JP2881772B2 - Focus detection device - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a focus detection device for a camera, and more particularly to an improvement in focus detection accuracy by calculating an image shift amount.

B. Conventional Techniques and Problems Thereof FIG. 10 shows an example of a focus detection device of a camera that detects a shift amount between two images formed by dividing a pupil of an objective lens and determines a focus adjustment state of the objective lens. Show.

The light beam of the object light incident via the objective lens 10 passes through the half mirror 11, the sub-mirror 12, the field mask 13, the field lens 14, the stop 15, and the re-imaging lens 16 to form an image on the image sensors 17a and 17b. These image sensors 17a, 1
The pair of light images formed on 7b are separated from each other in a so-called front focus state in which the focus adjustment state of the objective lens 10 forms a sharp image of the subject before the planned focal plane, and conversely, a so-called rear focus state after the planned focal plane. When they are pins, they approach each other, and when they are connected to a predetermined focal plane, light images are arranged at a certain interval between them. Therefore, each light image is converted to an image sensor 17a,
The focus adjustment state of the objective lens 10 can be determined by performing photoelectric conversion by 17b and arithmetically processing these signals to determine the amount of shift between the pair of optical images.

Therefore, the photoelectric conversion outputs of the pair of image sensors 17a and 17b are transmitted to the microprocessor via the interface unit 18.
It is sent to 19 and its calculation unit 21 determines the amount of deviation, and drives the motor 24 via the control unit 22 to move the objective lens 10 in the optical axis direction. The in-focus state is displayed on the display 23 in the viewfinder. The image signals from the image sensors 17a and 17b are temporarily stored in the memory unit 20.

By the way, the screen plate indicated by reference numeral 25 in FIG.
As shown in FIG. 11, a focus detection visual field display (visual field frame) 251 is drawn, and can be seen by the camera user through the finder. On the other hand, the image sensors 17a, 17
The range where the light image is projected on b is indicated by reference numeral 252 in FIG. Although this projection range 252 is invisible to the user, if a subject whose focus is to be detected is placed within the frame of the detection visual field display 251, the light image is captured by the image sensors 17a and 17b.

However, the conventional focus detection device has a disadvantage that the focus detection becomes unstable when, for example, a black and white edge pattern is positioned at the end of the field frame 251. This can be explained as follows with reference to FIG.

12 (a1) and (b1) show the contrast distribution of the light image on the image sensors 17a and 17b corresponding to the range of the field frame 251. (a2) and (b2) show the image sensors 17a and 17b, a
3) and (b3) are comparison ranges of the image sensors 17a and 17b when L = 0, 1, 2,..., −1, −2. Is shown. (A3), (b
The correlation amount C (L) is calculated by changing the image range to be compared as in 3), and the deviation amount is calculated by interpolating the value of L that gives the minimum value of the correlation amount C (L).

According to this method, as shown in FIGS. 12 (a1) and (b1), when an edge portion containing a large amount of information is located at the end of the focus detection range FR, the detection accuracy is reduced.

 There are also the following problems.

That is, when a small amount of information is localized in a partial area GR as in the case where the image sensor output is as shown in FIG. 13, L = 0, 1, 2,..., −1, −2 in FIG. When the images in the area of the wide focus detection range FR are compared with each other while shifting the data, the data of the area that does not include information but includes only the noise component (the area excluding GR from FR) is included in the comparison data. That is, if the area to be compared with the case as shown in FIG. 13 is widened, the detection accuracy is reduced.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a focus detection device capable of accurately detecting a focus even if an edge is present at an end of a detection range of an image sensor output or an information amount is localized.

C. Means for Solving Problems Referring to FIG. 1 showing an embodiment, the present invention provides:
A pair of image sensors 17a and 17b each including a plurality of pixel arrays, receiving a pair of light beams of the same object light, and performing photoelectric conversion, and a pair of the image sensors 17
The present invention is applied to a focus detection device including image shift calculating means 213 for calculating a relative displacement of a light image on both image sensors from first image data and second image data based on the outputs of a and 17b. . According to the first aspect of the present invention, the setting means 212 calculates the contrast of the pattern of the object light based on at least one of the image data, and sets the reference image range used in the image shift calculating means 213 based on the contrast. Based on the reference image range set by the setting unit 212, the image shift calculating unit 213
The above-mentioned problem is solved by calculating the relative displacement amount of the image by comparing the image data.

According to the second aspect of the present invention, in an initial reference image range including a plurality of pixels in at least one of the image data,
In the case where there is a large contrast of the pattern of the object light, there is provided a setting unit 212 for changing and setting a reference image range used by the image shift calculating unit 213 so as to include a pixel having a large contrast. Based on the obtained reference image range, the image shift calculating unit 213 solves the above-described problem by comparing the first and second image data and calculating the relative displacement amount of the image.

According to a third aspect of the present invention, in the case of image data in which a large contrast is localized, the setting unit 212 changes the reference image range to an area around the image data in which the large contrast is localized.

In the case of image data having a large contrast at the end of the reference image range, the setting means 212 changes the reference image range so that the large contrast does not exist at the end.

D. Function In the invention of claim 1, the setting means 212 calculates the contrast of the pattern of the object light based on at least one of the image data, and based on the contrast, the reference image used in the image shift amount calculating means 213. Set the range. The image shift amount calculating unit 213 compares the first and second image data based on the set reference image range and calculates the relative displacement amount of the image.

In the invention of claim 2, the setting unit 212 includes a pixel having a large contrast when the object light has a large contrast in an initial reference image range including a plurality of pixels in at least one of the image data. The reference image range used by the image shift calculator 213 is changed and set. The image shift calculating unit 213 calculates the relative displacement amount of the image by comparing the first and second image data based on the set reference image range.

In the invention according to claim 3, in the case of image data in which a large contrast is localized, the setting unit 212 changes the reference image range to an area around the image data in which the large contrast is localized.

In the invention according to claim 4, in the case of image data having a large contrast at the end of the reference image range, the setting unit 212 changes the reference image range so that the large contrast does not exist at the end.

In the above sections D and E for describing the configuration of the present invention, the drawings of the embodiments are used for easy understanding of the present invention, but the present invention is not limited to the embodiments.

E. Embodiment An embodiment will be described with reference to FIGS.

The overall schematic configuration of the focus detection device in this embodiment is
As shown in FIG. 10, the calculation method of the image shift amount by the calculation unit 21 is different. Therefore, the calculation method will be mainly described below.

 FIG. 1 shows details of the calculation unit 21 according to the present embodiment.

The calculation unit 21 compares a pair of light images for calculating a shift amount according to the light image pattern with a minute area information amount calculation unit 211 that quantifies the light image pattern from the memory unit 20 (hereinafter, referred to as a range). Range setting unit 21 for determining a reference image range)
2 and an image shift calculator 21 that shifts a pair of optical image patterns and calculates a shift amount based on the correlation amount C (L) at that time.
With 3.

Next, the measurement of the light image pattern by the minute area information amount calculation unit 211 will be described.

As shown in FIG. 5B, an image sensor 17 including a plurality of pixels 17 ₁ , 17 ₂ ...
And it is divided into a plurality of minute regions MR ₁ ~MRm including four pixels (FIG. 5 (c)). The micro regions may be arranged so as to overlap each other as shown in FIG. In this case, the number of minute regions increases, but this is more preferable if there is no problem in memory capacity or calculation time. Further, the number M of pixels constituting the minute region MR is 2 to 10
Although the degree is appropriate, it needs to be sufficiently small with respect to the focus detection range FR.

The information amount dk is obtained for each of the minute regions MR as follows.

When the first image output data corresponding to one image sensor 17a is a ₁ ... An, the difference data ei = ai−
ai _{+ 1} is calculated (step S2 in FIG. 2). The difference data is
Every other difference ei = ai-ai ₊₂ or every other ei = ai-ai ₊₃
And so on. Regarding the difference data obtained in this manner, the information amount dk of each minute region MR is represented by dk = Σ | e [M (k−1) + m] | where M = 4, Σ is the sum of m = 1 to M Represents an operation.

(Step S2 in FIG. 2). It can be seen that the greater the value of the information amount dk, the greater the information amount of the k-th minute region MRk.

Since the information amount dk defined in this way has a roughly proportional relationship with the detection accuracy, a predetermined detection accuracy can be achieved by providing a reference image range such that the information amount exceeds a predetermined value. Is easily understood. Therefore, it is a very effective parameter when determining the reference image range.

In the range setting unit 212, thus by comparing the information amount dk of each minute region MR which is calculated to determine the starting point l ₁ and end l ₂ of the reference image within the above range in step S3 of FIG. 2. Then, in the image shift calculation unit 213, based on the determined reference image range and the second image data for the first image data,
The correlation amount C (L) relating to L pixel shift is , And the value of L that gives the minimum is calculated as the image shift amount (step S4 in FIG. 2).

Next, will be described in detail starting l ₁ and method for determining the end point l ₂ of the reference image region in step S3, as described above, in this embodiment, the range setting unit 212 is unnecessary area removal unit 212A and scope end height Since the information processing apparatus includes the information amount avoiding unit 212B, the range end high information amount avoiding unit 212B will be described first with reference to FIG.

First, in step S21, the size of the focus detection range FR is determined to a size set in advance as an initial value. Next, in step S22, an area wider by one minute area on both sides of the focus detection range FR is added to the new focus detection range FR '.
And In step S23, the sum of the information amounts dk regarding the minute regions included in both the regions FR and FR 'is obtained, and these are set as d _TOT and d _TOT ', respectively. In step S24, it is determined whether the ratio d _TOT / d _TOT ′ of the sum is greater than a reference value γ, and it is determined whether the boundary of the focus detection range FR includes a portion having a relatively large information amount. Detect. Here, γ
It is a constant of 0.5 or more and 1 or less. If d _TOT / d _TOT ′> γ holds, it is determined that there is no portion where the amount of information is extremely large at the end, and the flow advances to the next step S25. Also, d _TOT / d _TOT ′>
If γ is not satisfied, the process moves to step S27, and the image obtained by expanding one minute region on both sides of the previous focus detection range FR is set as a new reference image range, and the process returns to step S22. At step S25, thus determined position of the start point for the range of the focus detection range FR (pixel position) to determine the r ₁ and the position r ₂ for that endpoint. Then, in step S26, the values of the positions r ₁ and r ₂ thus determined are used as they are as the start point l ₁ and the end point l ₂ of the reference image range in the image shift calculation.

Referring to FIGS. 6 and 7, range end high information amount avoiding section
The operation of 212B will be specifically described.

FIG. 6 shows that a large amount of information is included at the left end of the focus detection range FR, and there is a problem in detection accuracy if the image shift amount calculation is performed with this focus detection range FR. Therefore, by executing the above-described steps S21 to S26, the focus detection range FR is expanded, and the end of the focus detection range FR does not include a large amount of information. In other words, with the initial focus detection range FR,
From the information dk shown in FIG. 6 (b), d _TOT / d _TOT '= 0.39, and d _TOT / d _TOT ' = 0.95 for the focus detection range FR after adding one minute area on each side. No problem. On the other hand, in the example of FIG.
In FIG. 7B, the information dk to d _TOT /
d _TOT ′ = 0.73, and there is no problem in the detection accuracy of the image at the edge of the area of the focus detection range FR.

Since such processing is performed, the range edge high information amount avoiding unit
212B has an effect of preventing a decrease in detection accuracy due to a high contrast portion being applied to an end of the focus detection range FR.

Next, the unnecessary area removing unit 212A constituting the calculating unit 21 will be described with reference to FIG.

In step S27 following step S26 in FIG. 3, values r ₁ and r ₂ determined in step S26 as values l ₁ and l ₂ of the reference image range.
Is set, and 0 is set in the start point change flag Q1 and the end point change flag Q2. If the flags Q1 and Q2 are 0, there is no change, and if they are 1, there is a change. Next step
In S28, the focus detection range FR [r _1, r _2] amount of information dk in small regions in the seek a position p k that maximizes.

In step S29, this information amount dk is compared with a predetermined value Th, and d
If k> Th, the process proceeds to step S30. In step S30,
The information amount dk is sequentially examined on both sides of the minute region located at k = p, and a minute region in which the information amount dk is equal to or less than the predetermined value Th is found. When found in the direction of r ₁ is the flag Q1 = 1, and stores the position k = q1. Further, when it finds the direction of r ₂ and flag Q2 = 1, and stores the position k = q2.

The values l ₁ and l ₂ of the reference image range are preferably set at the center positions of the small areas q ₁ and q ₂ thus stored, but are determined at the ends of q ₁ or q ₂ closer to or farther from p. You can also.

Specifically, the case of the optical image pattern shown in FIG. 5 (a) will be described with reference to FIG. 5 (d). The maximum amount of information dk is k = 6, that is, p = 6. Here, if the predetermined value Th = 10, is to become the information amount dk <Th at both sides, the r ₁ direction k = 7, i.e. q ₂ = 7 in the k = 5, i.e. q ₁ = 5, r ₂ directions Become.

Next, in step S31, the information amount dll in the reference image range l _{1 to} l ₂ thus determined is Is calculated from In step S32, it is determined whether or not the information amount dll of the reference image range obtained in this manner includes an information amount sufficient to perform an image shift calculation based on the information amount dll. For example, as shown in the drawing, the dll is compared with a predetermined value Th1 to determine whether the information amount is sufficient. Or dll / d _TOT > β
(Β is 0.5, for example) may be used to determine that the amount of information is sufficient, or may be determined by a combination of both. If the amount of information is determined to be insufficient in step S32, step S33
In reference image area is extended to r ₁ ~r ₂ again. As a result, although the information amount of each minute area MR is small, the information is distributed in a wide range, and if the area is made wider, it is possible to save a case where detection is possible. If the information amount is sufficient in step S32, step S33 is bypassed.

Here, the relationship between the reference image ranges l _{1 and} l ₂ determined by the above-described range setting unit 212 and the expansion of the focus detection range FR as an initial value will be described collectively.

In the subject pattern shown in FIG. 5A, the focus detection range FR
On the other hand, the reference image ranges l _{1 to} l ₂ are limited to (a) of only the portion where the information is localized. Therefore, the influence of noise from unnecessary portions can be reduced.

In the subject pattern of FIG. 6A, the focus detection range FR
On the other hand, when the range edge high information amount avoiding unit 212B is used, the reference image ranges l _{1 to} l ₂ are as shown in (b), and when the unnecessary area removing unit 212A is also used, as in (c). In any case, a portion having a large amount of information is not at the end of the reference image range, so that the detection accuracy is improved. That is, the range edge height information amount avoiding unit 212B or the unnecessary area removing unit 212A may be used independently.

As shown in FIG. 7 (a), in a subject pattern in which the amount of information is dispersed irrespective of location, the focus detection range FR given as an initial value is unchanged (d) the reference image range l _{1 to} l _2.
Becomes

Further, the discussion has been made on the assumption that the focal length range FR is equal to the detection field frame of FIG. 11, but it is needless to say that the present invention is not limited to this. For example, in the case where the detection visual field is divided into a plurality of detection visual fields and focus detection calculation is performed on each of them, each of the detection visual fields is treated as a focal length range FR.

 Finally, the image shift calculator 213 will be briefly described.

The first image output data a ₁ determined as described above.
The comparison area of the second image output data is shifted by one pixel with respect to the shift amount L with respect to the reference image range l _{1 to} l _{2 of} an as shown in FIG. , And the value of L that gives the maximum correlation is calculated from this by a known method. This equation is for the case of the shift as shown in FIG. 9, but the shift may be as shown in FIG. In the latter case, the reference image range is slightly wider. For example, l ₁ and l ₂ are determined at the ends of the minute regions q ₁ and q ₂ that are farther from P.

In the above description, the first and second image output data are described as being direct output from the image sensor. However, the data obtained by performing some sort of filter processing on the direct image output from the image sensor is referred to as the first and second image output data. It may be treated as the second image output data.

F. Effects of the Invention As described above, according to the present invention, it is possible to set the reference image range used in the image shift calculating unit based on the contrast of the pattern of the object light, or to set a pixel having a large contrast of the pattern of the object light. Change and set the reference image range used by the image shift calculation means to include, or,
A reference image range having an optimum spread or position is set according to the subject pattern projected on the image sensor, and an image shift comparison is performed with respect to the other image. As a result, stable focus detection becomes possible.

[Brief description of the drawings]

1 to 8 illustrate one embodiment of the present invention.
FIG. 1 is a detailed block diagram of the calculation unit. FIG. 2 is a flowchart showing an image shift calculation routine. FIGS. 3 and 4 are flowcharts showing a routine for determining the reference image ranges l _{1 to} l ₂ . 5A and 5B illustrate a subject pattern having localized information, wherein FIG. 5A illustrates a subject pattern, FIG. 5B illustrates a configuration diagram of an image sensor, and FIG. 5C illustrates a minute region MR; (d) is a diagram showing a calculation result of the amount of information d _K. 6A and 6B illustrate a subject pattern having an edge at the end of the focus detection range FR. FIG. 6A is a diagram illustrating the subject pattern, and FIG. 6B is a diagram illustrating a calculation result of the information amount dk. FIG. 7 illustrates a normal subject pattern.
(A) is a figure showing a subject pattern, (b) is a figure showing a calculation result of information amount dk. FIG. 8 is a view for explaining a modified example of how to take the minute region MR. FIG. 9 is a diagram for explaining the image shift calculation. FIG. 10 is a diagram showing the overall configuration of a focus detection device to which the present invention is applied. FIG. 11 is a diagram showing a focus detection area in the field frame and a light image projected on the image sensor in an overlapping manner. FIG. 12 is a diagram for explaining an image shift calculation of a subject pattern having an edge at an end of the focus detection range FR. FIG. 13 is a diagram showing a subject pattern in which the amount of information is localized. 10: objective lens, 13: field frame, 14: field lens, 15:
Aperture, 16: Re-imaging lens, 17a, 17b: Image sensor, 19:
Microprocessor, 21: calculation unit, 211: minute area information amount calculation unit, 212: range setting unit, 212A: unnecessary area removal unit, 212B:
Range edge height information amount avoidance unit

Continuation of the front page (58) Field surveyed (Int.Cl. ⁶ , DB name) G03B 3/00 G02B 7/28

Claims (4)

(57) [Claims] 1. A pair of image sensors each comprising a plurality of pixel arrays, respectively receiving a pair of light beams of the same object light and performing photoelectric conversion, and a first image data based on outputs of the pair of image sensors. A focus detection device comprising: image shift calculating means for calculating a relative displacement of the light images on both image sensors from second image data; and a contrast of the object light pattern based on at least one of the image data. And setting a reference image range to be used by the image shift calculating means on the basis of the contrast. Based on the reference image range set by the setting means, the image shift calculating means sets A focus detection apparatus for calculating a relative displacement amount of an image by comparing first and second image data. 2. A pair of image sensors each comprising a plurality of pixel arrays, respectively receiving a pair of light beams of the same object light and performing photoelectric conversion, and a first image data based on outputs of the pair of image sensors. An image shift calculating means for calculating a relative displacement of the light images on the two image sensors from the second image data, wherein at least one of the one image data includes a plurality of pixels within an initial reference image range. Setting means for changing and setting a reference image range used by the image shift calculating means so as to include a pixel having a large contrast when the object light pattern has a large contrast. Based on the set reference image range, the image shift calculating unit compares the first and second image data to determine the relative position of the image. Focus detection device and calculates the position quantity. 3. The image processing apparatus according to claim 1, wherein the setting unit changes the reference image range to an area around the image data in which a large contrast is localized, in the case of image data in which a large contrast is localized. 3. The focus detection device according to claim 1. 4. The image processing apparatus according to claim 1, wherein the setting unit changes the reference image range so that the image data having a large contrast at an end of the reference image range does not exist at the end. Item 3. The focus detection device according to item 1 or 2.