JP2001325595A - Method and device for image processing and charged particle microscopical inspecting device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method and a device for image processing and a charged particle microscopical inspecting device which can extract a real difference part (real defective part) from a difference image of two images having irregular luminance variation while making the misetraction rate as low as possible. SOLUTION: An irregular luminance variation component is extracted by using only one of two object images and the intensity and generation rate of the component are estimated to compute such a binarization threshold that the rate of misextraction is less than a specific rate.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing method and apparatus for extracting a difference between two images by calculating a difference between two image signals and binarizing the difference image signal. The present invention relates to an inspection apparatus including a review apparatus for a charged particle microscope such as an SEM provided with the above.

[0002]

2. Description of the Related Art In general, the following two methods are often used for setting a binarization determination threshold value.
(1) A method of setting a specific value by manual trial and error, (2) A method of calculating a threshold value from a histogram of an image by a discriminant analysis method (Reference A: Nobuyuki Otsu,
“Automatic Threshold Selection Method Based on Discrimination and Least Square Criterion”, IEICE Transactions, vol.J63-D, no.4, pp.349-
356 (1980)).

[0003]

However, in the method according to (1), the set value may not always be optimal depending on the skill of the setter. In addition, it is necessary to reset the determination threshold every time the imaging state of the image changes, such as when the intensity of the irregular luminance change changes. Further, the method according to (2) is based on the premise that the gray-scale value histogram of the image to be binarized has bimodality, and in an image for which bimodality is not guaranteed, the calculated threshold value is It may not always be an appropriate value.

[0004] An object of the present invention is to solve the above problems.
Provided is an image processing method and apparatus capable of extracting a true difference portion (true defect portion) from a difference image of two images including an irregular luminance change by reducing an erroneous extraction rate as much as possible. It is in. Further, another object of the present invention is to extract and determine a very minute defect from a detected image signal having an irregular luminance change detected by a charged particle microscope with an erroneous extraction rate as low as possible. To provide a charged particle microscope inspection apparatus.

[0005]

SUMMARY OF THE INVENTION In order to achieve the above object, the present invention is based on one of two image signals based on an irregular change in luminance contained in the image signal. A binary process that sets a rate of occurrence of erroneous extraction due to an irregular brightness change to a desired rate or less based on a creating process of creating an intensity distribution and an intensity distribution due to the irregular brightness change created in the creating process. A setting step of setting a threshold value for determining the conversion, a difference calculating step of calculating a difference between the two image signals, and a difference image signal calculated by the difference calculating step. An extraction processing step of extracting a difference between the two image signals using a binarization determination threshold, and an apparatus therefor. Further, the present invention provides a creation process of creating an intensity distribution due to an irregular luminance change included in this image signal based on one of the two image signals, and a creation process of the creation process. A setting for setting a binarization determination threshold value that sets a rate of occurrence of erroneous extraction due to an irregular brightness change in a predetermined image area range to be equal to or less than a desired rate based on an intensity distribution due to an irregular brightness change. A difference calculation step of calculating a difference between the two image signals, and a difference determination signal set in the difference calculation step using a binarization determination threshold value set in the setting step. An image processing method and apparatus, comprising: an extraction processing step of repeating extraction processing of a difference between the two image signals in the predetermined image area range.

Further, the present invention provides a method for manufacturing
The intensity distribution due to the irregular luminance change is shaded based on a difference image signal calculated from the one image signal and a pixel signal obtained by translating the one pixel signal in a predetermined direction by a predetermined amount. It is characterized in that it is created as a statistical value of the frequency distribution. In addition, the present invention further provides, with respect to the frequency of a certain size difference portion extracted from the extraction process, a ratio of erroneous extraction due to irregular luminance change or a ratio expected to be an actual difference portion. It has an output step of outputting.

According to the present invention, there is provided a creating process for creating an intensity distribution based on an irregular change in luminance contained in one of two image signals based on one of the two image signals. A setting step of setting a determination threshold, a difference calculation step of calculating a difference between the two image signals, and a difference image signal calculated as a difference in the difference calculation step,
An extraction processing step of extracting a difference between the two image signals using a binarization determination threshold set in the setting step, and an intensity distribution due to an irregular luminance change created in the creation step. Based on the size (for example, the area) of the different portion (one defective portion) extracted in the extraction process
And an apparatus for calculating the rate of occurrence of erroneous extraction according to the image processing method.

Further, the present invention provides a creating process for creating an intensity distribution based on an irregular luminance change included in one of two image signals based on one of the two image signals; A setting step of setting a determination threshold, a difference calculation step of calculating a difference between the two image signals, and a difference image signal calculated as a difference in the difference calculation step,
An extracting process of extracting a different portion (defective portion) between the two image signals using the binarization determination threshold value set in the setting process, and an extracting process of the different portion extracted in the extracting process. A calculation process for calculating a ratio expected to be an actual difference portion (actual defect portion) according to the size, and an image processing method and apparatus therefor.

Further, the present invention is characterized by further comprising an output step of displaying and outputting the ratio calculated in the calculation step. In the image processing apparatus, the image processing apparatus may further include a ratio input unit for inputting a ratio of occurrence of erroneous extraction due to an irregular luminance change, and a calculating unit for determining a binarization threshold according to the input ratio. It is characterized by having. Further, the present invention is a charged particle microscope inspection device such as an SEM, which is provided with the image processing device.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of a process for extracting a difference (defective portion) between two images by calculating a difference between two images and performing a binarization determination process on the difference image according to the present invention will be described with reference to the drawings. It will be described using FIG. A first embodiment of a method and an apparatus for inspecting a pattern or the like according to the present invention will be described with reference to FIG. Here, the inspection object 100 such as a wafer is scanned by the electron beam 30, electrons generated from the inspection object 100 by the irradiation of the electron beam are detected, and an electron beam image of the scanned portion is detected based on the intensity change. And inspecting a pattern or the like using the obtained electron beam image. The inspection system for this pattern or the like is configured by, for example, an SEM visual inspection system including a detection unit 101, an image extraction unit 102, an image processing unit 103, and an overall control unit 104 that controls the whole. The appearance inspection system may use a charged particle microscope other than the SEM. That is, a charged particle beam including an ion beam or the like other than the electron beam may be used. The present system includes an inspection room 105 in which the inside of the room is evacuated, and a spare room (not shown) for loading and unloading the inspection object 100 in the inspection room 105. Is configured. Inspection room 105
The inside is roughly divided into an electron optical system 106 and an electron detection unit 10.
7, a sample chamber 109, and an optical microscope unit (not shown) for positioning the inspection object 100.
The electron optical system 106 includes an electron gun 31 that emits the electron beam 30.
A condenser lens 32 for focusing the electron beam 30, a scanning deflector 34 for scanning the electron beam 30, and turning on the electron beam.
A blanking electrode (not shown) for turning on / off, an objective lens 33 for further focusing an electron beam, a reflector (not shown) for forming a conical shape and having a secondary electron multiplication effect, Ex
It comprises a B deflector (not shown) and a Faraday cup (not shown) for detecting a beam current, and is controlled by the optical system controller 110 based on a command from the overall controller 104. Among the electron detectors 107, for example, secondary electrons,
An electron detector 35 for detecting electrons such as reflected electrons is provided in the inspection room 1
05, for example, below the objective lens 33. The output signal of the electronic detector 35 is output to the inspection room 105.
, And is amplified by an amplifier 36 and the like installed outside the box. Sample room 1
Reference numeral 09 includes a sample stage (not shown), a stage 2, a length measuring device (not shown) for monitoring the position of the stage, and a substrate height detector (not shown) for inspection. The stage 2 is driven and controlled by the stage control unit 111 based on an instruction from the overall control unit 104, and is also measured by data measured by the position monitoring length measuring device or by the board height detector for inspection. The height data of the inspection board is fed back to the overall control unit 104.

Therefore, the electron beam emitted from the electron gun 31 is
After passing through the condenser lens 32 and the objective lens 33, the beam diameter on the sample surface is reduced to about the pixel size. At this time, a negative potential is applied to the sample by the ground electrode 38 and the retarding electrode 37, and the electron beam is decelerated between the objective lens 33 and the inspection object (sample) 100. Increase resolution. When the electron beam is irradiated, electrons are generated from the inspection object (wafer) 100. By detecting the electrons generated from the inspection object 100 in synchronization with the repetitive scanning of the electron beam in the X direction by the scanning deflector 34 and the continuous movement of the inspection object 100 in the Y direction by the stage 2, A two-dimensional electron beam image of the inspection object is obtained. Electrons generated from the inspected object 100 are captured by the detector 35 and amplified by the amplifier 36. Here, in order to enable high-speed inspection, an electrostatic deflector having a high deflection speed is used as the deflector 34 for repeatedly scanning the electron beam in the X direction. It is desirable to use a thermal field emission type electron gun that can shorten the irradiation time because the current can be increased, and it is desirable to use a semiconductor detector that can be driven at high speed as the detector 35.

Next, the image extracting unit 102 is processed as described below. That is, the electron detection signal detected by the electron detector 35 is amplified by the amplifier 36 and the A / D
Digital image data (gradation image data) by the converter 39
120. Then, for example, it is converted into an optical digital signal by a light emitting element and transmitted through an optical transmission path, received by a light receiving element, converted into digital image data (gradation image data), and output. The preprocessing circuit 40 performs image correction such as dark level correction, electron beam fluctuation correction, and shading correction on the digital image data. The corrected digital image data is subjected to a filtering process such as a Gaussian filter, an average value filter, or an edge enhancement filter to improve the image quality. In addition, the image distortion is corrected as needed.

Next, the digital image signal (gradation image signal) 121 whose image quality has been improved from the preprocessing circuit 40 is stored in the image memory 41 provided in the image processing unit 103. As the image memory 41, for example, a reference reference image signal D (x,
y) and the processing target image signal F (x,
The second image memory 41b for storing y) may be used. In the case of such a configuration, the digital image signal 121 output from the preprocessing circuit 40 is transmitted to the overall control
The changeover switch 51 may be switched based on a command from the user. As a result, by reading the processing target image signal F (x, y) and the reference reference image signal D (x, y) from the image memory 41, the difference image signal sub is compared.
(X, y) can be extracted. Note that chips to be inspected 100 are repeatedly arranged like a wafer substrate.

Next, the image processing unit 103 according to the present invention will be described. That is, as shown in FIG. 2A, the reference image signal D (x, y) 81 having no defect and FIG.
A processing target image signal F having a defect 91 as shown in FIG.
The (x, y) 82 is output after being aligned in the pixel unit in the pixel unit alignment unit 42. The reference reference image signal D '(x,
y) and the processing target image signal F ′ (x, y), a difference extraction circuit 49 extracts a difference image signal sub (x, y) 83 including the defect 91 shown in FIG. As described above, the difference image signal sub (x, y) 83 is a binarization target image for review / analysis according to the present invention.

By the way, the noise generated in the image pickup system is superimposed on the reference image signal D (x, y) 81 and the image signal F (x, y) 82 detected by the electronic detection unit 107, Alternatively, an irregular luminance change is included due to the presence of an irregular texture on the imaging target (inspection target) 100.

Therefore, the difference image signal sub (x, y)
A difference signal 83 (hereinafter referred to as difference noise) generated due to an irregular luminance change is provided in addition to a difference signal of a difference portion between the reference image signal D (x, y) 81 and the processing target image signal F (x, y) 82. ) Will be included.

Therefore, according to the present invention, the intensity distribution of the difference noise is estimated in the density value histogram distribution creating circuit 47, and the threshold value calculating circuit 48 calculates the binarized threshold values thH and thL from the occurrence probability. By setting the difference image signal s in the defect determination circuit 50,
The rate of occurrence of erroneous extraction due to irregular luminance change in the binarized image determined for ub (x, y) 83 is
The ratio can be less than the specified ratio, and the review / analysis image can be accurately extracted.

Next, a first embodiment of the binarization determination threshold value calculation according to the present invention will be described with reference to FIG. FIG.
FIG. 5 is a diagram showing a processing flow of a first embodiment of the binarization determination threshold value calculation according to the present invention. First, the overall control unit 10
4 includes an optical pattern inspection device, an optical foreign material inspection device, and an S
Information on the inspection object 100 inspected by an appearance inspection apparatus (not shown) such as an EM appearance inspection apparatus, and position coordinates of the inspection result, such as a pattern defect or a foreign matter defect, on the inspection object 100, Defect features as needed
(For example, the area of the defect, the projection length (length) of the defect in the X-axis and Y-axis directions, the shape of the defect, and the like) are input using the input device 201 including a network or a recording medium, and are input to the storage device 203. It is memorized. In addition, the general control unit 104 stores design information (information such as chip arrangement information, memory cell area and peripheral circuit area in the chip, and logic area in the case of a system LSI) related to the inspection object 100.
The data is input from a CAD system or the like using the input device 201 and stored in the storage device 203.

Therefore, the overall control unit 104 can know a chip having no defect on the inspection object 100. Therefore, on the inspection object 100 mounted on the sample stage fixed on the stage 2, it is basically possible to position a chip having no defect at the electron beam irradiation position. By the movement and the scanning irradiation of the electron beam, the electron detection unit 107 detects the reference image signal D (x, y) in the local image area as shown in FIG. 2A and stores it in the image memory 41a. It is possible (step S31). Further, since the overall control unit 104 can know the position coordinates of the defect to be reviewed / analyzed together with the chip coordinates on the inspection object 100, the overall control unit 104 positions the chip having the defect 91 at the electron beam irradiation position. Becomes possible, and by the movement of the stage 2 and the scanning irradiation of the electron beam,
As shown in FIG. 2B, the electronic detection unit 107 can detect the processing target image signal F (x, y) in the local image area and store it in the image memory 41b (step S31). Note that the reference reference image signal D (x, y)
May be an image signal obtained from a chip adjacent to the chip in which the defect 91 exists.

[0020] Here, the probability of irregular luminance change of the two images if the imaging condition is the same, the intensity may be used either image signal because it can be considered the same as,
If there is an unexpected luminance change area in the different part, noise of threshold calculation may occur. Therefore, it is desirable to use the image signal 81 which is a reference image signal. In the image signal 81, reference numeral 90 denotes one wiring pattern having a width of about 0.1 to 0.4 μm.

Next, for example, the image signal 81 stored in the image memory 41a is shifted by one to several pixels in a predetermined direction in a predetermined direction and read out, so that a shifted image signal can be created. (Step S
32). Note that one pixel has an electron beam focus size of about 0.005 to 0.02 μm. At this time, the shifted image signal may be stored in an image memory (not shown). As described above, the image signal shifted by one to several pixels is generated by the difference extraction circuit 46 by the difference between the reference reference image signal D (x, y) and the shifted reference reference image signal D (x + n, y + m). This is for completely erasing the wiring pattern signal 90 except for the edge portion (one to several pixels). n and m are each 0 to several pixels. That is, it may be shifted in the X direction or the Y direction. Thus, in order to completely erase the signal from the wiring pattern except for the edge portion, it is necessary to use the same image signal. The reason is that, in the case of an image signal obtained from adjacent chips, the signal 90 from the wiring pattern is slightly different and cannot be completely erased when a difference signal is obtained.

Next, the difference extraction circuit 46 determines in step S3
In 3, the difference image signal subD (x, y) between the reference image signal D (x, y) and the shifted reference image signal D (x + n, y + m) is calculated. This difference image signal subD (x, y) is obtained by
It can be considered as an image signal in which difference noise that may be generated due to irregular luminance change including an edge portion of a very fine wiring pattern is extracted.

As described above, the distribution of the irregular luminance change is calculated directly from the image signal 81, such as calculating the luminance variance of a specific area from the image signal 81 by shifting the same image signal 81 and taking the difference. Compared to the case, only the intensity of the irregular luminance change can be extracted without being affected by the luminance difference due to the wiring pattern 90 imaged as the image signal. By setting the shift amount sufficiently smaller than the spatial frequency of shading, which is a gentle luminance change, only the intensity of the irregular luminance change can be extracted without being affected by shading.

Next, in step S34, the gradation value histogram distribution creation circuit 47 creates a gradation value histogram of the difference image signal subD (x, y) calculated in step S33. FIG. 4 shows an embodiment of the gray-scale value histogram created in step S34. The horizontal axis represents the intensity of the difference noise, and the vertical axis represents the frequency of occurrence. Since the difference noise is due to an irregular luminance change, the difference noise is centered on 0,
It can be expected that the distribution follows the probability distribution. Therefore, in step S35, the threshold value calculating circuit 48 applies a normal distribution curve to the distribution of the gray value histogram created by the gray value histogram distribution creating circuit 47.

The normal distribution curve is a general theoretical formula representing a probability distribution, and is expressed by the following formula (1).

Here, σ represents a standard deviation, and μ represents an average.

[0027]

(Equation 1)

The gray value of the difference image signal in this embodiment is
Since the distribution can be assumed to be evenly distributed in the positive direction and the negative direction, a curve with the average value μ set to 0 may be applied. Further, the normal distribution curve is such that the sum of the areas included under the normal distribution curve is 1
However, for convenience of calculation, a curve obtained by multiplying the probability value by a certain constant K may be applied. Here, when the wiring pattern 90 exists in the image signal 81, the difference image signal subD
At (x, y), the edge portion of the wiring pattern in the image signal appears as a signal, and curve fitting may not be performed normally. However, if the intensity of the luminance change of the pattern is sufficiently large compared to the intensity of the irregular luminance change,
Since the signal of the pattern edge portion in the difference image signal subD (x, y) is also sufficiently larger than the signal intensity of the difference noise component, it can be expected that both are separated and distributed.

Therefore, for example, in the above-mentioned density histogram, when the search is performed in both the positive and negative directions from the center of the density histogram and in the direction away from the center, the noise intensity value at which the frequency distribution amount changes from decreasing to increasing is calculated in each of the positive and negative directions. However, by setting only the range sandwiched by the noise intensity values as the calculation target, it is possible to exclude the signal of the edge portion of the pattern in the difference image signal. here,
The signal elimination method of the edge portion of the pattern is not limited to the above-described embodiment. For example, the value 0 on the horizontal axis of the histogram may be used.
May be used only in a certain range of data centered on the pattern, and any method may be used as long as it can eliminate signals at the edge of the pattern. FIG. 5 shows the normal distribution curve applied in step S35.

At this time, the binarization determination threshold is set to a (t
h), the area under the normal distribution curve in the range of ± a can be calculated. In the normal distribution curve, since the integrated value in the range of ± ∽ is known, the ratio of the area included in the range of ± a is known. This ratio represents the probability that the intensity of the difference noise caused by the irregular luminance change is equal to or less than the threshold value a. That is, the area ratio outside the range of ± a represents the ratio r at which erroneous extraction due to a change in luminance occurs.

Therefore, in step S36, the rate r of occurrence of erroneous extraction due to the change in luminance is determined by the threshold value calculation circuit 4
8, the threshold value calculation circuit 48
Can be obtained by calculating the threshold value that satisfies it. In this manner, the threshold value calculation circuit 48 calculates the threshold values thH and thL that make the ratio r of occurrence of erroneous extraction due to irregular luminance change in the specific local image region in the chip equal to or less than the specified ratio. be able to. In the difference extraction circuit 49, the absolute value of the difference | sub (x,
y) In the case of outputting |, the threshold value may be a threshold value thH on one side. As shown in FIG. 2, the method of specifying the ratio r is specified by the number of occurrences of erroneous extraction per local image (local image region) having a side length of about several μm to about several hundred μm in the chip. May be.

The inspection is repeated by changing the inspection position,
When a certain predetermined range is defined as one unit, it may be specified by the occurrence ratio of erroneous extraction in the unit or by the number of occurrences of erroneous extraction in the unit. Further, the threshold value calculating circuit 48 does not apply the normal distribution curve to the histogram of the difference image signal subD (x, y), and in the difference noise component of the histogram, ± a Values where the area ratio of the difference noise component not included in the section (a is a certain constant) satisfies the specified ratio may be calculated as the thresholds thH and thL.

Further, the difference extraction circuit 46 determines in step S3
In 2, instead of taking the difference from the shifted image,
A difference value between the target pixel and a luminance value of a pixel at an arbitrary position within a certain distance corresponding to the shift amount may be calculated as a target pixel difference value.

The form of the input signal is not limited to the form of an image having a two-dimensional pixel array as in the examples described above, but may be a one-dimensional pixel array such as an input image signal from a line sensor. Absent. That is, a one-dimensional pixel array signal detected from the electron beam detection unit 107 per one raster for deflecting and scanning the electron beam may be used.

As described above, the threshold value calculation circuit 4
8 are threshold values thH, t based on the rate of occurrence of erroneous extraction due to a change in luminance designated using the input device 201 or the like.
Set hL. Then, the defect determination circuit 50 determines the difference signal sub (x, y) output from the difference extraction circuit 49 using the threshold values thH and thL set above, so that the noise component is erroneous at the above ratio. In the extracted state, a binary image signal indicating a defect is obtained. That is, even if the noise component is very close to the difference signal indicating the defect, it is possible to extract the binarized image signal indicating the defect 91 without detecting the noise component. By providing the information to the storage device 104, the storage device 203 can store it as map information to be truly reviewed or analyzed.
At this time, the overall control unit 104 controls the defect determination circuit 50
By receiving the processing target image signal indicating the defect stored in the image memory 41b based on the position coordinates (map information) of the defect determined in the above, the processing target image signal to be reviewed or analyzed is obtained. . As a result, the overall control unit 104 can review or analyze the defect by outputting and displaying the obtained image signal to be reviewed or analyzed, for example, on the display device 202 or the like. Become. When reviewing, for example, the defect category (type:
For example, whether the defect is a short-circuit defect, a disconnection defect, a foreign matter, or the like.

Further, the overall control unit 104 can provide these data to an analyzer (such as a mass spectrometer, a cross-sectional observation device, or an X-ray spectrometer) via, for example, a network. Further, regarding the defect determined by the defect determination circuit 50, the feature extraction circuit 50a
Defect features (defect area, defect length, defect brightness
(Shading value) can be extracted and provided to the overall control unit 104.

Next, a second embodiment of the binarization threshold value calculation according to the present invention will be described with reference to FIG. FIG.
In the frequency distribution based on the area (area extracted as one defect) of an area extracted as a different part (defect part), a difference area of a certain area (one defect is extracted with a certain area) extracted by the feature extraction circuit 50a. FIG. 9 is a diagram showing a processing flow of a method of calculating a ratio of erroneous extraction due to an irregular luminance change and determining a binarization determination threshold with respect to the frequency of (area to be performed) S. FIG. 7 shows an embodiment of a graph used to calculate the erroneous extraction occurrence ratio.

First, in step S42, the threshold value calculation circuit 48 determines the area of the different portion extraction area (the area extracted as one defect in the defect determination circuit 50).
In n, it is determined how much the occurrence ratio S2 / S1 of the erroneous extraction is made. For example, in the specified local image region, the erroneous extraction occurrence ratio S2 (n1) / for an extracted region (one extracted defect) n1 having an area of 4 pixels
S1 (n1) is 30% or less, and the erroneous extraction occurrence ratio S2 (n2) / S1 (n2) is 0.5% or less for an extraction region (one defect to be extracted) n2 having an area of 8 pixels. Set forth in Here, steps S31 to S35
The steps are the same as those shown in FIG.

In the second embodiment, the processing is a routine for estimating the erroneous extraction occurrence rate using the detected image (step S3).
2 to S35 and S37) and a routine for extracting a defective portion (steps S38 to S41).

In the routine for estimating the erroneous extraction occurrence ratio, first, in step S32, the reference image signal D (x, y) in the local image area shown in FIG. An image signal shifted in a predetermined direction by a predetermined amount is created. Next, in step S33, a difference image signal subD (x, y) between the reference image signal in step S31 and the shifted image signal in step S32 is calculated. As described above, this difference image signal can be considered as an image signal obtained by extracting difference noise that may be generated due to irregular luminance changes of the input image. Next, in step S34, a gray level histogram of the difference image signal calculated in step S33 is created. Next, in step S35, a normal distribution curve is applied to the distribution of the gray value histogram.

Next, the threshold value calculating circuit 48 calculates the ratio r of the occurrence of erroneous extraction in the local image area at the threshold value ± a (thH, thL) set in step S39 described later with respect to the defect determination circuit 50. , In the threshold value calculation circuit 50 or the overall control unit 104 by the method described above. Here, the above step S33
The threshold value calculating circuit 48 is calculated based on the gray level histogram of the difference image signal in the local image area calculated by
Threshold value ± a (thH, thL) 8 set at
5, the occurrence ratio of erroneous extraction is represented by r, and the feature extraction circuit 50a
The area of the erroneous extraction (area corresponding to one defect) extracted from n is n pixels, the number of combinations of the arrangement of the n pixels is A, and the erroneous extraction region (one difference portion (1 Assuming that the number of pixels adjacent to the periphery of the defective portion)) is m (i) pixels, the ratio R (n) in which the erroneously extracted region is composed of n pixels is expressed by the following equation (2).

[0042]

(Equation 2)

In the above equation (2), (1-r) is 1
When r is sufficiently close to 0, that is, when r is sufficiently close to 0, the ratio R (n) in which the erroneously extracted region is composed of n pixels may be expressed by the following equation (3).

[0044]

(Equation 3)

As described above, the threshold value calculation circuit 48 or the overall control unit 104 determines whether a local region having an area of n pixels (an area corresponding to one defect determined by the defect determination circuit 50 is n pixels). The erroneous extraction occurrence frequency S (n) within the image area (image calculation range) can be estimated using the following equation (4).

[0046]

(Equation 4)

That is, in step S37, the threshold value calculating circuit 48 or the overall control unit 104 sets the erroneous extraction area n detected by the feature extraction circuit 50a on the horizontal axis and the erroneous extraction occurrence frequency S (n) on the vertical axis. The calculated area / erroneous extraction occurrence frequency curve is calculated in the local image region. An example of the calculated curve 95 is shown in FIG. As is clear from the curve 95, if the area (extraction area) n of one defect to be determined as a defect in the defect determination circuit 50 increases,
The rate at which noise components are erroneously extracted as defects becomes close to 0. On the other hand, if the area n of one defect becomes smaller,
The rate at which noise components are erroneously extracted as defects will increase.

On the other hand, a routine for extracting a defective portion is as follows.
A difference signal sub (x, y) between (x, y) and the processing target image signal F (x, y) is calculated (step S38). Next, in step S39, the threshold value calculating circuit 48 determines an appropriate threshold value ± a (thH, thH, based on the information input through the overall control unit 104 using the input device 201 and the like.
thL) 38 is set.

Next, the defect judging circuit 50 proceeds to step S4
At 0, a binarization determination process is performed on the difference signal based on the set appropriate threshold value, and the difference between the image signal to be processed and the reference image (basically a portion indicating a defect)
Is extracted.

Next, in step S41, the threshold value calculation circuit 48 or the overall control unit 104 calculates the area n and frequency (for example, the number of occurrences) S of the different portion (one defective portion) extracted by the feature extraction circuit 50a. Is calculated in the local image region. The frequency S may be obtained from the defect determination circuit 50. This calculated curve 9
6 is shown in FIG. The horizontal axis indicates the area n of the different portion (one defect portion), and the vertical axis indicates the frequency (for example, the number of occurrences) of the area. As is apparent from the curve 95, if the area (extraction area) n of one defect to be determined as a defect in the defect determination circuit 50 increases, the number of occurrences in the local image area also decreases. On the other hand, if the area n of one defect becomes small, the rate of noise components being erroneously extracted as defects increases, so that the number of occurrences in the local image region also increases.

Next, in step S42, the threshold value calculation circuit 48 or the overall control unit 104 calculates the erroneous extraction occurrence ratio S2 / S1 corresponding to the extraction region area (the area determined as one defect) n. The calculated data can be stored in the storage device 203 and output and displayed on the display device 202 or the like. Since the value of the area (size) n of the defect to be determined and detected by the defect determination circuit 50 is determined in advance, the value can be designated using the input device 201 such as a keyboard. Thus, the erroneous extraction occurrence ratio S with respect to the area n of the defect to be detected
2 / S1 is the area / area calculated in step S37.
It can be easily calculated by combining the erroneous extraction occurrence frequency curve 95 and the extraction area / frequency curve 96 calculated in step S41. Further, the threshold value calculating circuit 48 or the overall control unit 104 stores the calculated area / erroneous extraction occurrence frequency curve 95 and the extracted area area / frequency curve 96 in the storage device 203 and outputs the calculated data to the display device 202 or the like. It may be displayed.

For example, as shown in FIG. 7, when the area of the extraction region (the area of one defect) is n1, the relationship between the frequency S1 (n1) of the extracted region and the number S2 (n1) of erroneous extractions is 10. : 4, the extraction region area is n1 at the threshold value set for calculating these relationships.
In this case, the erroneous extraction rate is 40%, that is, the probability of being actually determined as a different portion (defect portion) is 60%, and this can be output and displayed on the display device 202 or the like. Further, the frequency of the curve 96 is integrated with respect to the range of the extraction region (the area of one defect) from n1 to n2,
Further, the number of erroneously extracted occurrences of the curve 95 may be integrated, and the ratio of these integrated values may be taken as the erroneously extracted occurrence ratio. In this case, the probability that the area of one defect is determined as a defective portion in the range of n1 to n2 is obtained, and this can be output to the display device 202 or the like.

Next, in step S43, it is determined whether the erroneous extraction occurrence ratio S2 / S1 with respect to the extraction region area n output and displayed on the display device 202 or the like has desired characteristics. As a result, if the desired characteristic is satisfied, in step S44, the satisfaction is input to the threshold value calculation circuit 48 using an input device or the like, and the value is determined as a threshold value. If not, the flow returns to step S39 to reset the threshold value in the threshold value operation circuit 48 using an input device or the like, and the same processing is repeated until the desired characteristic is satisfied. If the table for the desired characteristics is stored in, for example, the storage device 203, the overall control unit 104 or the threshold value calculation circuit 48 determines the erroneous extraction occurrence ratio S2 / S1 with respect to the extraction region area n. Can be automatically determined in step S43. Even in this case, it is possible to confirm whether or not the desired characteristic is appropriate by outputting a table of the desired characteristic stored in the storage device 203 to the display device 202 or the like.

Using the erroneous extraction occurrence ratio calculated in this way as a guide, the two image signals D (x, y) and F by the defect determination circuit 50 are set so as to satisfy the desired erroneous extraction ratio in a desired extraction area. In the process of extracting a different portion (defective portion) of (x, y), a suitable binarization determination threshold can be determined. Here, the threshold value calculation circuit 4
8 is a case where a standard multiple of the standard deviation σ of the normal distribution curve applied in step S35 is given as an initial value of the binarization determination threshold value set in step S39, and the desired characteristic is not satisfied in step S43. To
The threshold may be changed according to a specific rule such as increasing or decreasing the binarization threshold, and automatic calculation may be performed without human intervention until desired characteristics are satisfied. In this case, it is necessary to provide the desired characteristic data to the threshold value operation circuit 48.

The specific rule for varying the above-mentioned threshold for binarization determination is not limited to the rule for monotonically increasing or decreasing the threshold, but the direction and amount of the variation are determined according to the approach to the desired characteristic. May be changed. next,
An embodiment of a method for inputting the rate r at which erroneous extraction occurs due to irregular luminance changes will be described with reference to FIG. That is, the display device 202 includes the reference image D (x, y) stored in the image memory 41a, the processing target image signal F (x, y) stored in the image memory 41b, and the difference extraction circuit 4
9, a histogram with the horizontal axis representing the luminance value of the difference image obtained from the density value histogram distribution creation circuit 47, and a binarized judgment image obtained from the defect judgment circuit 50. A signal or the like is displayed.
Then, the user operates the input device 2 on the screen of the display device 202.
By using 01, the erroneous extraction occurrence ratio (noise occurrence frequency) r can be input as, for example, 0.3%.

As a result, in step S36, the binarization determination threshold value is automatically calculated back according to the input numerical value r and the histogram, and the binarization determination threshold ± a (thH, thL) is added to the histogram. Can be displayed in a superimposed manner. The displayed binary image is
The difference image signal sub is determined by the binarization determination threshold value.
(X, y) is binarized. By the way, the input screen does not need to display all the image signals shown here, and may display any combination of arbitrary image signals including the binarized image signal.

In this manner, the user does not input the binarization determination threshold value itself, but sets the binarization determination threshold value by inputting the rate r at which erroneous extraction occurs. be able to.

Next, when the difference image signal sub (x, y) is binarized and judged at certain judgment thresholds thH and thL, the difference (defect) extracted at a certain area n
The erroneous extraction ratio S2 / S1 of the different part having
An example of the display screen displayed on the screen 2 will be described with reference to FIG. The screen includes a reference image signal D (x,
y), processing target image F (x, y), difference image sub
(X, y), a histogram 96 and an erroneous extraction occurrence number curve 95 with the area of a region (one defect) extracted by the feature extraction circuit 50a when the local image region area is about 50 nm ² , and a defect determination circuit The binarized judgment image signal and the like obtained from 50 are displayed. Then, the user sets the occurrence ratio (noise occurrence frequency) r of the erroneous extraction to, for example, 1
It can be entered as 0%. At this time, 2
The value itself of the valuation determination threshold ± a (thH, thL) may be input. Further, the number curve 95 of erroneously extracted occurrences with respect to the area n may be displayed at the same time.

By moving the dotted line vertically drawn on the histogram with a mouse or a cursor key, it is possible to specify the extraction area area n for which the erroneous extraction ratio S2 / S1 is calculated. The designated extraction area area n and the erroneous extraction ratio S2 / S1 in the area n are displayed as 50 nm ² and 35% beside the histogram screen.
By doing so, for example, in two images that are input images, the area n of the expected difference (defect)
If the lower limit of can be assumed, the ratio S2 / S1 of the erroneous extraction in the area can be quantitatively presented. A judgment threshold can be set.

The case of the SEM appearance inspection apparatus and the SEM review apparatus has been described above. However, the optical appearance inspection apparatus (optical pattern inspection apparatus, optical foreign matter inspection apparatus, etc.)
Can also be used for the defect determination. In particular, ultraviolet light (including annular illumination) is used as illumination light, and a TDI sensor capable of detecting ultraviolet light is used as a detector. As described above, when ultraviolet light is used, the defect to be detected is 0.05 to
The reason is that the probability of occurrence of erroneous extraction due to irregular luminance change increases in that case. However, it is clear that the above-described binarization determination method can be applied to an inspection apparatus using visible light.

According to the embodiment described above, the irregular luminance change component of the image is extracted, and the intensity of the difference noise due to the irregular luminance change component and the occurrence ratio are calculated. The threshold value that makes the rate of occurrence of erroneous extraction due to the threshold equal to or less than the specified rate can be calculated and set. As a result, a binarization determination threshold value suitable for removing an irregular luminance change component can be set. Can be set. Further, according to the above embodiment, at a certain threshold value, the ratio of erroneous extraction in a certain extraction region area is displayed, so that the user can evaluate the validity of the numerical value.
Further, according to the above-described embodiment, it is possible to provide a user interface for inputting the specified ratio, and to input a ratio at which erroneous extraction occurs, which is suitable for removing irregular luminance change components. It is possible to set a value threshold. Further, according to the above embodiment, at a certain threshold value, by providing a user interface for displaying the ratio of erroneous extraction in a certain extraction region area, the user can evaluate the validity of the numerical value, By modifying the input threshold value with reference to, a suitable binarization threshold value can be set.

Next, an embodiment in which the above-described inspection apparatus is applied to a line for manufacturing a pattern such as an electric circuit will be described with reference to FIG. For example, manufacturing process A,
In a manufacturing line where a pattern such as an electric circuit is formed on a semiconductor substrate via B, C,..., A reference image signal D (x, y) and a processing target image (inspection image) F in which a pattern is normally formed As an inspection apparatus for comparing (x, y) and inspecting a pattern for a defect or adhesion of a foreign substance using the present binarization determination threshold value determination method, for example, between the manufacturing process B and the manufacturing process C, By installing, it is possible to acquire information such as how many defects have occurred or how the number of defects changes over time.

When the above-described inspection device is used as a review device or an analysis device, the defect image is determined by the present binarization determination threshold value determination method based on the approximate defect position detected by the visual inspection device. For example, the extracted defect image is displayed on the display device 202 and inspected or analyzed in detail to determine what kind of defect (defect category, for example, pattern defect (disconnection or short-circuit)) Is analyzed, and is input to the overall control unit 104 using the input device 201, and
For example, it can be stored in the storage device 203 together with data (position coordinates and feature amounts on the inspection object 100) on the defect inspected by the visual inspection device.

Further, by this analysis, it is possible to determine in which manufacturing process a defect has occurred. For example, if an abnormality occurs in the manufacturing process A, it is checked in advance that an abnormality of a pattern having a similar shape occurs, or a foreign matter having a similar shape occurs, and the abnormality is detected by the input device 2.
01 and stored in the storage device 203, for example. Then, the overall control unit 104 can identify which manufacturing process caused the defect by comparing the preliminary inspection result with the inspection result of the defect obtained from the image processing unit 103. As described above, information about the identified manufacturing process (including an alarm indicating which process is a problem) is automatically output to a manufacturing line management system (not shown) via a network. The countermeasures can be taken immediately, and the yield can be improved. Also, like this
By constructing a manufacturing system including an inspection device provided with the binarization determination threshold value determination method according to the present invention, it is possible to produce a high-quality product with few defects and foreign matters.

[0065]

According to the present invention, a true difference portion (true defect portion) is extracted from a difference image between two image signals containing an irregular luminance change with an erroneous extraction rate as low as possible. It has an effect that can be made possible. In particular, in the case of a charged particle beam image as the two image signals, a great effect can be obtained. Further, according to the present invention, it is possible to extract and judge a very minute defect from a detected image signal having an irregular luminance change detected by a charged particle microscope with the erroneous extraction rate as low as possible. There is an effect that an inspection device including a review device using a particle microscope can be realized.

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration of an SEM visual inspection device which is an embodiment of an inspection device including a review device using a charged particle microscope according to the present invention.

FIG. 2 is a diagram showing an embodiment of an image signal targeted by the present invention.

FIG. 3 is a process flowchart showing a first embodiment until a binarization determination threshold value is calculated according to the present invention.

FIG. 4 is a diagram showing an embodiment of a calculated gray value histogram according to the present invention.

FIG. 5 is a diagram showing an embodiment in which a curve is applied to a gray value histogram according to the present invention.

FIG. 6 is a processing flowchart showing a second embodiment for calculating an erroneous extraction occurrence ratio with respect to an extraction region area according to the present invention.

FIG. 7 is a diagram showing an example of a graph used for a method of calculating a false extraction occurrence ratio with respect to an extraction region area.

FIG. 8 is a diagram showing an embodiment of a user interface screen.

FIG. 9 is a diagram illustrating an example of a user interface screen.

FIG. 10 is a diagram for explaining a method of using an inspection apparatus in which the binarization determination threshold value automatic determination method according to the present invention is installed in a production line.

[Explanation of symbols]

2 stage, 31 electron gun, 32 condenser lens, 33 objective lens, 34 scanning deflector, 35 electron detector, 36 amplifier, 39 A / D converter, 40 preprocessing circuit, 41 ... Image memory, 42... Pixel-based positioning unit, 46... Difference extracting circuit, 47... Gray-scale histogram distribution creating circuit 48. ... Characteristic extraction circuit, 100 ... Object to be inspected, 101 ... Detector, 102 ...
Image take-out unit, 103: Image processing unit, 104: Overall control unit, 105: Examination room, 106: Electro-optical system, 107 ...
Electronic detection unit, 109: sample chamber, 120: digital image data (gradation image data), 201: input device, 202:
Display device, 203 ... Storage device.

 ────────────────────────────────────────────────── ─── Continuing on the front page (72) Inventor Ryo Nakagaki 292 Yoshida-cho, Totsuka-ku, Yokohama-shi, Kanagawa Prefecture F-term in Hitachi, Ltd. Production Engineering Laboratory 5B057 AA03 BA01 BA24 CE06 CE12 DA03 DA08 DC22 DC33 5L096 BA03 CA24 EA43 FA37 GA08 GA51 GA55 HA07

Claims (17)

[Claims] 1. A process for creating an intensity distribution due to an irregular change in luminance contained in one of two image signals based on one of the image signals, and A setting step of setting a binarization determination threshold value based on the intensity distribution due to the irregular luminance change, the rate of occurrence of erroneous extraction due to the irregular luminance change being equal to or less than a desired ratio; A difference calculation step of performing a difference calculation on the signal; and a difference between the two image signals using a binarization determination threshold set in the setting step with respect to the difference image signal calculated by the difference in the difference calculation step. An extraction processing step of extracting a part. 2. A process for creating an intensity distribution based on an irregular luminance change included in one of the two image signals based on one of the image signals; and A setting process of setting a binarization determination threshold value based on an intensity distribution due to an irregular luminance change so that a rate of occurrence of erroneous extraction due to an irregular luminance change in a predetermined image area range is equal to or less than a desired rate. A difference calculation step of calculating a difference between the two image signals; and a difference image signal calculated by the difference in the difference calculation step, using a binarization determination threshold value set in the setting step. An extraction processing step of repeating extraction processing of a difference between the two image signals in a predetermined image area range. 3. The method according to claim 1, wherein in said creating step, the intensity distribution due to the irregular luminance change is calculated from the one image signal and a pixel signal obtained by translating the one pixel signal by a predetermined amount in a predetermined direction. 3. The image processing method according to claim 1, wherein the image data is created as a statistic of a grayscale frequency distribution based on the calculated difference image signal. 4. The method according to claim 1, further comprising the step of outputting a rate of erroneous extraction due to irregular luminance change or a rate expected to be an actual difference with respect to a frequency of a difference having a certain size extracted from the extraction process. 3. The image processing method according to claim 1, further comprising: 5. A creating process for creating an intensity distribution based on an irregular luminance change included in one of two image signals based on one of the image signals, and a binarization determination threshold value. A difference calculation step of calculating a difference between the two image signals; and a binarization determination threshold set in the setting step for the difference image signal calculated by the difference in the difference calculation step. An extraction process of extracting a difference portion between the two image signals using a value, and a difference extracted in the extraction process based on an intensity distribution due to an irregular luminance change created in the creation process. A calculating step of calculating a rate of occurrence of erroneous extraction according to the size of the portion. 6. A creation process for creating an intensity distribution due to an irregular change in luminance contained in one of the two image signals based on one of the image signals, and a binarization determination threshold value. A difference calculation step of calculating a difference between the two image signals; and a binarization determination threshold set in the setting step for the difference image signal calculated by the difference in the difference calculation step. An extraction processing step of extracting a difference portion between the two image signals using a value, and a ratio expected to be an actual difference portion according to the size of the difference portion extracted in the extraction processing step. And a calculating step of calculating the following. 7. The image processing method according to claim 5, further comprising an output step of outputting the ratio calculated in said calculation step. 8. A creating section for creating an intensity distribution based on an irregular luminance change included in one of the two image signals based on one of the image signals, and a creating section for creating the intensity distribution. A setting unit for setting a binarization determination threshold value based on the intensity distribution due to the irregular luminance change, the rate of occurrence of erroneous extraction due to the irregular luminance change being equal to or less than a desired ratio; A difference operation unit for performing a difference operation on the signal, and a difference image signal calculated by the difference operation unit in the difference operation unit.
An image processing apparatus, comprising: an extraction processing unit configured to extract a difference between the two image signals using a binarization determination threshold set by the setting unit. 9. A creating section for creating an intensity distribution based on an irregular change in luminance contained in one of the two image signals based on one of the image signals, and a creating section for creating the intensity distribution. A setting unit that sets a binarization determination threshold value based on the intensity distribution due to the irregular luminance change so that the rate of occurrence of erroneous extraction due to the irregular luminance change in a predetermined image area range is equal to or less than a desired rate; And a difference calculation unit that calculates a difference between the two image signals, and a difference image signal calculated by the difference calculation unit in the difference calculation unit.
An extraction processing unit that repeats extraction processing of a difference portion between the two image signals in the predetermined image area range using a binarization determination threshold value set by the setting unit. Image processing device. 10. The method according to claim 1, wherein the generating unit calculates the intensity distribution due to the irregular luminance change from the one image signal and a pixel signal obtained by translating the one pixel signal in a predetermined direction by a predetermined amount. The image processing apparatus according to claim 8, wherein the image processing apparatus is configured to generate a statistical value of a grayscale frequency distribution based on the calculated difference image signal. 11. The method according to claim 1, further comprising the step of outputting a rate of erroneous extraction due to an irregular luminance change or a rate expected to be an actual difference with respect to a frequency of a difference of a certain size extracted from the extraction processing unit. The image processing apparatus according to claim 8, further comprising an output unit configured to perform the operation. 12. An image processing apparatus according to claim 8, wherein said setting section has input means for inputting a binarization determination threshold value. 13. An image processing apparatus according to claim 8, wherein said setting section has an input means for inputting a rate of occurrence of erroneous extraction due to irregular luminance change. 14. A creating section for creating an intensity distribution based on an irregular luminance change contained in one of two image signals based on one of the image signals, and a binarization determination threshold value. A setting unit that sets the following; a difference calculation unit that calculates a difference between the two image signals; and a difference image signal calculated by the difference calculation unit.
An extraction processing unit that extracts and processes a difference between the two image signals using a binarization determination threshold value set by the setting unit; and an intensity distribution due to an irregular luminance change created by the creation unit. An image processing apparatus comprising: a calculating unit for calculating a rate of occurrence of erroneous extraction according to a size of a different part extracted by the extraction processing unit. 15. A creating section for creating an intensity distribution based on an irregular luminance change included in one of two image signals based on one of the image signals, and a binarization determination threshold value. A setting unit that sets the following; a difference calculation unit that calculates a difference between the two image signals; and a difference image signal calculated by the difference calculation unit.
An extraction processing unit that extracts and processes a different portion between the two image signals using a binarization determination threshold value set by the setting unit; And a calculating unit for calculating a ratio expected to be an actual difference. 16. The image processing apparatus according to claim 14, further comprising an output unit for outputting the ratio calculated by said calculation unit. 17. An eighth or ninth or fourteenth or fifteenth aspect.
A charged particle microscope inspection device, comprising the image processing device according to any one of the preceding claims.