JPH10341375A - Image-pickup device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the sensitivity of a display image, to facilitate optimum threshold value setting and to reduce the use frequency of a reference image (heat) source by reading an image which becomes the reference for a uniform level, at plural points in the perpendicular direction of arrayed reading elements and discriminating an element in which the difference between the maximum and minimum values of read data obtained for each pixel is extremely small, as a defective element. SOLUTION: When rediscriminating a defective pixel, reference data are sent to a defective element extraction/replacement control part 92. A P-P arithmetic part 13 holds the P-P noise level of each element in the case of picking up the image of a uniform reference heat source. When picking up the image of a target, the P-P noise levels are successively outputted from the P-P arithmetic part 13 synchronizing with data read (A/D conversion) from array-detecting elements. Then, the image which becomes the reference of a uniform level is read at plural points in a perpendicular direction (j) of arrayed reading elements and the element in which the difference between the maximum value MAXS(j) and the minimum value MINS(j) of the read data SD obtained for each pixel is extremely small, is discriminated as the defective element.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image pickup apparatus, and more particularly, to extracting a defective element based on read data of an image serving as a reference by an array-shaped read element, and extracting pixel data of the defective element at the time of imaging a target. The present invention relates to an imaging device that replaces pixel data of another element. A representative example of this type of device is an infrared imaging device, and the imaging characteristics (image quality) of the device greatly depend on the characteristics of the array-shaped reading element.
However, in general, there are variations in the element characteristics of the array-shaped reading element, and the characteristics change due to changes in the operating environment (ambient temperature, power supply voltage, etc.) and aging.
If a reading element (defective element) exceeding the limit is left (used), the image quality is significantly deteriorated. For this reason, when the power of the apparatus is turned on or when the apparatus is in operation (when the image quality is deteriorated, etc.), defective elements in the array-shaped read element are extracted based on read data of a reference image (heat source), and pixel data of the element is extracted. Replacement with pixel data of another element is performed.

[0002]

2. Description of the Related Art FIGS. 19 to 23 are diagrams (1) to (5) for explaining a conventional technique. FIG. 19 shows the configuration of a conventional infrared imaging apparatus. In the figure, reference numeral 50 denotes a target object (target space) which is a subject; 60, a reference heat source used when calibrating the imaging characteristics of the apparatus (such as extraction of a defective element); 70 is a conventional infrared imaging device, and 80 is a monitor such as a CRT.

The reference heat source 60 has, for example, a structure in which Peltier elements are two-dimensionally arranged, and can generate a uniform temperature within a range of room temperature ± 30 ° C. on a rectangular plate-shaped surface to be imaged. The periphery of the imaging surface is covered with a black (non-reflective) case, realizing a state similar to so-called black body radiation. In the infrared imaging device 70, 1 is a polygon mirror for scanning (scanning) a target object, 2 is an imaging lens, 3 is an array detecting element in which m infrared detecting elements are substantially arranged in a line, and 4 is a preamplifier of an image signal ( PA), 5 is an A / D converter, 6
Is a FIFO type image memory (DPRAM or the like) capable of storing at least one screen (frame) of pixel data SD;
7 is a write counter (WC), 8 is a read counter (R
C) and 9 are sensitivity correction units of the array detection element, 10 is a display level / gain adjustment unit that adjusts brightness and contrast of a display image, and 11 is a D / A conversion unit.

Further, 71 is a main control of the apparatus (imaging control C).
1, a main control unit that performs defective pixel extraction / replacement control C2, etc.)
Reference numeral 72 denotes a defective pixel extraction / replacement control unit which detects a defect of the detection element based on the read data of the reference heat source 60 and controls replacement of the defective pixel data by the element. Reference numeral 12 denotes each element when the reference heat source 60 is read. RMS (effective noise) calculation unit for read data, 13 is a PP calculation unit, 14 is a DC calculation unit, and 15 to 17 are the above calculation results δ, P, and DC when the target 50 is read. A comparator (CMP) 18 for comparing the predetermined threshold value (threshold value) RMSTH, PTH, DCTH to determine the presence or absence of a defective element, and an OR gate 18 for generating a replacement enable signal SE for defective pixel data based on the comparison result of each CMP A circuit (O), 19 is a RAM for generating an offset address ± α for reading replacement pixel data, and 20 is an addition for changing a read address of the image memory 6. It is a (+).

At the time of imaging the target object 50, the read signal of the array detecting element 3 is A / D converted in synchronization with the pixel clock GCK, and the obtained pixel data SD is sequentially written to the image memory 6. At the same time, the stored data RD is read out from the image memory 6 in a FIFO mode, and after sensitivity correction, display level and gain (contrast) adjustment,
The video signal VS obtained by the D / A conversion is output to the monitor 80.
Will be displayed.

In such a configuration, if a defective element exists in the array detection element 3, the display image quality is significantly deteriorated. Therefore, the reference heat source 60 is inserted at the time of turning on the power of the apparatus or during operation (when the image quality is deteriorated, etc.), a defective element is extracted based on the read data, and pixel data of the element is replaced with pixel data of another element. Need to be replaced. Conventionally, defective elements are determined based on the effective (RMS) noise, PP noise, and DC noise obtained from the read data of the reference heat source 60, and pixel replacement at the time of imaging is performed. Hereinafter, this will be described in detail.

In FIG. 19, when a reference heat source 60 having a uniform surface temperature T is scanned by the mirror 1, reference image data SD for one screen is obtained. The capture of the reference pixel data SD for one screen starts with i = 1 to m where j = 1, and j = n
Ends with i = 1 to m. FIG. 20A focuses on the reference pixel data S (j) of the i-th element (i-element). The vertical axis indicates the detection level (temperature), and the horizontal axis indicates j (scan).
The number of samplings in the direction is 1 to n. Due to the fact that the surface temperature of the reference heat source 60 is not completely constant, the sensitivity of the i-element, the noise characteristics, and the like, the reference pixel data S (j) has a variation in amplitude as illustrated.

[0008] The effective noise δ (i) of the i element is defined by equation (1).

[0009]

(Equation 1)

Here, S _a (i) is S _a for the i element.
(J) is the average value. The effective noise δ (i) corresponds to a standard deviation. The PP noise P (i) of the i element is defined by equation (2).

[0011]

(Equation 2)

Here, _MAX S (j) is the pixel data S
The maximum value in (j), _MIN S (j), is the minimum value. The DC noise DC (i) of the i element is defined by equation (3).

[0013]

(Equation 3)

Here, S_a(I) shows the current reference temperature T
The average value read this time, S_a(I) 'is the same as the previous
This is the average value of the previous time when the reference temperature T was read. FIG.
(B) shows the configuration of the RMS operation unit 12. Reference heat source 60
At the time of imaging, n reference pixel data for the i element
Data S (j) is cumulatively added to the same address in RAM1.
And then 1 / n and the average value S _a(I). one
On the other hand, the n pieces of reference pixel data S (j)
It is sequentially stored in the storage block corresponding to the element, and then each pixel
Error e = S (j) -S for data_a(I) and even that
Square error e^TwoAre found, and these are the same
Cumulative square error of i element added to dress
Σe^TwoBecomes The above operation is performed for all the elements i (= 1 to m).
Then, the RAM 3 stores the Δe of all the elements.^Two(1)-$ e^Two
(M) is stored.

When the target object 50 is imaged, Δe ² is sequentially read from the RAM 3 in synchronization with the reading of the pixel data RD from the image memory 6, and is read by 1 / (n−1) to obtain the variance δ ^2. At the same time, the obtained variances δ ² are synchronized with the pixel clock GCK, and the registers REG 1 and REG
2 is sequentially shifted and transferred. In this case, now RAM3
When the data reading from the image memory 6 is started one pixel ahead of the data reading from the image memory 6, i
At the timing of processing the pixel data RD (i) of the element, REG2 is the (i−1) th variance δ ² (i−1) one pixel before (upper), and REG1 is the variance δ ² (i) of the i element. ),
Then, the multiplier (×) outputs the (i + 1) -th variance δ ² (i + 1) one pixel later (lower). Here, the variance δ ² (i) is used to determine whether or not the pixel data RD (i) of the i element is to be replaced. On the other hand, CMP is the variance δ of the upper pixel.
² (i-1) is compared with the variance δ ² (i + 1) of the lower pixel. If δ ² (i-1) <δ ² (i + 1), an upper replacement signal US = 1 is output. In other cases, lower substitution (U
S = 0).

Here, the variance δ ² (i) is used as the effective noise δ (i) in the equation (1), in order to avoid the route calculation. Hereinafter, the variance δ ² (i) is referred to as RMS noise δ (i) for simplicity of explanation. FIG.
(A) shows the configuration of the PP calculation unit 13. Reference heat source 60
At the time of imaging, CMP1 is the input reference pixel data a
Is compared with the storage data b of the RAM1, and if a> b, the data write signal WE = 1 of the RAM1 is output. on the other hand,
The CMP2 compares the input reference pixel data a with the storage data b of the RAM1, and outputs a data write signal WE = 1 for the RAM2 when a <b. Therefore, of the n pieces of pixel data S (j) for the i element, the maximum value _MAX S (j) is stored in the RAM 1 and the minimum value _MIN S (j) is stored in the RAM 1.
2 is stored. When the above operation is performed for i = 1 to m, the maximum value _MAX S (j) and the minimum value _MIN S (j) for all the elements i (= 1 to m) are stored in the RAMs 1 and 2, respectively.

When the target object 50 is imaged, RA is synchronized with the reading of the pixel data RD of the i-element from the image memory 6.
_MAX S (j) and _MIN S (j) of the i element are read from M1 and M2, respectively, and PP noise P (i) = _MAX S (j)-
_MIN S (j) is determined. The value of P (i) is used to determine whether or not the pixel data RD (i) of the i element is to be replaced. FIG.
1 (B) shows the configuration of the DC operation unit 14.

At the time of imaging of the reference heat source 60 at a certain point in time, the n pieces of pixel data S (j) obtained when the uniform temperature T is read by the i element are cumulatively added to the same address in the RAM 1 (one bit lower at the same time).シ フ ト) by the shift, and finally the average value S _a (i). The above operation is performed for all elements i
(= 1 to m), the RAM 1 stores the average values S _a (1) to S _a (m) of all the elements, respectively. All these data are synchronized with the frame clock FCK and stored in the RAM 2
To the previous average value S _a (i) ′.

At the other time, the same processing is performed for the same temperature T, and the current average value S is stored in the RAM 1.
_a (1) to S _a (m) are stored. When the target object 50 is imaged, the pixel data RD of the i element from the image memory 6
The average values S _a (i) and S _a (i) ′ of the i-element are read from the RAMs 1 and 2 in synchronization with the reading of (i), respectively.
DC noise DC (i) = S _a (i) −S _a (i) ′ is obtained. The value of DC (i) is the pixel data RD of the i element.
It is used to determine the presence or absence of replacement in (i).

Returning to FIG. 19, at the time of imaging the target object 50,
The CMP 15 compares the RMS noise δ (i) of the i-element with a predetermined threshold value RMSTH, and outputs a defect detection signal RED = 1 when δ (i)> RMSTH. If the RMS noise is large, the display brightness (or color) changes finely and almost randomly even when a target portion at the same temperature is imaged, which causes flickering of the displayed image.

The CMP 16 has a PP noise P (i)
And a predetermined threshold value PTH, and outputs a defect detection signal PED = 1 when P (i)> PTH. Even when the PP noise is large, the displayed image flickers. Also, CMP17 is D
C noise DC (i) is compared with a predetermined threshold value DCTH, and D
When C (i)> DCTH, a defect detection signal DED = 1 is output. If this DC noise is large, even if a target portion having the same temperature is imaged, its display luminance (or color) gradually changes over time, which may cause unsightly horizontal stripes in the display image.

When any of the above defect detection signals is generated, the OR gate circuit O18 outputs a replacement enable signal SE = 1 for the defective pixel. The RAM 19 reads the offset address value ± α according to the address input signal US = 0/1 at that time because of SE = 1. + Α is the read pixel data RD
(I) A relative address indicating the lower replacement pixel data RD (i + 1), and -α is a relative address indicating the upper replacement pixel data RD (i-1). Thus, a high quality image can be displayed on the monitor 80 by replacing the pixel data of the defective element with the upper or lower pixel data.

FIG. 22 shows the relationship between RMS noise and defective pixel replacement control. In FIG. 22A, the horizontal axis indicates the elements i = 1 to m (= 18), and the vertical axis indicates the RMS noise level. By setting the threshold value RMSTH to a or b, the noise level (flicker) of the displayed image can be suppressed as desired. In FIG. 22B, for example, when the threshold value RMSTH is set to a relatively high value a, the total number of defective elements i = 8 and 14 is 2
One. Pixel data R (8) of defective element i = 8 is RM
Pixel data R of lower element i = 9 with smaller S noise
The pixel data R (14) of which the lower element is replaced by (9) and the defective element i = 14 is the upper element i having smaller RMS noise.
= 13 pixel data R (13). Therefore, pixel data having a large RMS noise is deleted (replaced) from the display image, and the image quality is improved from the viewpoint of the RMS noise.

In FIG. 19, at the same time, P> PT
The determination output of H, DC> DCTH is also a cause of replacement.
However, in this case, the determination as to whether to perform the upper replacement or the lower replacement is made according to the magnitude of the RMS noise. FIG.
In (C), when the threshold value RMSTH is set to a lower value b, the defective element i = 3, 5, 8, 10, 12, 14, 1, 1
The total is 6, 17 for a total of eight. In the same manner as described above, each defective pixel data is replaced with an upper / lower pixel.
The more strict the MSTH, PTH, and DCTH, the better the image quality.

FIG. 23 shows an example of a display image. In FIG. 23A, a pyramid-shaped target having a uniform surface temperature is displayed on the monitor 80. For example, a low-temperature background is displayed in black, and a high-temperature target is displayed in white. FIG. 23 (B) corresponds to the defective element detection state of FIG. 22 (B), and the defective pixel data RD (8) for one row of display is replaced with the defective pixel data RD for one row.
(14) has been replaced. Accordingly, the variation (fluctuation) in luminance (color) as occurs in the eighth and fourteenth rows of the display screen is improved, but the smoothness of the inclination of the shoulder of the pyramid is lost from the viewpoint of the shape.

FIG. 23 (C) corresponds to the defective element detection state of FIG. 22 (C), and the shape of the target object is changed as a result of the upper / lower replacement of defective pixel data in a larger number of display rows. It is different from its original form. Under such circumstances,
Conventionally, the reference heat source 60 is inserted at the time of turning on the power of the apparatus or during operation (when the image quality is deteriorated, etc.), and each threshold is set by trial and error in which the displayed image is visually evaluated.

[0027]

However, according to the above prior art, defective elements are detected based on the magnitude of RMS, PP and DC noise, so that pixel replacement of an insensitive or insensitive element cannot be performed. Was. For this reason, if the sensitivity of a certain element is so small that the sensitivity cannot be corrected, sufficient luminance (color) and contrast (color change) cannot be obtained in the display pixels in that row.

In the above-mentioned prior art, a threshold value for detecting a defective element is set by visual evaluation of a reference image or the like.
In addition to trial and error to determine which threshold has an effect on the image quality, it is not only difficult to set an optimal threshold, but also the threshold setting operation has become extremely complicated.
Further, in the above-described conventional technique, it is necessary to insert the reference heat source 60 every time the image pickup apparatus is calibrated, so that the image of the target object is interrupted during that time. In order to automatically insert and remove the reference heat source 60, it is necessary to always provide mechanical parts and an electric circuit for the insertion and removal of the reference heat source 60, which leads to an increase in size and cost of the apparatus.

The above-mentioned problems are not necessarily unique to the infrared imaging apparatus, but may occur in this type of imaging apparatus using ordinary light (such as CCD) or ultrasonic waves.
The present invention has been made in view of the above-mentioned problems of the prior art,
An object of the present invention is to provide an imaging device in which the sensitivity of a display image is improved, an optimum threshold value for defective pixel replacement is easily set, and the frequency of use of a reference image (heat) source is preferably significantly reduced. Is to do.

[0030]

The above-mentioned object can be attained by, for example, the structure shown in FIGS. That is, the present invention (1)
The image pickup device (for example, an infrared image pickup device) extracts a defective element based on read data of a reference image (for example, the reference heat source 60) by the array-shaped reading element 3, and extracts pixel data of the defective element at the time of imaging a target. Is replaced with pixel data of another element, an image serving as a reference of a uniform level is read at a plurality of points in the vertical (j) direction of the array-like reading element, and a maximum value _{MAX of} read data SD obtained for each element is read. An element in which the difference between S (j) and the minimum value _MIN S (j) is extremely small is determined as a defective element.

Normally, if an element having a certain degree of sensitivity is used, even when an image serving as a reference for a uniform level is read, a delicate level of unevenness on the surface and the operating environment of the element (temperature, bias voltage, noise signal component, etc.) ) And the maximum value _MAX S (j) and the minimum value _MIN S of the read data in response to
There may be a predetermined difference or more from (j). On the other hand, an element having no sensitivity can produce only an extremely small difference (for example, a difference of about the quantization error due to the A / D converter). Therefore, by determining such an element as a defective element, pixel replacement of an insensitive element can be effectively performed, thereby improving the sensitivity of a displayed image.

The above-mentioned problem can be solved, for example, by the structure shown in FIG. That is, in the image pickup apparatus of the present invention (2), in the image pickup apparatus based on the above-mentioned premise, a plurality of reference images of two different levels are read in the vertical direction of the array-like reading element, and the read data of the one level is read. The difference between the average value T _Ha (i) for each element and the average value T _La (i) for each element of the other level of read data is sufficiently smaller than the value corresponding to the difference between the different levels. This is determined as a defective element.

When an image serving as a reference at a different level is read, an element having low sensitivity cannot detect a level difference sufficiently smaller than the reference level difference. The same applies to an element having no sensitivity. Therefore, by determining such an element as a defective element, pixel replacement can be effectively performed not only for the element having no sensitivity but also for an element having a low sensitivity, thereby improving the sensitivity of a display image.

The above-mentioned problem can be solved, for example, by the structure shown in FIG. That is, in the image pickup apparatus of the present invention (3), in the image pickup apparatus based on the above, one or two or more types of noise levels for each element i (= 1 to m) based on the read data SD of the reference image {for example, Effective noise δ (i), P−
Noise level detectors 12 to 14 for detecting P peak noise P (i), DC noise DC (i)};
A plurality of thresholds RMSTH1 to RMSTHk, PTH1 to PMS1 to RMSTH1 to RMSTHk having different sizes provided for different types of noise levels, respectively.
PTHk, DCTH1 to DCTHk, and the strictest threshold value for each of the one or more noise levels are selected, and the number of defective elements in which any of the detected noise levels exceeds the corresponding threshold value is determined. A threshold control unit 23 that counts all the elements and switches the selection of each of the thresholds to a looser one until the obtained number of defective elements becomes a predetermined number or less.

According to the present invention (3), a plurality of thresholds can be commonly used for each device irrespective of the variation in the characteristics of the array-shaped reading element, and such a device is easy to manufacture. Also, by simply setting the total number of defective elements of the array-like reading element, an optimum threshold setting (selection) state for obtaining an optimum pixel replacement state (an easy-to-view image) is automatically obtained. Easy.

The above-mentioned problem can be solved, for example, by a structure similar to that shown in FIG. That is, the imaging device of the present invention (4)
In the imaging apparatus as the premise, a noise level detection unit that detects one or two or more types of noise levels for each element based on read data of a reference image;
A plurality of thresholds having different magnitudes respectively provided for more than one kind of noise level, and a loosest threshold for each of the one or more kinds of noise levels are selected, and any one of the detected noise levels corresponds to the selected one. A threshold control unit that counts the number of defective elements exceeding the threshold for all elements of the array-shaped reading element and switches selection of each of the thresholds to a stricter one until the obtained number of defective elements becomes a predetermined number or more. is there.

In the present invention (4), contrary to the above-described present invention (3), the threshold control unit 23 selects the loosest threshold and selects each threshold until the number of defective elements reaches a predetermined number or more. Switch to the tougher one. According to such a method, the same array-shaped reading element is set to the same threshold value setting state as that of the present invention (3). The above-mentioned problem is solved by, for example, the configuration shown in FIG. That is, in the imaging device according to the present invention (5), in the imaging device as the premise, a noise level detection unit that detects a plurality of types of noise levels for each element based on read data of a reference image; A plurality of thresholds having different sizes provided for each noise level and the number of defective elements in which each of the detected noise levels exceeds each corresponding threshold were counted for all the elements in the array-like reading element, respectively. A maximum noise type detection unit 24A that extracts a noise level type corresponding to the maximum count value by comparing each count value, and selects a strictest threshold value for each of the plurality of types of noise levels, and The number of defective elements whose noise level exceeds the corresponding threshold value is counted for all elements of the array-like reading element, and the obtained number of defective elements is a predetermined number. Until a lower, each time, the maximum noise type detector is one that includes a threshold controller 24 to switch towards loose selection threshold noise level species detected.

According to the present invention (5), the selection is started from the strictest threshold, the noise level type having the maximum error count is detected each time, and the selection of the threshold is loosened preferentially from the noise level type. With this configuration, pixel replacement can be performed in a balanced manner for each noise level type. The above-mentioned problem is solved by, for example, a configuration similar to that of FIG. That is, in the imaging device according to the present invention (6), in the imaging device as the premise, a noise level detection unit that detects a plurality of types of noise levels for each element based on read data of a reference image; A plurality of thresholds having different sizes provided for each noise level and the number of defective elements in which each of the detected noise levels exceeds each corresponding threshold were counted for all the elements in the array-like reading element, respectively. A minimum noise type detection unit that extracts a noise level type corresponding to the minimum count value by comparing each count value; and selecting a loosest threshold value for each of the plurality of types of noise levels, and selecting one of the detected noise levels. The number of defective elements whose level exceeds the corresponding threshold value is counted for all of the array read elements, and the number of obtained defective elements is equal to or more than a predetermined number. Until, in each case, the minimum noise type detector is one that includes a threshold controller for switching towards strict selection threshold noise level species detected.

In the present invention (6), contrary to the above-mentioned present invention (5), the selection is started from the loosest threshold value, the noise level type having the smallest error count is detected each time, and the noise level is detected. With the configuration in which the selection of the threshold value is preferentially switched from the type to the stricter one, pixel replacement can be performed in a balanced manner for each noise level type. The above-mentioned problem is solved by, for example, the configuration shown in FIG. That is, in the imaging device according to the present invention (7), in the imaging device based on the above premise, a noise level detection unit that detects a predetermined noise level for each element based on read data of a reference image,
A threshold for determining the presence or absence of a defect in each element by multiplying the average value of all the elements obtained for the detected noise level of each element by a predetermined coefficient and comparing with the detected noise level of each element. And a threshold generation unit that generates the threshold.

According to the conventional method in which a predetermined threshold value is set externally, the number of pixel replacements caused by a certain noise level type becomes extremely large in accordance with the characteristic variation of the array-like reading element. An extremely low state may occur. According to the present invention (7), the threshold value is generated by multiplying the average value of all the elements of the noise level by a predetermined coefficient to generate the threshold value. For example, the threshold value of the array element having a large RMS noise is generated slightly larger. The threshold value of the array element having a small RMS noise is set to be slightly smaller, so that a balanced pixel replacement can be performed for each noise level type.

Preferably, in the present invention (8), instead of providing a plurality of thresholds having different magnitudes for one or more noise levels in the present inventions (3) to (6), Alternatively, a plurality of coefficients having different sizes provided for each of two or more noise levels and an average value of all the elements are obtained for each of the noise levels of each element detected for one or more noise levels. A threshold generation unit that generates a threshold by multiplying each of the obtained average values by the selected coefficient, and the threshold control unit selects a coefficient instead of selecting a threshold.

Therefore, regardless of the characteristic variation of the array reading element, an optimum pixel replacement state balanced for each noise level type can be easily obtained. The above-mentioned problem is solved by, for example, the configuration shown in FIG. That is, the present invention (9)
In the imaging apparatus described above, in the imaging apparatus as the premise, each read data SD of a reference image is corrected with predetermined correction data XG for each element separately obtained.

In the conventional method of directly evaluating the noise level for each read data of the reference image, for example, R
As a result of applying a predetermined gain to the pixel data of the element that has passed the evaluation of the MS noise by various processes (sensitivity correction process, contrast increase process, etc.) for the subsequent screen display, the RMS noise increases at the same time, The screen may be difficult to see.

According to the present invention (9), each read data of the reference image is corrected in advance with separately obtained predetermined correction data (for example, sensitivity correction data) for each element, so that each corrected data can be obtained. The noise level can be evaluated for the read data, and the defective pixel can be determined with a value more faithful to the display data. The above-mentioned problem is solved by, for example, the configuration shown in FIG. That is, in the imaging device of the present invention (10), signal levels (various detection noise levels, sensitivity detection levels, etc.) for extracting defective elements detected for each element can be separately obtained from the imaging device on the premise above. The correction is performed using predetermined correction data (for example, sensitivity correction data) for each element.

According to the present invention (10), the presence or absence of a defective element can be determined based on the noise level after correction, and the defective pixel can be determined with a value more faithful to the display data. The above-mentioned problem is solved by, for example, the configuration shown in FIG. That is, the present invention (1)
The imaging device of 1) is an imaging device based on the above premise,
An image serving as a reference for a uniform level is read at a plurality of points in the vertical direction of the array-like reading element, and the average value Sa (i) ′ for all the elements from the previous reading data is calculated for each element.
′ Is determined as a defective element if the difference between the value obtained by subtracting the average value Saa for all the elements from the average value Sa (i) of each element of the current read data exceeds the predetermined threshold value. is there.

The present invention (11) relates to improving the detection of the DC noise level. The DC noise level represents the variation of the detection level when an image of the same level is read by the same element. In the conventional method, if the level of the reference image fluctuates, the variation is directly applied to the element. There is a disadvantage that the detected DC noise level is included. In this regard, in the present invention (11), the average value for all the elements is subtracted from the average value for each element of the read data of the image serving as the reference of the uniform level. Here, each average value for each element may indicate a detection value that varies with the level of the reference image at that time in accordance with the fluctuation of the detection level (detection ability) for each element. However, when the average value of all the elements is obtained, the influence of the variation of each element is canceled out. As a result, the average value faithfully represents a representative level of the reference image at that time. Therefore, by subtracting the average value for all elements from the average value for each element, a net variation for each element is obtained.
Therefore, when this is performed between the previous time and the current time, and the difference between them is obtained, even if the level of the previous time and the current time of the reference image is differently changed regardless of the fluctuation of the level of the previous time and the current time of the reference image, , A net change (DC noise level) between the previous time and the current time is obtained for each element.

The above problem can be solved, for example, by the structure shown in FIG. That is, in the image pickup apparatus of the present invention (12), in the image pickup apparatus based on the above-mentioned premise, an image serving as a reference of a uniform level is read at a plurality of points in the vertical direction of the array-like reading element, and the p-th (p. Are read data R1 to R3, R4 to R of the maximum group and the minimum value up to the qth (the third in the figure) from the maximum group up to the third group.
6 is extracted, and an element in which a difference between predetermined read data in the maximum group and predetermined read data in the minimum group exceeds a predetermined threshold value is determined as a defective element.

The present invention (12) relates to the improvement of the PP noise level detection. The PP noise level indicates a difference between the maximum value and the minimum value when an image at the same level is read by the same element. In general, such a maximum value and / or a minimum value are superimposed with some noise signal. In many cases, it is difficult to accurately detect the PP noise level. In this regard, according to the present invention (12), since the signal level such as the maximum value and / or the minimum value in which the noise signal is superimposed can be excluded from the evaluation target, an accurate PP noise level can be detected. .

The above problem can be solved, for example, by the structure shown in FIG. That is, in the image pickup apparatus of the present invention (13), in the image pickup apparatus based on the above-mentioned premise, an image serving as a reference of a uniform level is read at a plurality of points in the vertical direction of the array-like reading element, and from the maximum value to the p-th element for each element. Each of the read data of the minimum group from the maximum group and the minimum value to the q-th is extracted, and the difference between the average value of the read data of the maximum group and the average value of the read data of the minimum group exceeds a predetermined threshold. The element is determined as a defective element.

Therefore, the influence of the noise component is canceled (or mitigated) in each average value, and an accurate PP noise level can be detected. The above problem is solved by, for example, the configuration of FIG. That is, in the imaging device according to the present invention (14), the pixel data of the defective element at the time of imaging the target is replaced (interpolated) by the average value of the pixel data of the elements on both sides of the defective element in the imaging device on which the above-mentioned premise is taken. Things.

Therefore, the defective pixel portion is smoothly interpolated,
An easy-to-view image with less discomfort can be obtained. The above-mentioned problem can be solved by, for example, the configuration shown in FIGS. That is, in the imaging apparatus according to the present invention (15), in the imaging apparatus based on the above-described premise, read data of a uniform level continuous in the element array direction of the array-shaped reading element and / or a direction forming a predetermined angle at the time of imaging the target object. When the columns are detected, these are stored in the memory as read data columns of the reference image, and are used for extracting defective elements.

In the present invention (15), imaging data equivalent to the imaging data of the reference image is extracted from the imaging data of the target, and these are stored in the memory, and are used in the subsequent defective element extraction processing. Therefore, the frequency of use of the original reference image (heat) source can be greatly reduced, the apparatus can be simplified, and the re-calibration of the imaging characteristics can be performed without interrupting the imaging of the target.

Preferably, in the present invention (16),
In the present invention (15), read data strings of the same and different levels extracted from the same and different imaging frames are stored in the memory. Therefore, there are many opportunities to obtain valid reference data, and it is possible to recalibrate the imaging characteristics without interrupting the imaging of the target.

[0054]

Preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings. FIG. 1 is a diagram illustrating a schematic configuration of an infrared imaging device according to an embodiment. The same or corresponding parts as in FIG. 19 are denoted by the same reference numerals, and description thereof will be omitted. In FIG. 1, reference numeral 90 denotes an infrared imaging apparatus according to the embodiment, 91 denotes a main control unit, 92 denotes a defective pixel extraction / replacement control unit, and 93 denotes a reference data extraction unit.

The defective pixel extraction / replacement control unit 92
Is drawn in the same manner as the defective pixel extraction / replacement control unit 72 in FIG. 19, but the characteristic part of the present invention will be described later.
The main control unit 91 performs reference data extraction control C3 in addition to imaging control C1 and defective element extraction / replacement control C2. The reference data extraction unit 93 extracts a portion of pixel data equivalent to the image of the reference heat source 60 at a uniform level from the pixel data SD at the time of capturing the target object, stores these in a memory, and retains them as reference data RFD. At the same time, when it is necessary to re-determine the defective element, the reference data RFD is provided to the defective element extraction / replacement control unit 92. Therefore, in this case, the reference heat source 60 is used when the power of the apparatus is turned on, but is not always necessary during readjustment on the way.

Hereinafter, embodiments of each characteristic portion according to the present invention will be described. However, it goes without saying that these may be used alone or in combination in an apparatus. FIG. 2 shows the first
FIG. 4 is a diagram illustrating the infrared imaging apparatus according to the embodiment, and relates to control for extracting and replacing defective pixels of an insensitive element. FIG.
In (A), reference numeral 21 denotes a first element for determining an insensitive element.
Is a sensitivity determination unit. The PP calculation unit 13 holds the PP noise levels P (1) to P (m) of each element when the uniform reference heat source 60 is imaged. When the target object 50 is imaged, PP noise levels P (1) to P (m) are sequentially output from the PP calculation unit 13 in synchronization with data reading (A / D conversion) from the array detection element 3. You. The comparator CMP compares the PP noise levels P (1) to P (m) with the first sensitivity threshold value KTH to obtain P <K
In the case of TH, it is determined that there is no sensitivity, and the first sensitivity error detection signal KED = 1 is output. The pixel data SD of the element for which KED = 1 is replaced with the pixel data SD of another element.

FIG. 2B is a timing chart of the first sensitivity detection / determination processing. The vertical axis represents the detection level when the reference heat source 60 at the uniform level is imaged, and the horizontal axis represents the j (scan) direction. For example, since the element i = 3 has an appropriate sensitivity, it has a detection level that faithfully represents the unevenness of the reference heat source temperature. PP noise level of the element P (3) =
MAX S (j) -MIN S (j), where P (3)
> KTH, it is not determined to be a defective element. Also, since P (3) <PTH, it is not determined as a defective element even from the PP noise level.

On the other hand, the output of element i = 5 hardly changes. Even if there is a change, it is a quantization error in the A / D converter 5 (that is, the least significant bit of the pixel data SD changes). In this case, it is determined that there is no sensitivity (or the sensitivity is extremely low) by P (5) <KTH, and the element i = 5
Is determined to be a defective element. 3 and 4 are views (1) and (2) for explaining an infrared imaging device according to the second embodiment.
The present invention relates to defective pixel extraction / replacement control of an element having low sensitivity.

In FIG. 3, reference numeral 22 denotes a second sensitivity judging section for reading the reference heat sources 60 at different temperatures and detecting an element having low sensitivity based on the obtained read data. In a second sensitivity determining section 22, the circuit consisting of RAM1 is an average value T _Ha of the first every element on the basis of the read data of the reference temperature T _H at a certain time _{(1) ~T Ha (m)} , Keep these. Circuit consisting RAM2 obtains the second reference temperature T _L at other times the mean value T _La (1) for each element based on each read data _{(<T H) ~T La (} m), holding them I do.

At the time of imaging the target object 50, the RAM 1 reads T from the RAM 1 in synchronization with the data reading from the array detecting element 3.
_Ha (i) and T _La (i) are sequentially read from the RAM 2, and the difference signal D _Ta (i) = | T _Ha (i) −T
_La (i) | are obtained in order. The comparator CMP is
By comparing the difference signal DT _a of each element (1) ~DT _a (m) and a second sensitivity threshold KTH, DT _a (i)
If <KTH, the sensitivity is determined to be small, and the second sensitivity error detection signal KED = 1 is output. Incidentally, the magnitude of the second sensitivity threshold KTH in this case is determined according to the difference between the reference temperature T _H, T _L. And
The pixel data SD of the element for which the second sensitivity error detection signal KED = 1 has been replaced with the pixel data SD of another element.

FIG. 4 is a timing chart of the second sensitivity detection / judgment process. The vertical axis indicates different levels of reference heat source 6
The detection level at the time of imaging each of 0 is shown, and the horizontal axis is the j (scan) direction and the temperature T of the reference heat source 60. Element i = 3
Has an appropriate sensitivity, and the detection level when the heat source temperature T is linearly changed becomes a characteristic. On the other hand, the element i = 5 has low sensitivity, and the detection level when the heat source temperature T is linearly changed becomes a characteristic. However, in the present embodiment, two temperatures _TH and _TL are imaged twice. At that time, the average of each read data in the scan (j) direction is taken, and these are taken as T _Ha (3), T _Ha (5), T _Ha
La (3) and T _La (5). DT of element i = 3
_a (3) = | T _Ha (3) −T _La (3) |> KTH, and it is not determined to be a defective element. On the other hand, D of element i = 5
T _a (5) = | T _Ha (5) −T _La (5) | <KTH, and it is determined to be a defective element. This second sensitivity threshold KT
An example of the magnitude of H is such that the sensitivity of the element cannot be corrected to a required level with the finite (correcting too much an image is distorted) sensitivity correction data in the sensitivity correction unit 9 of FIG.

FIG. 5 is a diagram for explaining an infrared imaging apparatus according to the third embodiment. An optimum display image (pixel replacement state) is obtained by automatically setting thresholds for noise levels of RMS, PP, and DC. Is obtained. In the figure, reference numeral 23 denotes a threshold control unit for automatically setting each threshold for extracting defective elements based on the evaluation of the total number of replacement pixels of the array detection element 3.

RMS, PP, DC operation units 12 to 1
Reference numeral 4 denotes each detection noise level δ of RMS, PP, and DC for each element based on read data obtained by reading the reference heat source 60 in advance.
(I), P (i), and DC (i). The same applies to the sensitivity determination units 21 and 22. At this point, a defective element relating to the sensitivity can be determined. On the other hand, RMS, PP, DC
The threshold value has not yet been determined for each of the noise levels, and therefore the defective element has not been determined yet. That is, RMS, P
A plurality of thresholds are previously provided for each of the noise levels -P and DC. For RMS noise, threshold values RMSTH1 to RMSTHk are provided.
.. <RMSTHk. TH1 <RMSTH2 <. The same applies to the noise levels of PP and DC.

In the threshold control section 23, CTR1 is a counter for counting the total number of defective elements of the array detection element 3, CTR2 is a counter for generating a common threshold selection signal STH, and TG1 is a timing control section for threshold control. At the start of the threshold control, reset signals R1 and R2 are output from the timing generator TG1, and the counters CTR1 and R2 are reset.
CTR2 is reset together. Due to the reset of CTR2, the selectors SEL1 to SEL3 each have the minimum threshold RM.
Select STH1, PTH1, DCTH1, which is the selection of the most severe threshold for each detected noise level.
After that, the noise level of each element is sequentially read from each of the arithmetic units 12 to 14 and compared with the selected threshold value. If δ (i)> RMSTH1, the error detection signal R
If ED = 1, P (i)> PTH1, error detection signal PE
If D = 1, DC (i)> DCTH1, error detection signal D
ED = 1. At the same time, there are cases where the error detection signal KED = 1 and / or the error detection signal KED = 1.
Each error detection signal is logically ORed by the OR gate circuit 18, and the output signal SE is, at this point, the threshold control unit 2
3 is used.

The timing generator TG1 generates an inspection clock SC1 for each processing of one element. When the signal SE = 1 at the processing timing of a certain element, the defective element number counter CTR1 is incremented by one. When SE = 0, the count value does not change. Thus, when processing of all elements is completed,
The CTR 1 counts the total number of defective elements for the array detection element 3.

At that time, the comparator CMP1 is set to CT
The count value Q of R1 is compared with the threshold CLTH of the number of defective elements. When Q> CLTH, a count-over signal OV = 1 is output. The signal OV = 1 is notified to the TG1, whereby the TG1 outputs one clock signal SC2 and counts up one CTR2 for one cycle of detection. Further, TG1 outputs a reset signal R1 to initialize CTR1, and again enters the counting mode for the number of defective elements. On the other hand, since the contents of CTR2 are incremented by one, SEL1 to SEL3 have a threshold RM one step lower than the previous time.
STH2, PTH2, and DCTH2 are selected respectively.

This process is repeated, and even if the processing of all the elements in a certain cycle is completed, if the count value Q of CTR1 does not exceed the threshold value CLTH, the threshold control ends, and the threshold selection signal STH of CTR2 becomes Fixed in state. Therefore, a pixel replacement state in which the total number of defective pixel replacements is limited to the desired threshold value CLTH is automatically obtained.

It is to be noted that the above-mentioned threshold value relating to, for example, RMS noise is set to RMSTH1>RMSTH2>.
In this case, the threshold control unit 23 selects the loosest threshold. The other thresholds PTH and DCTH are arranged in the same manner. In this case, the threshold control ends when the count value (the number of defective elements) Q of the CTR 1 exceeds the predetermined threshold CLTH, and the same array detection elements 3 settle down to a threshold setting substantially similar to the above.

FIG. 6 is a diagram for explaining an infrared imaging apparatus according to the fourth embodiment. Among the numbers of detection errors of RMS, PP, and DC noise, the threshold is preferentially set in descending order of the number of errors. The case where loosening is performed to obtain a fine and finally balanced defective pixel replacement state for each error type is shown. In the figure, reference numeral 24 denotes a threshold control unit in this case, and reference numeral 24A denotes a maximum noise type detection unit.

The threshold control operation including the defective element number counter CTR1, the comparator CMP1, and the timing generator TG1 may be basically the same as described above. In the maximum noise type detector 24A, the counter CTR5 outputs a reset signal R
It is reset by 1 and counts the number of occurrences RQ of the error detection signal RED related to RMS noise. Similarly, C
TR6 counts the number of occurrences PQ of the error detection signal PED related to PP noise, and CTR7 counts the number of occurrences DQ of the error detection signal DED related to DC noise. The comparators CMP2 to CMP4 count the respective count values RQ, P
The decoder DEC detects the maximum count value of Q and DQ, and outputs one of the count activation signals a / b / c of the threshold selection corresponding to the maximum count value RQ / PQ / DQ, respectively. Set to logic 1 level.

For example, when the error count value RQ of the RMS noise is the maximum, A = 1 and RQ due to RQ> PQ
> DQ makes C = 1, and the count energizing signal a becomes A
= 1 and C = 1, then a = 1. Also P
When the error count value PQ of -P noise is the maximum, RQ
A = 0 due to <PQ, and B = 1 due to PQ> DQ, and the count energizing signal b becomes b = 1 by decoding A = 0 and B = 1. When the error count value DQ of the DC noise is the maximum, C = 0 when RQ <DQ, and B = 0 when PQ <DQ, and the count energizing signal c
Becomes c = 1 by decoding of C = 0 and B = 0.

In the threshold value control section 24, the comparator C
When MP1 detects the number of defective elements exceeding the limit CLTH, when a = 1, the threshold selection counter C for RMS noise is set.
TR2 is incremented by 1, and when b = 1, the threshold selection counter CTR3 for PP noise is incremented by 1, and when c = 1, the threshold selection counter CTR4 for DC noise is incremented by 1.
Therefore, the threshold value of the noise having the largest error count value is preferentially alleviated, whereby the cause of the pixel replacement is made fair (balanced), and the required defective pixel replacement is carried out with the minimum necessary threshold value relaxation for the entire apparatus. The state is obtained.

If the maximum error count value is not found (that is, a to c = 0, etc.), a predetermined priority order (preferably, a cyclic priority order that the previous selection is not selected this time) is used. Each time, one of the thresholds is relaxed. By the way, if the above-mentioned threshold value relating to, for example, RMS noise is set to the relationship of RMSTH1>RMSTH2...> RMSTHk, the threshold value control unit 24 will select from the loosest threshold value. The other thresholds PTH and DCTH are arranged in the same manner. In this case, the threshold control ends when the count value (the number of defective elements) Q of the CTR 1 exceeds the predetermined threshold CLTH.
Is set to a threshold value substantially similar to the above.

However, in this case, the CMP2 to CMP4 operate to detect the one with the smallest error count value and make the threshold corresponding thereto one step stricter. Also, a minimum error count value cannot be found (ie, a to c = 0, etc.).
If so, a predetermined priority (preferably a cyclic priority)
, One of the thresholds is relaxed each time. FIG. 7 is a diagram illustrating an infrared imaging device according to a fifth embodiment.
A case is shown where thresholds for RMS, PP, and DC noise are optimally set based on detection characteristics (average characteristics of all the elements) of the array detection element 3 itself.

In the figure, 25 to 27 are RMS, P-
This is an average value calculation unit that calculates an average value of P and DC noise for all elements (i = 1 to m). For example, focusing on the RMS noise, when controlling the threshold value, an average value of the RMS noises δ (1) to δ (m) is obtained in advance, and the average value is stored in the register R.
Hold at EGR. At the time of defective element extraction (at the time of threshold value control), a desired threshold value RMSTH is generated by multiplying this average value by an appropriate coefficient (for example, 1.5).

Since the generation threshold value RMSTH is generated based on an average value specific to the array detection element 3,
The values reflect the characteristics of the individual array detection elements 3 having various variations. That is, when the array detection element 3 having a large RMS noise is employed, the threshold RMS
TH is also generated to be relatively large, so that the number of defective elements due to RMS noise does not increase remarkably. Also RM
When the array detection element 3 having a small S noise is adopted, the threshold value RMSTH is also generated at a small value, so that the RMS
The number of defective elements caused by noise does not decrease remarkably. That is, the number of defective elements caused by the RMS noise is made uniform. The same applies to other PP and DC noise.

Preferably, a plurality of thresholds having different magnitudes can be generated for each of RMS, PP and DC noises, and the same threshold control as in FIG. 5 or FIG. 6 is performed. FIG. 7 shows a case where the threshold control unit 24 of FIG. 6 is provided.
Here, instead of providing a plurality of thresholds having different sizes, a plurality of coefficients having different sizes (for example, 1.2,
1.4,... 2.2). Such devices are easy to manufacture, since the multiple coefficients can be the same for any device.

As in the case of FIG. 6 (or FIG. 5), the threshold value control unit 24 starts the selection from the strictest coefficient of 1.2, and the total number of defective elements of the array detecting element 3 becomes a predetermined number. RMS, PP and / or DC until below
The noise coefficient is gradually reduced. Alternatively, the threshold control unit 24 starts the selection from the loosest coefficient of 2.2, and keeps the RM until the total number of defective elements of the array detection element 3 exceeds a predetermined value.
The coefficients of S, P-P and / or DC noise are made stricter sequentially.

FIGS. 8 and 9 are views (1) and (2) for explaining an infrared imaging apparatus according to the sixth embodiment.
The figure shows a case where pixel data obtained at the time of imaging of 0 or a noise level detected based on the pixel data is multiplied by predetermined correction data. The noise level detected based on the read data of the reference heat source 60 faithfully represents the noise level of each element i (= 1 to m) itself. On the other hand, the image data SD of the target 50 read by each element is subjected to sensitivity correction by the sensitivity correction unit 9 in FIG. 1, and if necessary, the display level (gain) and the display gain (color) by the display level / gain adjustment unit 10. The level is adjusted (e.g., contrast) and displayed on the monitor 80.

In such a configuration, for example, a large gain is applied to the imaging data SD (8) of the element i = 8 having a relatively low sensitivity, and the imaging data SD (9) of the element i = 9 having a relatively high sensitivity. Is multiplied by a smaller gain, and sensitivity correction for obtaining uniform display sensitivity is performed.
On the other hand, as a result, the RMS, P-
The noise components of P and DC substantially increase, and the display data SD
Each noise component of (9) is substantially reduced. For this reason, there is a difference between the case where the evaluation of the noise level is performed on the reference data before the sensitivity correction and the case where the evaluation is performed on the reference data after the sensitivity correction. When the evaluation is performed after the correction, there may be a case where it is better to determine the defective pixel.

Therefore, in the sixth embodiment, predetermined correction data information is reflected on the noise level value detected when the reference heat source 60 is imaged. This method is applicable to the configurations shown in FIGS. FIG. 8 shows a case where sensitivity correction for each element is added to the reference data. In the figure, 28 is a multiplier (x).

The correction data XG is, for example, gain data for correcting the sensitivity of each element to a required value. Although not shown, the gain data XG is generated in advance by the sensitivity correction unit 9 performing the sensitivity stabilization process when the reference heat source 60 for sensitivity correction is imaged. When extracting defective elements,
Each read data SD (i) of the reference heat source 60 is multiplied by gain data XG (i) for each element, and the noise level δ (i), P (i), DC (i) for each element is determined based on the output data.
Ask for each. Since the information of the sensitivity correction data is reflected in each of these detection noise levels, it is possible to determine a defective pixel from the viewpoint of a display image. That is, although the RMS noise of a certain element is not so large as the element itself, its sensitivity is somewhat small, so that the sensitivity is corrected and amplified to the required sensitivity. As a result, the RMS noise component of the element is also amplified, which may exceed the threshold RMSTH. Such an element may be regarded as a defective element from the viewpoint of RMS noise even if the second sensitivity determination described later is passed.

The second sensitivity determination section 22 can determine the sensitivity based on the pixel data SD after the sensitivity correction. Since the sensitivity correction by the sensitivity correction unit 9 is naturally limited, the size of the gain data XG also has an upper limit. Therefore, there is an element that can have only the required sensitivity or less even when evaluated with the pixel data SD after the sensitivity correction, and the second sensitivity determination unit 22 determines such an element as a sensitivity defective element.

For the first sensitivity judging unit 21, since it is not meaningful to use the pixel data SD after the sensitivity correction, the pixel data SD before the sensitivity correction is used. Of course, the pixel data SD after the sensitivity correction may be used,
Further, the second sensitivity determination unit 22 determines whether the pixel data S before the sensitivity correction is correct.
D may be used. FIG. 9 shows a case where sensitivity correction for each element is added to the detection noise level. In the figure, 2
8 to 30 are multipliers (×).

When calculating the average value of the RMS, PP and DC noises, the sensitivity correction data XG = 1 is fixed. As a result, an average value specific to the array detection element 3 is obtained in the same manner as in the case of FIG. 7, and threshold values RMSTH, PTH, and DCTH specific to the array detection element 3 can be generated. On the other hand, when a defective element is extracted, the detection noise level δ (i), P (i), DC (i) of each of the calculation units 12 to 14 is multiplied by the sensitivity correction data XG (i) for each element. Each of the results is compared with the corresponding threshold value RMSTH, PTH, DCTH. Since information of the sensitivity correction data is reflected in each of these detection noise levels, a well-balanced defective pixel determination can be performed based on a threshold value unique to the array detection element 3 and from the viewpoint of a display image.

The sensitivity correction data XG at the time of calculating the average value of the RMS, PP and DC noises may be used as the gain data XG (i) for each element. in this case,
The information of the sensitivity correction data is also reflected on the threshold value unique to the array detection element 3. FIGS. 10 and 11 are diagrams (1) and (2) for explaining an infrared imaging apparatus according to the seventh embodiment.
It relates to improvement of C noise level detection.

In short, the DC noise is slightly different due to a change in the operating environment (temperature, bias voltage, etc.) of the element when the same reference temperature T is read by the same element at a certain time interval. It refers to a phenomenon in which the temperature T ± α is detected. Scan (j) at a certain point in time of the i element
Since each type of sample data SD (j) contains various types of instantaneous noise components, the conventional DC operation unit 14 of FIG. DC level S with noise component canceled
a (i). If the current DC level (average value) of the i element is Sa (i) and the previous DC level (average value) at which the same temperature is read is Sa (i) ′,
The DC noise level DC (i) of the i-element this time = | Sa
(I) -Sa (i) '|.

However, the previous and current reference temperatures T in the reference heat source 60 are not always completely the same. Actually, the plate surface temperature has subtle unevenness (fluctuation). If the reference temperature for the i-element itself varies by ΔT, this directly appears as the DC noise level of the i-element, which is inconvenient. there were. FIG. 10 shows the configuration of the DC operation unit 14 for improving this.

The circuit composed of the RAM 1 obtains an average value Sa (i) for each element based on the image data of the reference heat source 60 at a uniform temperature T at a certain point in time, and holds these. Thereafter, these are transferred to the RAM 2 and the previous DC level S
a (i) ′. Further, the circuit composed of the RAM 1 obtains the average value Sa (i) for each element based on the imaging data of the reference heat source 60 at a uniform temperature T (not necessarily the same as the previous time) at another subsequent time and holds these values. I do. These are the current DC level Sa (i).

Before extracting the defective element, the RAM 2
The previous average value Saa ′ of all the elements is obtained based on the previous DC level Sa (i) ′ of each element, and this is stored in the register RE.
Store it in G2. Further, based on the DC level Sa (i) of each element of the RAM 1 at this time, an average value Saa of all the elements at this time is obtained, and this is stored in the register REG1. At the time of defective element extraction, the DC noise level DC (i) of each element at this time is calculated as follows: DC (i) = | Sa (i) −Saa− ｛Sa (i) ′ −
Saa ′｝ | Since the sign has no meaning, an absolute value is used.

By transforming the above equation, DC (i) = | Sa (i) −Sa (i) ′ − ｛Saa−
Saa ′｝ |. The first half term ｛Sa (i) -Sa (i) on the right side of the above equation
'｝ Is the same as the DC operation of the conventional equation (3). Therefore, if the reference temperature T for the i-element has a variation of ΔT, this directly appears as the DC noise level of the i-element.

However, in the present embodiment, the latter term {Saa-Saa ′} on the right side of the above equation exists. Here, the current average value of all elements Saa is, so to speak, the average value of all current plate surface temperatures. Even if the detection sensitivity of each element fluctuates, these fluctuation components are offset as a result of the entire element. It can be said that this accurately represents the reference heat source set temperature. Similarly, the previous total element average value Saa ′ accurately represents the previous reference heat source set temperature. Therefore, if there is a difference between the previous reference heat source temperature and the present reference heat source temperature, a temperature component ΔT ′ corresponding to the difference also appears from the term {Saa−Saa ′}. This difference ΔT ′ is very close to the difference ΔT for the i element. Therefore, the component ΔT corresponding to the temperature difference of the reference heat source 60 is removed from the DC noise level DC (i) for the i element, and a more accurate DC noise level can be detected.

FIG. 11 is a timing chart of the DC operation unit 14 of FIG. FIG. 11A shows the previous DC level detection processing. The set temperature of the reference heat source 60 in this example is low, and 20.1 ° C. is detected as the average value Saa ′ of all elements. Focusing on a certain element i = 8, since the sensitivity of the element at this time happened to be low, Sa (8) ′ −
-△ T (8) 'is obtained as Saa'. The -ΔT
(8) 'is a net variation from the reference heat source set temperature Saa', and is no longer related to the reference heat source set temperature Saa 'at this time.

FIG. 11B shows the current DC level detection process. The set temperature of the reference heat source 60 in this example is high,
24.9 ° C. is detected as the average value Saa of all the elements. Since the sensitivity of the element i = 8 at this time happened to be high, the Sa (8) -Saa at this time was + ΔT (8).
Is required. The ΔT (8) is a reference heat source set temperature Saa.
, And is no longer related to the reference heat source set temperature Saa at this time.

FIG. 11C shows a threshold value judgment process at the time of defective pixel extraction. For element i = 8, see FIG.
(A), (B) the net fluctuations ΔT (8) ′, ΔT
(8) and relatively large DC noise level DC
(8). This is because the detection sensitivity of element i = 8 is-
It is faithful that the fluctuation is relatively large from ΔT (8) ′ to + ΔT (8). Therefore, element i = 8 is DC
(8) It is determined as a defective element by> DCTH.

Thus, according to the seventh embodiment,
An accurate DC noise level can always be detected regardless of the set temperature (or temperature error) of the reference heat source 60. FIG. 12 is a diagram for explaining an infrared imaging device according to the eighth embodiment.
It relates to improvement of P noise level detection. In the figure, 13
A is a PP calculation unit according to the eighth embodiment, and 13a extracts p (three in the figure) pixel data in the order from the largest pixel data S (j) of each i element. The maximum value group detection unit 13b is a minimum value group detection unit that extracts q (three in the figure) pixel data in order from the minimum value.

When the image of the reference heat source 60 at the uniform level is picked up, the P-
The case of performing P noise level detection calculation will be described. In the maximum value group detector 13a, the CMP1 compares the input pixel data SD with the storage data R1 (the minimum value is 0 at the beginning) of the RAM1, and outputs a write enable signal WE1 = 1 of the RAM1 when SD> R1. Then, the input pixel data SD is overwritten on the write address of the i element of the RAM 1 (= the storage address of the storage data R1). If SD ≦ R1, WE1 = 0 is output, and data writing to RAM1 is not performed. That is, the circuit relating to the RAM 1 is a maximum value detection circuit that unconditionally detects the largest one from the input j pixel data series.

The circuit relating to the RAM 2 is also basically a maximum value detecting circuit. However, a selector SEL1 is provided on the data line of the RAM2, and data written in the RAM2 has certain conditions. That is, CMP2 is S
When D> R2 is detected, CMP1 simultaneously detects SD> R1
Is detected, the stored data R1 of the RAM 1 at that time (that is, the pixel data R1 which was the largest in the pixel data series up to that time) is written to the RAM 2, and the CMP
1 detects SD ≦ R1, the input pixel data SD at that time (that is, pixel data which does not exceed the maximum pixel data R1 but is larger than the storage data R2 of the RAM 2) is stored in the RAM 2 Write to. The same applies to the circuit relating to the RAM 3.

Therefore, the maximum value group detector 13a sets j
When the pixel data series are input, the RAM1 to R
AM3 stores the third (generally the p-th) pixel data downward from the maximum. Minimum value group detector 13
b, the CMP4 includes the input pixel data SD and the RAM
4 with the stored data R4 (initially the maximum value FF),
If SD <R4, the write enable signal WE of the RAM4
4 = 1 is output, and the input pixel data SD is written to the i-element write address of the RAM 4 (= the storage address of the storage data R4).
Overwrite If SD ≧ R4, WE4 = 0 is output, and no data is written to the RAM4. That is, the circuit relating to the RAM 4 is a minimum value detection circuit that unconditionally detects the smallest one from the input j pixel data series. In the same manner as described above, the circuit composed of RAM5 and RAM6 is a conditional minimum value detection circuit. Therefore, when j pixel data sequences are input to the minimum value group detection unit 13b, the RAM4 to RAM6 store the third (generally q) pixel data in order from the smallest.

The selectors SEL3 and SEL6 select pixel data R1 to R3 of the maximum value group in accordance with the external selection signal SW.
And any combination of pixel data R4 to R6 of the minimum value group. For example, the pixel data R2 and R5 (or R3 and R6) are selected. This is because pixel data R1 that is likely to contain peak noise and pixel data R that is likely to contain bottom noise
4 is a selection to exclude from the target of the PP noise level evaluation.

When a defective element is extracted, a PP noise level P (i) = | R2 (i) -R5 (i) | is obtained for each element, and is used for defective element determination. Note that the previously obtained PP noise level P (i) itself may be stored in the RAM. FIG. 13 is a diagram for explaining an infrared imaging device according to the ninth embodiment, and relates to another improvement of PP noise level detection.

In the figure, 13B is a PP calculation unit according to the ninth embodiment, 13c is a maximum value group detection unit,
13d is a minimum value group detection unit. Here, the pixel data R1 to R3 of the maximum value group and the pixel data R4 to R6 of the minimum value group are respectively averaged by the averaging unit. Therefore, an instantaneous noise component is removed (or reduced) from the averaged maximum value and minimum value, and an accurate P
-P noise detection level P (i) can be provided.

FIGS. 14 and 15 are views (1) and (2) for explaining an infrared imaging apparatus according to the tenth embodiment, in which pixel data of a defective element is replaced with an average value of pixel data above and below the defective element. Is shown. In FIG. 14A, 31
Is a defective pixel interpolation unit. When the target 50 is imaged, the image data SD (i) of each column (element arrangement direction) is sequentially written to the image memory 6 according to the write address WA of the write counter 7. Further, the stored data MD (i) of the image memory 6 is sequentially read in accordance with the read address RA of the read counter 8 which is delayed by a predetermined phase from the write counter 7, and is input to the terminal 0 of the selector SEL. The register REG latches each read data MD (i) by the pixel clock GCK and delays it by one pixel. The adder adds the read data MD (i) of the memory 6 and the delayed pixel data MD (i) ′, and halves (shifts down by one bit) the obtained addition output, and outputs the result to the terminal 1 of the selector SEL. To the side.

On the other hand, the comparators CMP15 to CMP17 evaluate the noise levels δ (i), P (i) and DC (i) of the i element in synchronization with the read address RA of the read counter 8, and determine whether or not the element is defective. Is determined. If the i element is not determined to be a defective element, the replacement enable signal SE = 0
(That is, the addition value α = 0 to the read address RA),
Read address RA is not changed. Selector SEL
Selects the read data MD (i) on the terminal 0 side according to SE = 0.

If the i element is determined to be defective with respect to any of the noise levels, the replacement enable signal SE = 1 (that is, the added value α to the read address RA = the value of the next pixel data MD (i + 1)). Offset address), and the read address RA is changed to the read address of the next pixel data. The selector SEL selects the read data MD (i) ′ on the terminal 1 side according to SE = 1.

FIG. 14B shows an operation timing chart of the defective pixel interpolation section. For example, when the element i = 4 is determined to be a defective element, at this timing, the replacement enable signal SE = 1 causes the content G5 of the storage data MD (5) to be read instead of the content G4 of the storage data MD (4). It is. On the other hand, REG holds the content G3 of the storage data MD (3) one pixel before, and averages these values (G3 + G
5) / 2 are formed. Then, the selector SEL has SE =
1, the terminal 1 side is selected, and pixel data (G3 + G5) / 2 is output as replacement pixel data for the defective element i = 4.

FIG. 15 is a view for explaining a display image as an example. FIG. 15A shows a case of defective pixel replacement similar to the conventional case. A pyramid-shaped target having a uniform surface temperature is displayed on the monitor 80. For example, a low-temperature background is black,
The hot targets are each displayed in white. Where i
= 8, 14 are determined to be defective, the defective pixel data RD (8) for one row of display is replaced with the lower one, and the defective pixel data RD (14) for one row of display is replaced with the upper one. For this reason, the smoothness of the inclination of the shoulder of the pyramid is lost from the viewpoint of the shape.

FIG. 15B shows the case of defective pixel replacement (interpolation) according to the tenth embodiment. Here, the defective pixel data RD (8) for one line of display is replaced (interpolated) with the average value of the upper and lower pixel data RD (7) and RD (9), and the defective pixel data RD ( 1
4) is replaced (interpolated) by the average value of the upper and lower pixel data RD (13) and RD (15). Therefore, the pixel of the replacement part is displayed without a sense of incongruity, and the inclination of the shoulder of the pyramid looks smooth.

FIG. 16 is a view for explaining an infrared imaging apparatus according to the eleventh embodiment. Image data of a portion equivalent to imaging a reference heat source from image data of a target (read data of a reference image) is shown. Is extracted and stored in a memory, and these are used at the time of defective element determination. In the figure, 93A is a reference data extraction unit, 32 is a uniform data determination unit, and 33 is one column of pixel data S in the element array direction.
RAM for storing (delaying) D, 34A a write control unit for reference pixel data (reference data), 35 a RAM for storing reference data, 36 a read control unit for reference data, 37A a reference data Is an extraction control unit.

At the time of imaging the target object 50, the extraction control section 37A outputs a reset signal R1 in synchronization with the start of reading of each column, and resets the flip-flop FF1. Thereafter, the imaging data SD (i) for one column is input, and these are temporarily stored in the RAM 33. On the other hand, the image data SD (i) is also input to the uniform data determination unit 32 at the same time, and it is checked whether the amplitude of one column is within a predetermined range.

Specifically, the circuit composed of the register REG1 detects and retains the maximum pixel data R1 in the image data sequence, and the circuit composed of REG2 detects and retains the minimum pixel data R2. Then, a PP noise level P = R1-R2 is obtained by the adder (+), and the comparator CMP3 determines the noise level P and a predetermined threshold LTH.
FF1 is set when P> LTH. If P ≦ LTH, FF1 is not set.

The extraction control unit 37A examines the contents of the FF1 when the image data for one column is input, and detects the over-detection signal O.
When V = 1, since a uniform data string is not detected in the element array direction, the process directly proceeds to the uniform data string determination processing for the next row. When the over-detection signal OV = 0, a uniform data string is detected, so that the write control unit 34A is energized, and the process proceeds to the next row uniform data string determination process. During this interval, the write control unit 34A stores the (delayed) data in the RAM 33 in the RAM 3
Store in 5

In this way, a reference pixel data string is extracted from one scan of a certain target, for example, with j = 3, 4, 8 columns. Since the target includes a wall and a background having a uniform temperature, there are many opportunities to acquire reference data. In the inset (a), a reference pixel data column is extracted from each column of j = 3, 4 and stored in the RAM 35. These happen to belong to the same average level (imaging temperature T1), and are preferably stored in the storage area of the same temperature group in the RAM 35. The writing control unit 34A can perform temperature-dependent writing control by obtaining the average of the pixel data sequence. On the other hand, the pixel data sequence of j = 8 belongs to a different average level (imaging temperature T2) and is stored in the storage area of another temperature group in the RAM 35. However, j = 10
Pixel data strings such as 0 and 200 are not extracted.

When the extraction and storage of the uniform data sequence are continued and a plurality of pixel data sequences for the same temperature T1 are extracted for one or more scans, the reference heat source 60 is imaged based on these. Similarly, the RMS, PP, and DC noise levels of each element can be obtained effectively. In an apparatus including the reference data extracting unit 93A, when the apparatus is powered on, the reference heat source 60 is imaged, and the apparatus is initialized (calibrated). Thereafter, the reference image data is appropriately extracted from the image data of the target object 50, and is stored and stored in the RAM 35. Then, periodically or when the display quality is degraded, the RAM 3
Using the reference data of No. 5, each noise level of RMS, PP, DC, etc. is recalculated. The read control unit 36 is the RAM 3
5, a storage data string necessary for noise level detection is read and provided to the noise level calculation units 12 to 14 and the like.

Now, focusing on a certain element i = 3, each noise level is satisfied in the judgment at the time of power-on, but when the target 50 is imaged thereafter, for example, the RMS noise level may increase. . The pixel data in such a state regarding the element i = 3 is already accumulated and stored in the RAM 35, and when a recalculation of each noise level is performed at a certain point, the element i = 3 is regarded as a defective element from the viewpoint of RMS noise. It is determined, and pixel replacement is performed thereafter. On the other hand, there may be an element that was initially determined to be a defective element, but is not determined to be a defective element this time.

Thus, the imaging characteristic of the apparatus is periodically or appropriately re-calibrated without the need to image the reference heat source 60 during the imaging of the target object 50, and a high quality monitor image is always obtained on the monitor 80. FIG. 17 and FIG. 18 are diagrams (1) for explaining an infrared imaging device according to the twelfth embodiment,
(2) shows a case where read data of a reference image is extracted at various angles from image data of a target.

In FIG. 17, reference numeral 38 denotes a frame memory (FM) capable of storing one screen of image pickup data SD, 39 denotes a write counter (WC), and 40 denotes a read pattern generating unit (40) for generating a read pattern of pixel data. RPG), 41 is a read address generation unit for generating a read address according to the read pattern, and 42 is an averaging unit for calculating an average value of the read data string.

The image data SD for one screen of the target 50 is stored in the frame memory 38. The extraction control unit 37B outputs a reset signal R1 in synchronization with the start of reading of the stored data in each column (although there are diagonal columns) of the frame memory 38, and resets FF1. Thereafter, the imaging data SD (i) for one column is read from the frame memory 38.

FIG. 18 shows some read patterns of the frame memory 38. FIG. 18A shows a case where stored data is read in the element array direction, and this read pattern is used. The data reading by the pattern is performed for all the stored data of the frame memory 38. When a data string having a uniform amplitude is detected, the data series is R
Stored in AM35.

FIG. 18B shows a case where stored data is read out at an inclination of an angle θ ₁ from the element array direction, and this read pattern is used. The read pattern is also used for all the stored data in the range that can be read. When a uniform data string is detected, the data series is stored in the RAM 35. As the angle θ increases, the number of rows that can be read also decreases.

Thereafter, the process proceeds in the same manner, and FIG.
From the element array direction to the angle θ ₇Read with the inclination of
This case shows this read pattern. like this
Then, in the illustrated example, θ = 0 to 45 ° per one image capturing screen,
Until θ matches the angle of the diagonal line of the imaging screen,
Search and extract useful data. θ exceeds this angle
Then, the extracted data on the element i = 1 or m side is missing.
Here, this range is not extracted.

Returning to FIG. 17, such a designation of the read pattern is performed by the extraction control unit 37B, and the read pattern generating unit 40 generates a read pattern signal RPT corresponding to the pattern code signal PC. The read pattern generation unit 40
Although not shown, an i counter for counting the row number of the frame memory 38 and a j counter for counting the column number are provided.
For example, in the case of a read pattern, j = 1 is fixed and i = 1
Count up to m. Next, j is fixed at 2, and i = 1 to m are counted up. This is performed until j = m.

In the case of a read pattern, i = 1
J starts from 1 while counting up to ~ m,
In addition, it counts up at a substantially equal pace to two or more predetermined values. Such a circuit is very similar to a circuit for generating a linear vector at an arbitrary angle in a bit map type graphic display. In the next round, while counting up to i = 1 to m, j starts from 2 and counts up to a predetermined value of 3 or more at a substantially uniform pace. This is performed within the possible range of j. Hereinafter, the same applies. Then, the read address generator 41 is connected to the read pattern generator 40.
Is converted into a read address RA of each corresponding pixel data in the frame memory 38.

Note that the count values i, j (or the read address of the FM 38) of each read pattern corresponding to the designated code PC may be generated in a ROM in which conversion data is stored in advance. The read data RD of the frame memory 38 is inspected by the uniform data determination unit 32 to determine whether the amplitude difference P for one column is within a predetermined range. When P> LTH, FF1 is set, whereby the extraction control unit 37B immediately stops processing of the column, outputs a reset signal R1, and starts reading control of the next column. The determination of a meaningless column is interrupted, and the process is accelerated. When OV = 0 when data reading for one column is completed, the data writing control of the writing control unit 34B is activated.

On the other hand, the RAM 33 stores (delays) the read data of the FM 38 for one column. When the angle θ of the data readout column increases, the number of data read out from the FM 38 may be greater than m, so the RAM 33 has a storage capacity of k (> m) pieces. The averaging unit 42 calculates an average value of k pieces of read data. Since the average value is the average value of all the elements, it indicates the accurate temperature of the reference image even if the sensitivity of a certain element is deteriorated. The writing control unit 34B stores the data series of the RAM 33 in the writing area of the corresponding temperature group based on the average value output of the averaging unit 42. When a plurality of data are read from the same i element, the write control unit 34B
Stores these average values in the RAM 35. Thus,
One pixel data is extracted for each element in any readout pattern. In this way, the reference data stored in the RAM 35 is used when extracting a defective element.

The above-mentioned extraction pattern is an example. If θ = 90 °, for example, uniform j reference data can be extracted for the i element. Although the above embodiments have been described as applied to an infrared imaging apparatus, the present invention can also be applied to this type of imaging apparatus using ordinary light, ultrasonic waves, X-rays, and the like. In each of the above embodiments, the case where the array detection elements 3 are one-dimensionally arranged has been described.

Further, in each of the above-mentioned embodiments, the description has been made mainly on the solution means by the hardware configuration.
It may be realized mainly by software processing using a C, DSP, general-purpose CPU, or the like. In addition, although a plurality of embodiments suitable for the present invention have been described, it goes without saying that various changes in the configuration, control, and combinations thereof can be made without departing from the spirit of the present invention. .

[0128]

As described above, according to the present invention, the sensitivity of the displayed image is improved, the optimum threshold value for defective pixel replacement is easily set, and the frequency of use of the reference image (heat) source is preferably reduced. It can be greatly reduced. Therefore, there are great contributions to the performance improvement, miniaturization and cost reduction of this type of apparatus.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating a schematic configuration of an infrared imaging device according to an embodiment.

FIG. 2 is a diagram illustrating an infrared imaging device according to the first embodiment.

FIG. 3 is a diagram (1) illustrating an infrared imaging device according to a second embodiment;

FIG. 4 is a diagram (2) illustrating an infrared imaging device according to a second embodiment.

FIG. 5 is a diagram illustrating an infrared imaging device according to a third embodiment.

FIG. 6 is a diagram illustrating an infrared imaging device according to a fourth embodiment.

FIG. 7 is a diagram illustrating an infrared imaging device according to a fifth embodiment.

FIG. 8 is a diagram (1) illustrating an infrared imaging device according to a sixth embodiment.

FIG. 9 is a diagram (2) illustrating an infrared imaging device according to a sixth embodiment.

FIG. 10 is a diagram (1) illustrating an infrared imaging device according to a seventh embodiment.

FIG. 11 is a diagram (2) illustrating an infrared imaging device according to a seventh embodiment.

FIG. 12 is a diagram illustrating an infrared imaging device according to an eighth embodiment.

FIG. 13 is a diagram illustrating an infrared imaging device according to a ninth embodiment.

FIG. 14 is a diagram (1) illustrating an infrared imaging device according to a tenth embodiment.

FIG. 15 is a diagram (2) illustrating the infrared imaging device according to the tenth embodiment.

FIG. 16 is a diagram illustrating an infrared imaging device according to an eleventh embodiment.

FIG. 17 is a diagram (1) illustrating an infrared imaging device according to a twelfth embodiment.

FIG. 18 is a diagram (2) illustrating an infrared imaging device according to a twelfth embodiment.

FIG. 19 is a diagram (1) illustrating a conventional technique.

FIG. 20 is a diagram (2) explaining a conventional technique.

FIG. 21 is a diagram (3) illustrating a conventional technique.

FIG. 22 is a diagram (4) explaining a conventional technique.

FIG. 23 is a diagram (5) illustrating a conventional technique.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 polygon mirror 2 imaging lens 3 array detection element 4 preamplifier (PA) 5 A / D conversion unit 6 image memory 7 write counter (WC) 8 read counter (RC) 9 sensitivity correction unit 10 display level / gain adjustment unit 11 D / A conversion unit 12 RMS operation unit 13 PP operation unit 14 DC operation unit 15 to 17 Comparator (CMP) 18 OR gate circuit 19 RAM 20 Adder (+) 21 First sensitivity determination unit 22 Second sensitivity Judgment units 23, 24 Threshold control unit 24A Maximum noise type detection unit 25-27 Average value calculation unit 28-30 Multiplier (×) 31 Defective pixel interpolation unit 32 Uniform data judgment unit 33 Delay RAM 34 Write control unit 35 Storage RAM 36 read control unit 37 extraction control unit 38 frame memory (FM) 39 write counter (WC) 40 read pattern generation unit ( PG) 41 read address generation unit 42 averaging unit 50 target (target space) 60 reference heat source 70 infrared imaging device 71 main control unit 72 replacement control unit 80 monitor 90 infrared imaging device 91 main control unit 92 defective pixel extraction / replacement control Unit 93 reference data extraction unit DEC decoder RAM random access memory CMP comparator CTR counter REG register SEL data selector TG timing generation unit + adder × multiplier

 ────────────────────────────────────────────────── ─── Continuing on the front page (72) Inventor Yuichi Matsuda 4-1-1 Kamikadanaka, Nakahara-ku, Kawasaki City, Kanagawa Prefecture Inside Fujitsu Limited

Claims (16)

[Claims] An imaging apparatus for extracting a defective element based on read data of an image serving as a reference by an array-like reading element and replacing pixel data of the defective element with pixel data of another element when capturing a target object. The image serving as a reference for the uniform level is read at a plurality of points in the vertical direction of the array-like reading element, and an element having an extremely small difference between the maximum value and the minimum value of the read data obtained for each element is determined as a defective element. An imaging device characterized by the above-mentioned. 2. An imaging apparatus for extracting a defective element based on read data of an image serving as a reference by an array-shaped reading element and replacing pixel data of the defective element with pixel data of another element at the time of capturing a target object. Reading a plurality of reference images at two different levels in the vertical direction of the array-like reading element, and averaging the reading data of one level for each element and the averaging of reading data of the other level for each element; An imaging apparatus, wherein an element whose difference from a value is sufficiently smaller than a value corresponding to the difference between the different levels is determined as a defective element. 3. An imaging apparatus for extracting a defective element based on read data of an image serving as a reference by an array-shaped reading element, and replacing pixel data of the defective element with pixel data of another element when capturing a target object. A noise level detection unit for detecting one or more noise levels for each element based on read data of an image serving as a reference; and a plurality of different sizes provided for the one or more noise levels, respectively. Selecting a threshold value and a strictest threshold value for each of the one or more noise levels, and counting the number of defective elements in which any of the detected noise levels exceeds a corresponding threshold value for all elements of the array-type reading element. And a threshold control unit that switches the selection of each of the threshold values to a looser one until the obtained number of defective elements is equal to or less than a predetermined number. Imaging device. 4. An imaging apparatus for extracting a defective element based on read data of an image serving as a reference by an array-like reading element and replacing pixel data of the defective element with pixel data of another element when capturing an image of a target. A noise level detection unit for detecting one or more noise levels for each element based on read data of an image serving as a reference; and a plurality of different sizes provided for the one or more noise levels, respectively. A threshold value, and the least strict threshold value is selected for each of the one or more noise levels, and the number of defective elements in which any of the detected noise levels exceeds the corresponding threshold value is counted for all elements of the array read element. And a threshold control unit that switches selection of each of the thresholds to a stricter one until the obtained number of defective elements is equal to or more than a predetermined number. Imaging device. 5. An imaging apparatus for extracting a defective element based on read data of an image serving as a reference by an array-like reading element, and replacing pixel data of the defective element with pixel data of another element at the time of capturing a target object. A noise level detection unit that detects a plurality of types of noise levels for each element based on read data of a reference image; a plurality of thresholds having different sizes provided respectively for the plurality of types of noise levels; The number of defective elements whose noise level exceeds each corresponding threshold is counted individually for all the elements of the array-like reading element, and a noise level type corresponding to the maximum count value is extracted by comparing the obtained count values. A maximum noise type detection unit, selecting a strictest threshold value for each of the plurality of types of noise levels, and selecting one of the detected noises The number of defective elements whose level exceeds the corresponding threshold value is counted for all elements of the array-shaped reading element, and the noise detected by the maximum noise type detection unit is detected each time until the number of obtained defective elements becomes equal to or less than a predetermined number. An image pickup apparatus, comprising: a threshold control unit that switches selection of a threshold value of a level type to a loose one. 6. An imaging apparatus for extracting a defective element based on read data of an image serving as a reference by an array-shaped reading element and replacing pixel data of the defective element with pixel data of another element when capturing a target object. A noise level detection unit that detects a plurality of types of noise levels for each element based on read data of a reference image; a plurality of thresholds having different sizes provided respectively for the plurality of types of noise levels; The number of defective elements whose noise level exceeds each corresponding threshold value is counted individually for all the elements of the array-like reading element, and the noise level type corresponding to the minimum count value is extracted by comparing the obtained count values. A minimum noise type detection unit, for each of the plurality of types of noise levels, selecting a loosest threshold value, and selecting any of the detected noises The number of defective elements whose bell exceeds the corresponding threshold value is counted for all the elements of the array-shaped reading element, and the noise detected by the minimum noise type detection unit is updated each time until the obtained number of defective elements becomes equal to or more than a predetermined number. An image pickup apparatus comprising: a threshold control unit that switches selection of a threshold of a level type to a strict one. 7. An imaging apparatus for extracting a defective element based on read data of an image serving as a reference by an array-shaped reading element and replacing pixel data of the defective element with pixel data of another element at the time of imaging a target object. A noise level detecting unit for detecting a predetermined noise level for each element based on read data of a reference image; multiplying an average value of all the elements obtained for the detected noise level of each element by a predetermined coefficient; An imaging apparatus comprising: a threshold generation unit configured to generate a threshold for determining whether each element has a defect by comparing the detected noise level of each element with the detected noise level. 8. Instead of providing a plurality of thresholds having different magnitudes for one or more noise levels, a plurality of coefficients having different magnitudes respectively provided for one or more noise levels, A threshold generation unit that calculates an average value of all the elements for each of the noise levels of each element detected for one or more noise levels, and multiplies the obtained average value by a selected coefficient to generate a threshold. The imaging apparatus according to claim 3, wherein the threshold control unit selects a coefficient instead of selecting a threshold. 9. An imaging apparatus for extracting a defective element based on read data of an image serving as a reference by an array-shaped reading element, and replacing pixel data of the defective element with pixel data of another element at the time of capturing a target object. An image pickup apparatus, wherein each read data of a reference image is corrected with predetermined correction data for each element obtained separately. 10. An imaging apparatus for extracting a defective element based on read data of an image serving as a reference by an array-like reading element, and replacing pixel data of the defective element with pixel data of another element at the time of capturing a target object. An image pickup apparatus, wherein a signal level for detecting a defective element detected for each element is corrected by separately obtained predetermined correction data for each element. 11. An imaging apparatus for extracting a defective element based on read data of an image serving as a reference by an array-shaped reading element and replacing pixel data of the defective element with pixel data of another element when capturing a target object. , An image serving as a reference for the uniform level is read at a plurality of points in the vertical direction of the array-like reading element, and the average value of all elements is subtracted from the average value of each element of the previous reading data and the element of the current reading data. An image pickup apparatus characterized in that an element whose difference from a value obtained by subtracting the average value of all elements from each average value exceeds a predetermined threshold value is determined as a defective element. 12. An imaging apparatus for extracting a defective element based on read data of an image serving as a reference by an array-shaped reading element and replacing pixel data of the defective element with pixel data of another element when capturing a target object. A plurality of images are read in the vertical direction of the array-like reading element, and the read data of the maximum group from the maximum value to the p-th and the read data of the minimum group from the minimum value to the q-th are extracted for each element. In addition, an imaging device is characterized in that an element whose difference between predetermined read data in the maximum group and predetermined read data in the minimum group exceeds a predetermined threshold value is determined as a defective element. 13. An imaging apparatus for extracting a defective element based on read data of an image serving as a reference by an array-shaped reading element, and replacing pixel data of the defective element with pixel data of another element when capturing a target object. A plurality of images are read in the vertical direction of the array-like reading element, and the read data of the maximum group from the maximum value to the p-th and the read data of the minimum group from the minimum value to the q-th are extracted for each element. In addition, an imaging device is characterized in that an element whose difference between the average value of each read data of the maximum group and the average value of each read data of the minimum group exceeds a predetermined threshold value is determined as a defective element. 14. An imaging apparatus for extracting a defective element based on read data of an image serving as a reference by an array-shaped reading element and replacing pixel data of the defective element with pixel data of another element at the time of capturing a target object. An image pickup apparatus, wherein pixel data of a defective element at the time of imaging a target object is replaced by an average value of pixel data of elements on both sides of the defective element. 15. An imaging apparatus for extracting a defective element based on read data of an image serving as a reference by an array-type reading element and replacing pixel data of the defective element with pixel data of another element at the time of capturing a target object. Detecting a continuous uniform level read data string in the element array direction of the array-shaped read element and / or a direction forming a predetermined angle with the array at the time of imaging of the target object, these are read into the memory as a read data string of a reference image. An imaging device, wherein the image is stored and used for extracting defective elements. 16. The imaging apparatus according to claim 15, wherein read data strings of the same and different levels extracted from the same and different imaging frames are stored in a memory.