xb and image sensors Za and zb. Furthermore, in the case of condition C, the image pairs are taken as a pair of image sensors xb and Xc and image sensors zb and Zc for the optical systems x and Z.

In the case of condition A, three choices are possible;
It is advantageous in terms of detection accuracy to use the pair c. This point will be explained in detail later.

Next, the focus detection operation will be explained with reference to the flowchart shown in FIG.

In step S1, the calculation/control unit 202 reads the aperture F value and the exit pupil position PO from the memory unit 104 of the photographing lens.

In step S2, the arithmetic/control unit 202 determines, based on Table 1, which of conditions A, B, and C the attached photographic lens is already under. This allows the image sensor in which vignetting occurs to be identified. Next, in step S3, the image sensor starts accumulating the charge, and when a predetermined amount of charge is accumulated, the image data is transferred to the memory section 20 in step S4.
6. Here, the accumulation time is calculated by the calculation/control unit 202.
and is controlled by the interface unit 205 in a well-known manner. In this case, since the transfer takes time, it is more efficient to omit the transfer of the output of the image sensor that is determined to cause vignetting.For example, if the image sensor is a CCD
When using the CCD, only the CCD determined to cause vignetting among the three CODs is prevented from performing transfer.

Next, in step S5, it is determined whether or not vignetting has occurred, and if there is vignetting, the process proceeds to step S6, and if there is no vignetting, the process proceeds to step S8.

If there is vignetting, the process proceeds to step S6, where a focus detection calculation is performed using the outputs of the pair of image sensors without vignetting, and based on this result, the movable lens 103 is driven in step S6, and the image is displayed on the display device 207. Do this.

If there is no vignetting, the process proceeds to step S8 and subsequent steps to find the optimal amount of defocus. In the processing after step S8, the defocus amount is roughly determined as follows.

■Aperture of the re-imaging optical system used (xa in Fig. 4(b),
xb, Focus detection is first performed on image output.

(2) When only one highly correlated image shift amount is obtained from the focus detection results, the defocus amount is determined from that image shift amount.

■If a highly correlated image shift amount cannot be obtained from the image sensor outputs of a pair of reimaging optical systems with a long center distance, focus detection is performed using the image sensor outputs of a pair of reimaging optical systems with a short center distance. .

(If a highly correlated image shift amount is found from the focus detection results from the image sensor output of the re-imaging optical system with a short inter-axis distance, then the defocus amount is calculated.

■When multiple highly correlated image deviation amounts are determined from the image sensor outputs of a pair of re-imaging optical systems with a long center distance, they are converted into defocus amounts and True focus is calculated by comparing the defocus amount determined from the image sensor output.

Discriminate.

This will be explained in detail below.

In the case of the above-mentioned condition A, that is, when no vignetting occurs in any of the three re-imaging optical systems, three combinations of image sensors are possible when detecting a shift between two images. That is, regarding the optical system X, image shift detection is possible for each pair of image sensors Xa and xb, xb and Xc, and Xa and Xc. Here, the larger the distance between the axes, the smaller the amount of defocus corresponding to a predetermined image shift, so the focus detection accuracy is higher, and conversely, the shorter the distance between the axes, the more focus detection becomes possible over a wider range of defocus amounts. It has the characteristic of becoming.

Therefore, first in step S8, image sensors Xa and X related to reimaging lenses Xa and
Image shift amount Qi with good correlation based on the output image data of c
(i=1.2...) is calculated. Here, the number of image shift amounts Qi found is stored in the memory. If only one image shift amount is found, step S9 is affirmed, and step S10
The image shift amount Q is converted into a defocus amount.

If the amount of defocus is large, image shift cannot be detected from the outputs of Xa and
Regarding the pair of image sensors Xa and xb (or xb and Xc) corresponding to a and xb, a well-correlated image shift amount Rj (j=1.2...
・) Find.

Then, in step S12, it is determined whether I=O, and if I=0, in step S13, the image shift amount R1 obtained from the outputs of the image sensors Xa and xb is set as the defocus amount ZRi.
Convert to

Of course, we first consider multiple pairs of (Xa, Xc) and (Xa,, Xb), or (X a + X c ) and (
Xa, Xb) and (Xb,
The magnitude and direction of the defocus amount may be determined from the image shift amount for the pair c).

By doing the above, for a bright photographic lens, it is possible to expand the defocus determination area or the front and rear focus determination area and to expand the focusing accuracy near one focus.

The above is the specific procedure of the method described in items 0 to 0 above.

In FIG. 8, the periodic pattern is image sensor Xa.

This shows the image output when projected onto Xb and Xc, and the ★ marks correspond to the same portions on the subject. Moreover, FIG. 9 shows an image sensor Xa.

The above-mentioned correlation amount C(L) is shown when the mutual shift amount is changed with respect to the image output of Xc. In Figure 9, the scale marked at the top represents the pixel shift scale,
At the bottom, there is a scale indicating the amount of defocus for the corresponding photographic lens. As is clear from the figure, there are a plurality of image shift amounts Q, , Q, , Q3. for which the correlation is judged to be good. Q,... exists. On the other hand, FIG. 10 shows the same thing as above for the outputs of image sensors Xa and xb.
The scale markings at the top are different from those in the illustration.

This is because the distance between the detection axes (the distance between the pair of re-imaging lenses) is different, and when the re-imaging lenses are lined up at equal intervals as shown in Figure 4(b), the image shift for the same amount of defocus of the taking lens This is because the amount is twice as different in the case of FIG. 9 (using xa and xc) and the case of FIG. 10 (using xa and xb).

As can be seen by comparing FIG. 9 and FIG. 10, it is possible to eliminate false focusing between Q and Q by comparing the scales of defocus.

The flow of processing for eliminating erroneous detections due to such periodic patterns will be explained based on the flowchart of FIG. 7.

When a plurality of correlation quantities @Qi are determined for the outputs of the image sensors Xa and Xc in step S8, if the target object has a periodic pattern, the number I of correlation peaks is plural, so I≠1 and I#O, and the process proceeds to step S11. , image shift amount R with good correlation regarding image data of image sensors Xa and Xb
Calculate j.

Next, since I≠0, the process proceeds to step S14, and Qi.

Defocus amounts ZQi and ZRj corresponding to Rj are calculated. In step S15, in order to eliminate false detection, it is determined whether or not, and ZQi that satisfies this condition is selected.

only becomes smaller than the predetermined value. Therefore z Q
z and ZQ4 are selected, and Z Q i other than when ZRj having approximately the same value as ZQi exists is removed. Thereafter, in step 816, the one ZQ with the smallest absolute value of the defocus amount among the selected ZQi is determined as the defocus amount.

According to the above configuration, it is possible to eliminate false focusing even for periodic patterns, which was previously impossible. Note that, as described above, the aperture xa of the re-imaging optical system.

If xb and xc are arranged at equal intervals, there is still a possibility that ZQ4 will be selected as a false focus, but since the probability that the lens is in a very defocused position is low, in step S16, ZQ with the smallest absolute value is selected. In most cases, there is no problem in selecting the correct value.Step S15 is a method of excluding defocus amounts other than those that closely match the defocus amounts ZQi and ZRi calculated from the detection means having two different baseline lengths. If so, the method is not limited to the above method.

Next, a method for completely eliminating false focusing will be described.

FIG. 11 is a diagram similar to FIG. 4, but the apertures xa, xb, and xc of the reimaging lens are arranged at non-uniform intervals 111ab≠12bc. In this case, when a periodic pattern similar to that described above is projected onto the image sensors Xa and Xc corresponding to the apertures Xa and XQ, a graph of correlation regarding the image output becomes as shown in FIG. 12. Further, a graph of the correlation regarding the image outputs of the image sensors Xa and Xb corresponding to the apertures xa and xb is as shown in FIG. As is clear from the figure, the distance Q between the axes of the apertures xa and xa, = (Qab + Qbc
) and the interaxial path between the apertures xa and xb lt Q2 = (Q a
b) When there is no simple integer ratio relationship between
Rj will never match. Here, as for the degree of non-uniform spacing, the larger distance between the axes is Q4. Let Q be the smaller distance between the axes, where the nearest integer value represents the value below the decimal point.

When t>0.02 is necessary, t>0.04 is preferable, and t>0.1 is sufficient.

In this way, if the inter-axial distance of the re-imaging lens (xa, xb) and the inter-axial distance of (x b , x c) are slightly different, false focusing can be reliably eliminated.

Note that this embodiment can also be applied to a so-called external light triangular distance measuring device, in which case xa.

Rather than being re-imaging lenses, xb and xc are lenses that directly image the object onto the image sensor.

Further, in the above embodiments, only the focus detection devices in the focus detection areas provided above and below off-axis separate the target object into the same three optical images and each is received by the corresponding image sensor. The focus detection device in the focus detection area provided at the position Y in FIG. 2(b), that is, around the optical axis, can be constructed in the same manner. Alternatively, the present invention may be applied to a device that does not have an off-axis focus detection device and detects focus only around the optical axis.

Furthermore, although the number of divided pupils is three in the above description, it may be four or more if disadvantages such as the reduction in the area of the pupil and the increase in the low luminance limit and the increase in the scale of the apparatus are ignored. However, when the focus detection area is set to an image height range of about 5 to 1011I11, the optimum number of pupils divided into one is three.

Effects of the G1 Invention According to the inventions of claims 1 and 2, a subject image is separated into at least three substantially identical light images having parallax, and
The relative positions of two pairs of optical images are detected, and the relative positions of each pair are determined.

Since the detection results are compared with each other and highly correlated focus detection results due to false focusing due to a periodic pattern are excluded, it is possible to detect the focus of a subject having a one-period pattern.

According to the invention of claims 3 to 5, the focus detection calculation is performed by determining the relative positional relationship between at least two pairs of optical images from the photoelectric conversion outputs of three or more reimaging optical systems. Accuracy and detectable defocus range can be improved.

[Brief explanation of the drawing]

Figure 1 shows the claims and the correspondence diagram. FIGS. 2 to 12 illustrate one embodiment. Figure 2(a) is a block diagram showing the overall configuration, Figure 2(b)
) is a front view of the focus detection optical system viewed from the film side.
Figure 3 is an enlarged view of the same, Figures 4 (a) to (c) are front views showing the field diaphragm, diaphragm plate, and IC board, respectively, and Figure 5 explains the vignetting of the detected light flux for the focus detection optical system X. FIG. 6 is a diagram explaining the case where the exit pupil of the photographing lens at various positions is projected onto the pupil via the field lens, FIG. 7 is a flowchart showing the processing sequence of focus detection calculation, Fig. 8 is a diagram explaining the periodic pattern projected on the three image sensors, and Fig. 9 is the correlation amount C(L) obtained from the output of a pair of image sensors with a long baseline length.
FIG. 10 is a graph of the correlation amount C(L) obtained from the outputs of a pair of image sensors with short baseline lengths. FIG. 11 is a front view of each optical element of the re-imaging optical system in which apertures are arranged at non-uniform intervals. FIG. 12 is a diagram corresponding to FIG. 9 using the re-imaging optical system of FIG. 11. FIG. 13 is a diagram corresponding to FIG. 10 using the re-imaging optical system of FIG. 11. FIG. 14 is a diagram illustrating a periodic pattern projected onto two image sensors. 100: Interchangeable lens 2oO: Camera body 202:
Calculation/control unit 210: Focus detection optical system 211: Field diaphragm 212: Field lens 213: Aperture plate 214: Re-imaging lenses 214xa to 214x
c: Re-imaging lens 215: Ic substrate xaA+xc
: [! Xa"Xc: Image sensor A: Imaging optical system B C: Detection means D E: Signal formation means F G: Photoelectric conversion means H I: Signal formation means La"Lc: Detection light flux: Photoelectric conversion means: Exclusion means: Re Imaging optical system: detection means

Claims (1)

[Claims] 1) An imaging optical system that separates and forms a subject image into at least three substantially identical light images having parallax, and photoelectrically converts and outputs each of the formed at least three light images. at least three photoelectric conversion means, a detection means for detecting the relative positions of at least two pairs of the optical images based on output signals from these photoelectric conversion means, and a detection means for detecting the relative positions of each pair obtained by the detection means. an elimination means for eliminating highly correlated focus detection results due to false focusing due to a periodic pattern by comparing detection results with each other; A focus detection device comprising a signal forming means for forming a signal. 2) In the apparatus according to claim 1, when the distances between the axes of at least two pairs of pupils of the imaging optical system are l_1 and l_2, the axes are arranged so that there is no relation of an integer ratio between l_1 and l_2. A focus detection device characterized in that a distance between the two points is determined. 3) The optical image of the target object obtained in the primary image plane through the photographing lens is divided into three or more split pupils arranged in a straight line at predetermined intervals in the primary image plane, and three or more identical images are obtained. a re-imaging optical system that separates and re-images a second optical image; a plurality of photoelectric conversion means that photoelectrically converts and outputs each of the re-formed second optical images; and the photoelectric conversion means detection means for detecting the relative positions of at least two pairs of said second light images based on output signals from said second light images; and among the relative position detection results of said two pairs of second light images, focus detection was possible. 1. A focus detection device comprising: signal forming means for forming a signal for focus adjustment based on a detection result. 4) In the focus detection device according to claim 3, when a plurality of focus detection results are obtained, a focus adjustment signal is formed from the focus detection results for the second optical image by the pair of split pupils having the longest axis distance. A focus detection device characterized by: 5) In the focus detection device according to claim 3, the detection means converts the output of the photoelectric conversion means corresponding to the second optical image formed by the pair of split pupils having the longest axis distance into the second optical image formed by the pair of split pupils having the shortest distance. The relative position of each pair of second light images is calculated by sequentially using the outputs of the photoelectric conversion means corresponding to the light images, and the first relative position obtained is taken as the focus detection result. Focus detection device.