JP4255565B2 - Method and apparatus for inspecting periodic pattern unevenness - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
INDUSTRIAL APPLICABILITY The present invention is applied to a periodic pattern unevenness inspection method and apparatus, and in particular, to inspect periodic pattern irregularities (unevenness) based on image data obtained by imaging an object having a periodic pattern. In particular, the present invention relates to a method and apparatus for inspecting unevenness of a periodic pattern.
[0002]
[Prior art]
Examples of objects having a periodic pattern include a shadow mask of a color CRT display, a color separation filter of a color image pickup tube, a color filter of a liquid crystal display device, a mesh electrode of an electron tube, a VDT filter, a photomask, a Fresnel lens, and a lenticular. There are precision industrial products such as lenses and many others.
[0003]
As a technique for inspecting these objects, for example, a second-order differential filter process in a predetermined direction is performed on image data obtained by imaging the inspection object, which is disclosed in JP-A-6-258249. A method for detecting a direction streak defect or a device disclosed in Japanese Patent Application Laid-Open No. 6-265476 that emphasizes unevenness of transmittance of an inspection object by image processing and displays it in an easy-to-see manner to assist human visual inspection. Are known.
[0004]
[Problems to be solved by the invention]
These inspection technologies have inspection performance as they are, but for example, if it is necessary to inspect a target to which fine foreign matters etc. that are to be removed later by cleaning etc. are attached, In order to prevent detection, it is necessary to lower the detection sensitivity for the fine foreign matter, while increasing the detection sensitivity for the inspection target. However, in the conventional inspection technique, for example, a process of smoothing image data is generally performed in order to reduce the detection sensitivity for fine foreign matters. When the inspection is performed based on the image data that has been subjected to such smoothing processing, there is a problem that the inspection performance is remarkably lowered because the image quality is deteriorated by the processing.
[0005]
Therefore, since the present applicant detects that it is not defective but is not defective, such as a fine adhered foreign material to be removed, even if it is an inspection object that has noise when detecting the defect, it is high. Specializes in pattern inspection technology that can provide inspection performance. Wish This has already been proposed in Hei 10-197815.
[0006]
That is, basically, a noise region is extracted from inspection target image data obtained by imaging the inspection target to generate noise region extraction image data, and the noise region of the inspection target image data indicated by the noise region extraction image data Is replaced with a representative value of pixels around the noise region to obtain noise region-removed image data, and by extracting a defect to be inspected based on the noise region-removed image data, the cause of noise is detected at the time of defect detection. Even if it is an inspection object in which fine foreign matter is present, it is possible to inspect irregularities of the periodic pattern with high accuracy.
[0007]
Hereinafter, this technique will be described in detail with a specific example of a pattern inspection apparatus. FIG. 1 shows a schematic configuration of the pattern inspection apparatus. Reference numeral 1 denotes an imaging apparatus, 2 denotes a main body of the pattern inspection apparatus, 3 denotes a display device, 4 denotes an inspection target, 5 denotes an inspection stage, 6 denotes a light diffusion plate, 7 Is an illumination light source, and 8 is a stabilized power source. The imaging device 1 is not particularly limited as long as it is a device that images a two-dimensional region and outputs an imaging signal. For example, a CCD camera that uses a two-dimensional solid-state imaging sensor such as a CCD can be used.
[0008]
The main body 2 of the pattern inspection apparatus executes each processing necessary for inspection including processing such as image processing described in detail below. The main body 2 is configured by hardware and software of a data processing device such as a personal computer, and can be configured only by a general-purpose or dedicated data processing device. In addition, a general-purpose data processing device and a dedicated image processing device can be combined to perform image processing at high speed. Further, it has an interface for inputting an image pickup signal output from the image pickup apparatus 1 to convert it into digital data and storing it as image data in a storage device. Although not shown in FIG. 1, the input / output device includes known peripheral devices such as a keyboard, a mouse, and a printer. Also, data for setting inspection conditions according to the production items in the manufacturing apparatus is input from a production management system connected to a LAN (local area network), or inspection status data is output to the production management system. You may do it.
[0009]
As the display device 3, a display device such as a CRT or a liquid crystal can be used. The display device 3 displays an imaging screen to be inspected, an image in the process of extracting a defect, and a screen indicating the position, size, and type of the extracted defect. In addition, a menu display for operating the pattern inspection apparatus and a display showing the processing status of the pattern inspection apparatus are performed.
[0010]
The inspection object 4 has the form of a web or a sheet, and when the inspection is performed, it is placed on the surface of the stage 5 as shown in FIG. The stage 5 holds the inspection object 4, and in the inspection area at the central portion of the inspection object 4, the stage 5 is made of a plate material such as transparent glass or plastic that allows light to pass through or is made a space to prevent light from passing. Do not. The imaging apparatus 1 performs imaging with light passing through the inspection object 4.
[0011]
The light diffusing plate 6 diffuses the light from the illumination light source 7 to form back illumination with uniform characteristics over the entire inspection region. The light diffusing plate 6 may be made of a glass whose surface is matted, a transparent material such as plastic, or a material such as plastic containing a light diffusing substance such as a white pigment. The illumination light source 7 can be a fluorescent lamp, an incandescent lamp, a halogen lamp, or the like, and is not particularly limited. However, the illumination light source 7 needs to have a change in brightness (luminance) within a predetermined range acceptable for imaging for inspection while the imaging apparatus 1 performs imaging. The stabilized power supply 8 is for that purpose, and keeps the brightness of the illumination light source 7 stable. As the stabilizing power source 8, a DC power source is usually used. However, in some cases, for example, a high frequency AC power source is used in combination with a fluorescent lamp having a long afterglow time.
[0012]
Next, the operation of the pattern inspection apparatus having the above configuration will be described. The basic processing steps executed in this pattern inspection apparatus are shown in the flowchart of FIG. As shown in FIG. 2, the process is composed of three steps: a noise area extraction process in step 1, a noise area removal process in step 2, and a pattern defect extraction process in step 3. Before the process shown in FIG. 2, the pattern inspection apparatus shown in FIG. 1 takes an image without setting the inspection object 4 on the inspection stage 5 for the purpose of correcting unevenness (distribution) of the brightness of the transmitted illumination. Imaging is performed by the apparatus 1, and image data for correction is stored in the main body 2 (details will be described later). In addition, imaging is performed by the imaging apparatus 1 with the inspection object 4 set on the inspection stage 5, and inspection image data (inspection object image data) is stored in the main body 2. The process shown in FIG. 2 starts from that state.
[0013]
In the noise area extraction process of Step 1, a process of generating a noise area extraction image data by extracting a noise area from inspection object image data obtained by imaging an inspection object is performed. In the noise region removal processing in step 2, processing for creating noise region-removed image data by replacing the noise region of the inspection target image data indicated by the noise region extraction image data with a representative value of pixels around the noise region. Is done. In the pattern defect extraction process in step 3, a process of extracting defects to be inspected based on the noise area removed image data is performed.
[0014]
An image of the image data in the process shown in FIG. 2 executed in the pattern inspection apparatus is shown in the schematic diagram of FIG. FIG. 3A shows image data obtained by imaging the inspection object 4, and this image data includes a black spot portion and a dust deposit portion. A black spot portion is a portion to be extracted as a defect, and a dust-attached portion is a portion that is not extracted as a defect. This inspection object is a shadow mask which is one of the articles to be inspected by the pattern inspection apparatus. FIG. 3B is an enlarged view of the dust adhering portion, and shows the state of the dust covering the through hole of the shadow mask.
[0015]
Image data is a collection of pixel values arranged in a matrix. The pixel value, that is, the pixel value is a value related to the brightness (transmittance, luminance, density, etc.) of the image at that position.
FIG. 3C is a graph showing pixel values for one row on the line L indicated by the one-dot chain line in FIG. 3A as a graph (profile such as luminance). On the line L, a black spot portion and a dust deposit portion are included. As shown in FIG. 3C, the pixel value in the black spot portion is slightly smaller than the peripheral pixel value, whereas the pixel value in the dust adhering portion is the peripheral pixel value. Is much smaller than.
[0016]
In the noise area extraction process of step 1, the pixel of the dust adhesion portion is extracted based on the characteristic difference between the pixel values of the black spot portion and the dust adhesion portion (details will be described later). FIG. 3D is a diagram showing, as a graph, pixel values corresponding to FIG. In the noise region removal in the next step 2, the part of FIG. 3D in FIG. 3C is replaced with the pixel value (representative value) of the part in the vicinity of FIG. 3D in FIG. Processing is performed. As a result, a diagram shown in FIG. 3E corresponding to the graph in which the attached portion of dust in FIG. 3C is removed is obtained.
[0017]
For convenience of explanation, FIG. 3 shows a process of processing for one row of pixels, but actual processing is performed on pixels arranged in a two-dimensional matrix. However, for convenience, image data other than a two-dimensionally displayed planar image will be described below using profile data corresponding to the luminance on the line L as described above.
[0018]
Next, the process in the pattern inspection apparatus will be described in detail. A preferred example of the process of extracting the noise region shown in step 1 of FIG. 2 is shown in the flowchart of FIG. FIG. 5 shows a schematic diagram corresponding to FIG. 3 in the process of extracting the noise region shown in FIG. FIG. 5A is a diagram showing inspection object image data obtained by imaging the inspection object 4. This image data includes a black spot portion, a white spot portion, and two dust deposit portions. The black spot portion and the white spot portion are portions to be extracted as defective, and the two dust-attached portions are portions that are not extracted as defective. This inspection object is a shadow mask as in the case of FIG.
[0019]
FIG. 5B is a graph (profile) on the line L in FIG. 5A. The line L includes a black spot portion, a white spot portion, and two dust deposit portions. It is. As shown in the graph of FIG. 5B, the pixel value in the black spot portion is slightly smaller than the surrounding pixel values. On the contrary, the pixel value in the white spot portion is slightly larger than the surrounding pixel values. In addition, the pixel values at the two dust-attached portions are extremely smaller than the surrounding pixel values.
[0020]
First, in the smoothing process in step 11 of FIG. 4, a process of smoothing the inspection target image data of FIGS. 5A and 5B obtained by imaging the inspection target is performed, and the smoothed image data is converted into the smoothed image data. create. The smoothing process is a process of weakening high spatial frequency components of image data by a spatial filter. For example, the pixel value P (i, j) of the pixel of interest and the pixel values P (i + 1, j), P (i + 1, j + 1), P (i, j + 1), P (i-1, j + 1), P (i−1, j), P (i, j−1), P (i−1, j−1), and P (i + 1, j−1), the new pixel value P1 of the target pixel A method of calculating (i, j) by the following equation (1) (average value filter), a method of calculating by the following equation (2) (median filter), and a method of calculating by the following equation (3) (mode Value filter), etc. are known.
[0021]
P1 (i, j) = (P (i, j) + P (i + 1, j) + P (i + 1, j + 1) + P (i, j + 1) + P (i-1, j + 1) + P (i-1, j) + P ( i, j-1) + P (i-1, j-1) + P (i + 1, j-1)) / 9 (1) P1 (i, j) = median value (P (i, j), P ( i + 1, j), P (i + 1, j + 1), P (i, j + 1), P (i-1, j + 1), P (i-1, j), P (i, j-1), P (i- 1, j-1), P (i + 1, j-1)) (2)
[0022]
However, the median value () is a value (median) whose size is centered in the numerical group in ().
[0023]
P1 (i, j) = mode (P (i, j), P (i + 1, j), P (i + 1, j + 1), P (i, j + 1), P (i-1, j + 1), P ( i-1, j), P (i, j-1), P (i-1, j-1), P (i + 1, j-1)) (3)
[0024]
However, the mode () is a representative value (usually the median value of the section) of the section with the highest frequency in the frequency distribution of the numerical group in ().
[0025]
The calculation of the above formulas (1) to (3) is performed with all pixels as the target pixel (i = 1, 2,..., N, j = 1, 2,..., M), and a new pixel value. Image data (smoothed image data) is obtained. This processing by the spatial filter can be repeated if necessary. By the processing in step 11, the smoothed image data in FIG. 5C can be created from the inspection target image data in FIG. 5B.
[0026]
Next, in the image subtraction process in step 12 of FIG. 4, a process of subtracting the pixel value of the corresponding pixel of the inspection target image data from the pixel value of each pixel of the smoothed image data is performed using the following equation (4). To create subtracted image data. By this processing, the subtraction image data in FIG. 5D can be created from the inspection target image data in FIG. 5B and the smoothed image data in FIG. 5C.
[0027]
P2 (i, j) = P1 (i, j) -P (i, j) (4)
Where P (i, j): pixel value of the array (i, j) of the inspection target image data
P1 (i, j): Pixel value of array (i, j) of smoothed image data
P2 (i, j): Pixel value of array (i, j) of subtraction image data
i: 1, 2, ..., n
j: 1, 2, ..., m
[0028]
Next, in the binarization process of step 13 in FIG. 4, a process of binarizing the subtraction image data is performed to generate binarized image data. As the threshold value to be set in the binarization process, for example, a well-known method for setting the threshold value by applying a discriminant analysis method of multivariate analysis can be used. By this processing, the subtraction image data (multi-valued data) in FIG. 5D is binarized using a threshold indicated by a broken line, so that the subtraction image data (binary data) in FIG. Can be created.
[0029]
As is apparent from a comparison between FIG. 5A and FIG. 5E, the part corresponding to the noise area due to dust in FIG. 5A has a pixel value “1”, and the other part has a pixel value “0”. "It has become.
[0030]
Next, in the expansion process of step 14 in FIG. 4, a process of expanding the image portion represented by the pixel value “1” is performed on the binarized image data, and the noise area extraction image data (expanded image data) is obtained. Generate. Since the method of expansion processing for this binarized image is well known, the description thereof is omitted here. By this processing, the binarized image data in FIG. 5E can be expanded to create the noise region extracted image data in FIG.
[0031]
As is clear when FIG. 5E and FIG. 5F are compared, the portion of the image represented by the pixel value “1” in FIG. 5E expands to expand the range. As a result, the portion of the noise area due to dust is completely included in the portion of the pixel value “1” in FIG.
[0032]
Next, the noise region removal process shown in step 2 of FIG. 2 by the pattern inspection apparatus will be described with reference to the flowchart of FIG. Similarly to the case shown in FIG. 3, FIGS. 7 and 8 are schematic views showing the process of extracting the noise region shown in FIG.
[0033]
First, in the labeling process in step 21 of FIG. 6, a label number for identifying each isolated region of the noise region extracted image data of FIG. 5 (F) obtained in the above-described noise region extraction process is assigned. To create labeled image data. The labeling process is, for example, a process for assigning a pixel number to each isolated region as a pixel value, and if the adjacent pixel has a pixel value “1”, the label number is assigned and If the pixel to be processed is the pixel value “0”, it is left as it is. Since this labeling process is well known, detailed description thereof is omitted here.
[0034]
By this processing, the noise area extracted image data of FIG. 5 (F) can be labeled to generate the noise area extracted image data of FIG. 7 (A1). FIG. 7A2 is a schematic diagram illustrating a planar image of labeled image data. A dashed-dotted line in FIG. 7A2 indicates a correspondence relationship with the line L in FIG. The same applies to the other drawings in FIGS.
[0035]
Next, in the label N extraction process of step 22 in FIG. 6, a process of leaving only the label N area of the labeled image data is performed to generate label N image data. With this process, the label N image data of FIG. 7B1 can be generated by performing the label N extraction process on the above-described labeled image data of FIG. 7A1. FIG. 7B2 is a schematic diagram showing a planar image of label N image data. As can be seen by comparing FIG. 7 (A2) and FIG. 7 (B2), only the isolated point with the number “1” is extracted here.
[0036]
Next, in the second expansion process in step 23 of FIG. 6, a process of expanding the label N image data is performed to generate label N expanded image data. By this process, the label N image data shown in FIG. 7C1 can be created by expanding the label N image data shown in FIG. 7B1. As can be seen by comparing the planar image of the label N expanded image data shown in FIG. 7C2 with FIG. 7B2, here, the isolated point with the number “1” is shown expanded.
[0037]
Next, in the second image subtraction process in step 24 of FIG. 6, a process of subtracting the pixel value of the corresponding pixel of the label N image data from the pixel value of each pixel of the label N expanded image data is performed. Label N subtraction image data is created. By this processing, the label N expanded image data of FIG. 7 (C1) is subjected to the second image subtraction process by the label N image data of FIG. 7 (B1) to generate the label N subtraction image data of FIG. 8 (D1). Can do. As can be seen by comparing the planar image of the label N subtraction image data shown in FIG. 8 (D2) with FIG. 7 (C2), here, only the expanded portion of the isolated point with the number “1” has a donut shape. Shown left. That is, a binary image in which pixels around noise due to dust on the left side of FIG. 5B are extracted is obtained.
[0038]
Next, in the representative value calculation processing in step 25 of FIG. 6, the pixel value of the pixel corresponding to the label N subtraction image data from the above-described inspection target image data (see FIGS. 5A and 5B) is obtained. Extraction and processing for calculating a representative value from the pixel value are performed. By this processing, the pixel values of the inspection target image data in FIG. 5B corresponding to the label N subtraction image data in FIG. 8D1 are extracted as image data shown in FIG. 8E1. . FIG. 8E2 is a schematic diagram showing pixels in the planar image extracted from the inspection target image data, and pixel values of these pixels are extracted by this processing. The difference between FIG. 8 (D2) and FIG. 8 (E2) is that the former is binary data while the latter is multi-value data. A representative value is calculated based on the extracted pixel value. As the representative value, an average value, a median value, a mode value, or the like can be used.
[0039]
Next, in the pixel replacement process of step 26 of FIG. 6, a process of replacing the pixel value of the pixel corresponding to the label N image data from the inspection target image data with the above-described representative value is performed, and the label N area removed image data Create By this process, the inspection target image data shown in FIG. 5B can be replaced, and the label N area removed image data shown in FIG. 8F1 can be created. As can be seen by comparing the planar image of the label N region removed image data shown in FIG. 8F2 with FIG. 5A, here, the isolated point with the number “1” is removed. .
[0040]
Next, in the iterative control process of step 27 in FIG. 6, the above-described process is repeated for all the label N areas, and the process of removing all the label N areas from the inspection target image data is controlled to obtain a noise area. Create removal image data.
[0041]
Next, the pattern defect extraction process shown in step 3 of FIG. 2 by the pattern inspection apparatus will be described with reference to the flowchart of FIG.
[0042]
First, in step 31, processing for correcting the light distribution is performed. In this light distribution correction process, the two-dimensional distribution of the amount of light when the inspection object 4 is illuminated by the transmission illumination system including the light diffusing plate 6, the illumination light source 7, the stabilized power supply 8 and the like shown in FIG. to correct. Also, the two-dimensional distribution of sensitivity due to the imaging optical system and the imaging device is corrected. In the light distribution correction process in step 31, imaging is performed without setting the inspection target 4 on the inspection stage 5 to obtain correction image data (light source image data).
[0043]
Further, the above-described noise-removed image data is divided by this correction image data. Division is performed between corresponding pixel values. Thereby, noise area-removed image data (transmittance image data) subjected to the light distribution correction process is obtained. Before performing this division, the noise area-removed image data is multiplied by an appropriate constant so that the data obtained by the division is within a predetermined value range, for example, data 0 to 255 having a gradation value of 8 bits. To do. As a result, it is possible to prevent the number of valid digits of data from being reduced.
[0044]
Next, in step 32, a process for smoothing the noise area-removed image data that has been subjected to the light distribution correction process is performed to create smoothed noise area-removed image data. The smoothing process is substantially the same as the smoothing process in step 11 of FIG. By this smoothing processing, the intensity of a specific region (high spatial frequency component) of the spatial frequency component of the noise region-removed image data is weakened, and smoothed image data is obtained.
[0045]
Next, in step 33, a secondary differential process is performed on the smoothed noise area-removed image data to create secondary differential noise area-removed image data. This secondary differentiation process is performed by processing image data with a spatial filter of secondary differentiation. The second-order differentiated image data is obtained by the configuration of the spatial filter and the set value.
[0046]
Next, in step 34, a process for smoothing the second-order differential noise region-removed image data is performed to create second smoothed noise region-removed image data. The smoothing process is the same as in step 32 above. By smoothing processing, the intensity of a specific region (high spatial frequency component) of the spatial frequency component of the secondary differential noise region-removed image data is weakened to obtain smoothed image data.
[0047]
Spatial filter processing as described above is performed so that the defective portion is emphasized from the noise region-removed image data. That is, when the reverse expression is performed, the spatial filter processing is performed so that the spatial frequency component of the defective portion is emphasized. In step 35, binarization processing is performed on the second smoothed noise region-removed image data obtained in this way, and defective extracted image data is created.
[0048]
Next, in step 36, a process for determining pass / fail is performed on the defective extracted image data, and pass / fail determination data is created. The determination of pass / fail is made in comparison with a determination standard that defines the position, size (area), number, etc. of defective portions.
[0049]
An example of the form of the defect extracted by the pattern inspection apparatus is shown in FIG. 10 as a schematic diagram. 10A shows a white spot defect, FIG. 10B shows a black spot defect, FIG. 10C shows a white stripe defect, and FIG. 10D shows a black stripe defect. In each of FIGS. 10A to 10D, the figure surrounded by the left rectangle is a captured image to be inspected, and the right figure is an enlarged view of each defect when the inspection object is a shadow mask. And the regularly arranged circles schematically show the shadow mask holes, respectively.
[0050]
The white spot defect shown in FIG. 10A is a defect in which holes larger than the standard are distributed, and the black spot defect shown in FIG. The white streak defect shown in (C) of FIG. 10 is a defect in which holes larger than the standard are distributed, and the black scratch defect shown in (D) is a defect in which holes smaller than the standard are distributed in a streak pattern. In each of FIGS. 10A to 10D, the enlarged diagram on the right side shows the defect size (range) smaller than the actual defect for explanation. In general, the defect size (range) is larger than the dust shown in FIG.
[0051]
Although the proposed pattern inspection technique has been described in detail with specific examples, the present invention is not limited to this. For example, the light distribution correction is performed on the noise area-removed image data in step 31, but the light distribution correction is performed on the inspection target image data immediately after obtaining the inspection target image data by imaging. Can be substituted. In general, image processing has a processing process in which a large difference does not occur even if the order is changed. Therefore, the processing steps described above are not necessarily limited to the order.
[0052]
Further, for example, an example of the pattern defect extraction process has been shown, but the process is appropriately set according to the content of the defect and is not necessarily limited to that example. Further, although the shadow mask has been described as an inspection target, the present invention is not limited to this.
[0053]
As described in detail above, in the periodic pattern unevenness inspection technology proposed by the present applicant, a dust portion is extracted from captured image data and replaced with a representative value such as an average value of surrounding pixels. Since the dust part (area) is removed, the dust part can be removed from the inspection image without degrading the image quality of the inspection image, and the inspection performance is not deteriorated as compared with the conventional method. There is an excellent advantage that inspection without detection becomes possible.
[0054]
However, in the proposed inspection technique, when a lot of electrical random noise is included in the captured image, such electrical noise is also recognized as dust. It has become clear that there is a new problem that inspection requires a huge amount of time because the dust removal process must be performed for noise.
[0055]
The present invention has been made to solve the above-described new problems, and even when a lot of electrical noise is included in an inspection image obtained by imaging an inspection object, it is accurate and quick. It is an object of the present invention to provide a periodic pattern unevenness inspection method and apparatus capable of inspecting a periodic pattern.
[0056]
[Means for Solving the Problems]
The present invention relates to a method for inspecting periodic pattern unevenness, a noise region extraction process for generating noise region extraction image data by extracting a noise region from inspection target image data obtained by imaging an inspection target, and the noise region From extracted image data It occurs only in one pixel unit Excluding electrical noise Consists of noise area of 2 pixels or more Electrical noise exclusion process to create residual noise area image data, Eliminate electrical noise The residual noise region removing step of replacing the inspection target image data corresponding to the residual noise region in the residual noise region image data with a representative value of pixels around the residual noise region to create residual noise region-removed image data; An electrical noise removing process for removing the excluded electrical noise and the electrical noise are removed from the residual noise area removed image data. A pattern defect extraction process for extracting defects to be inspected based on the residual noise area removed image data; , Have At the same time, the electrical noise removal process includes a noise area measurement process for measuring the area of each noise area extracted in the noise area extraction image data, and the measured area is one pixel. An electrical noise determination process for determining noise By doing so, the above-mentioned problems are solved.
[0057]
The present invention is also obtained by imaging an inspection object in a periodic pattern unevenness inspection apparatus. Et A noise region extraction means for extracting a noise region from the image data to be inspected to create noise region extraction image data, and the noise region extraction image data It occurs only in one pixel unit Excluding electrical noise Consists of noise area of 2 pixels or more Electrical noise exclusion means for creating residual noise area image data; Eliminate electrical noise A residual noise region removing unit that replaces the inspection target image data corresponding to the residual noise region in the residual noise region image data with a representative value of pixels around the residual noise region, and creates residual noise region-removed image data; Electrical noise removal means for removing the electrical noise that has been excluded from the residual noise area removal image data, and electrical noise has been removed Pattern defect extraction means for extracting defects to be inspected based on the residual noise area removed image data; , Has At the same time, the electrical noise exclusion means is electrically connected to the noise area measurement means for measuring the area of each noise area extracted from the noise area extraction image data, and the measured area is equivalent to one pixel. Electrical noise determination means for determining noise By adopting a configuration, the above-described problem is solved in the same manner.
[0058]
That is, in the present invention, the noise area is extracted from the image data to be inspected, and the electrical noise is excluded in advance when inspecting the periodic pattern unevenness after removing the area from the image data to be inspected. Because it is only necessary to perform noise removal processing only for noise caused by dust (fine foreign matter), even if there are many electrical noises in the image data to be inspected, only noise that becomes an obstacle to inspection is removed. Therefore, the periodic pattern irregularity inspection can be performed accurately and promptly. Note that the electrical noise present in the inspection target image data is generated only in units of one pixel, and therefore can be easily removed by a normal smoothing process or the like.
[0059]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
[0060]
In the present embodiment, the same apparatus as the inspection apparatus schematically shown in FIG. 1 is used, and basic processing is performed according to the procedure shown in the flowchart of FIG. Each process described in detail below is executed by the inspection apparatus body 2 in the same manner as described above.
[0061]
The flowchart of FIG. 11 is substantially the same as the flowchart of FIG. 2 except that the electrical noise elimination process of step 42 is performed between steps 1 and 2 of FIG. . Therefore, steps 41, 43, and 44 correspond to steps 1, 2, and 3 in FIG.
[0062]
FIG. 12 is image data corresponding to FIG. 3 schematically showing the features of the processing in steps 41 to 43 in the present embodiment. FIG. 12A shows image data of an inspection object (shadow mask) similar to FIG. 3A except that electrical noise is present, and FIG. 12B shows a dashed line in FIG. A profile (image data) of luminance (pixel value) on the line L (for one pixel) indicated by, and corresponds to FIG.
[0063]
First, a noise region is extracted from the profile data shown in FIG. 12B, and noise region extracted image data shown in FIG. 12C is created (step 41). Processing to exclude noise is performed (step 42).
[0064]
In this electrical noise exclusion process, the area of each extracted noise region is measured in pixel units using the noise region extracted image data. FIG. 12D shows an image of the measurement result. It is known that electrical noise appears only in units of one pixel, and noise caused by dust is sufficiently larger than electrical noise.
[0065]
Therefore, as shown in FIG. 12E, when the area is 1 pixel, electrical noise is used, and when it is 2 pixels or more, noise is generated. By excluding the former, residual noise area image data (detailed later) is obtained. Create). By excluding electrical noise in this way, the noise region extracted image data in FIG. 12C remains only against noise (residual noise) due to dust as shown in FIG. Noise area (dust) removal processing is performed (step 43).
[0066]
In the specific processing of step 43, the inspection target image data for the remaining noise region is replaced with the representative values of the pixels surrounding the region to create a residual noise region-removed image similar to that shown in FIG. That's true.
[0067]
Next, the outline of the process from the imaging of the sample (inspection object) to the quality determination of the periodic pattern will be described with reference to the flowchart of FIG.
[0068]
First, sample image data is input by imaging a sample, and a noise region is extracted from the sample image data (steps 51 and 52). Next, the area of each extracted noise region is measured, the noise having an area of 1 is regarded as electrical noise, is excluded from the extracted noise region, and only noise due to dust (residual noise region) remains. (Steps 53 and 54). Thereafter, residual noise area-removed image data is created by removing noise due to dust from the sample image (step 55).
[0069]
On the other hand, light source image data is input by imaging the light source 7 with the imaging device 1 without placing the sample (inspection object) on the stage 5 in FIG. 1 (step 56), and after dust is removed. The sample image data (residual noise area removed image data) is divided by the light source image data to obtain transmittance image data that is not affected by the shading of the light source (step 57). In order to remove electrical noise from the transmittance image data, smoothing processing is performed by applying an average value filter, a median filter, etc. (step 58), in order to emphasize unevenness of the periodic pattern Then, the secondary differential processing is performed on the smoothed image data (step 59). This secondary differentiation process is performed using a spatial filter corresponding to the shape of the unevenness (spots / streaks).
[0070]
Further, in order to remove the minute fluctuation region other than the unevenness, smoothing processing (average value filter) is performed (step 60), the image data after smoothing is binarized, and the non-defective product is based on the obtained binary image. Then, a defective product is determined (step 61).
[0071]
The processing procedure according to the present embodiment described above is substantially the same as a series of processing procedures according to the proposed inspection technique, except that two processes of area measurement in step 53 and exclusion of electrical noise in step 54 are added. Are the same. Therefore, although the noise region to be removed is noise due to dust even in the inspection technique, it is a residual noise region because electric noise is excluded in this embodiment.
[0072]
The processing procedure according to this embodiment will be described in further detail. First, the extraction of the noise region in step 52 will be described with reference to the flowchart of FIG. 14 and FIG. 15 schematically showing the characteristics thereof.
[0073]
This noise region extraction processing is substantially the same as the processing by the proposed inspection technique shown in FIG. 4 except that the expansion processing in step 14 is omitted.
[0074]
In step 71, smoothing processing is performed on the sample image data shown in FIG. The sample image data is the same as that shown in FIG. 5A except that white and black electrical noises are mixed. Here, smoothed image data is generated by applying an average value filter having a size larger than noise due to dust to the sample image. For example, when the magnitude of noise corresponds to 3 × 3 pixels, the size of the smoothing filter is about 5 × 5 pixels. FIG. 15B is a luminance profile on the line L of the sample image data in FIG. 15A, and the smoothed image data corresponding to the image data on the line is shown in FIG. C). Hereinafter, similarly, data on the line L is shown as a representative.
[0075]
Next, the subtracted image data in FIG. 15D is created by subtracting the sample image in FIG. 15B from the smoothed image data in FIG. 15C (step 72). In this case, the noise area due to dust is always extremely darker than the surrounding area, so this subtracted image data has a positive value. As a result of this processing, similarly, when the luminance is dark to electrical noise, it has a positive value, and conversely, when the luminance is bright, it has a negative value.
[0076]
Next, the subtracted image data is binarized to create a binary image corresponding to the noise region extracted image data shown in FIG. 15E (step 73). In this binarization processing, the above-described steps 71 to 73 are performed on a sample image whose dust position is known, and a threshold value as shown in FIG. Keep it. In the binary image data, the noise and the luminance due to dust are darker than the surroundings, and the electrical noise on the left side of FIG. 15A remains.
[0077]
Next, each process of measuring the area in step 53 and excluding the electrical noise in step 54 in the flowchart of FIG. 13 will be described in detail with reference to the flowchart of FIG. 16 and FIGS. 17 and 18 showing the features thereof. In FIG. 17 and FIG. 18, the planar image data is shown on the right side, and the luminance profile on the line L is shown on the left side. However, since all are binary images, there are two types of luminance, 0 and 1.
[0078]
FIG. 17A1 is a re-display of the noise area extracted image data (binary image) shown in FIG. 15E. The outer periphery of the noise area indicated by the black circle in the plane image data of FIG. The dashed circle represents the maximum range of the same noise region before binarization in the sample image data of FIG.
[0079]
First, in step 81, the area of each noise region of the noise region extracted image data is measured, and area measurement image data shown in FIGS. (B1) and (B2) are created. Next, in step 82, the area determination image data is subjected to noise determination in which area 1 is electrical noise and area 2 or more is noise due to dust, and the noise determination image whose images are shown in FIGS. 18C1 and 18C2. Create data.
[0080]
Next, in step 83, electrical noise excluded image data shown in FIGS. 18D1 and 18D, in which the electrical noise of area 1 is excluded, is created from the noise determination image data. In step 84, the electrical noise is excluded. The image data is expanded by a predetermined number of pixels, and the expansion processing is performed so that the maximum range of the broken line is obtained, and binarized residual noise area image data shown in FIGS. (E1) and (E2) is created. To do. 18D1 and 18E correspond to FIGS. 5E and 5F, respectively.
[0081]
Next, a process of removing noise (residual noise) due to dust extracted in the residual noise area image data of FIG. 18 (E1) from the sample image data shown in FIG. Execute according to the flowchart.
[0082]
This flowchart is only partially detailed, and is substantially the same as the noise region removal processing procedure shown in FIG. 6. Therefore, each step is indicated by the same number with a dash “′”, and the description thereof will be given. Is omitted. Accordingly, the image data representing the characteristics of each step is substantially the same as that shown in FIGS. However, the noise area is read as the remaining noise area. Further, as shown in FIG. 20, the sample image after the removal of the first residual noise region remains in the state in which the electrical noise that originally existed remains, as shown in FIG. 8 (F2). ) Is different.
[0083]
According to the flowchart shown in FIG. 19, after all noise due to dust is removed from the sample image, the processing of steps 56 to 61 shown in FIG. 13 is performed to inspect the irregularity of the periodic pattern. At this time, the electrical noise existing on the sample image data can be completely removed by the smoothing process in step 58 regardless of the number of the electrical noises. The periodic pattern unevenness inspection processing executed here is the same as the procedure shown in the pattern defect extraction flowchart shown in FIG.
[0084]
According to the embodiment described above in detail, even when a lot of electrical random noise is included in the sample image (inspection target image data), only noise due to dust is extracted from the noise detected from the sample image. Since the electrical random noise is removed by the smoothing process, the noise can be removed efficiently and at high speed, and the irregularity of the periodic pattern can be inspected accurately and quickly.
[0085]
Although the present invention has been specifically described above, the present invention is not limited to that shown in the above embodiment, and various modifications can be made without departing from the scope of the invention.
[0086]
For example, in the embodiment, the shadow mask is shown as the inspection target (sample), but the present invention is not limited to this.
[0087]
【The invention's effect】
As described above, according to the present invention, a periodic pattern can be inspected quickly and accurately even when a lot of electrical noise is included in an inspection image obtained by imaging an inspection object. Can do.
[Brief description of the drawings]
FIG. 1 is an explanatory diagram showing an outline of the configuration of a proposed pattern inspection apparatus
FIG. 2 is a flowchart showing an outline of a processing process in a proposed pattern inspection apparatus.
FIG. 3 is a schematic diagram showing characteristics of the processing process of FIG. 2;
FIG. 4 is a flowchart showing an example of a process for extracting a noise region in a proposed pattern inspection apparatus;
FIG. 5 is a schematic diagram showing characteristics of the processing process of FIG. 4;
FIG. 6 is a flowchart illustrating an example of a process for removing a noise region.
FIG. 7 is a schematic diagram showing an example of a processing process for extracting a noise region in the proposed pattern inspection apparatus, and features of the processing process in FIG. 6;
8 is another schematic diagram showing characteristics of the processing process of FIG. 6;
FIG. 9 is a flowchart showing an example of a process for extracting a pattern defect in an example of a process for extracting a noise region in a proposed pattern inspection apparatus;
FIG. 10 is a schematic diagram showing an example of a defect form extracted by a proposed pattern inspection apparatus.
FIG. 11 is a flowchart showing a procedure of basic processing in an embodiment according to the present invention.
FIG. 12 is a schematic diagram showing the characteristics of the basic processing procedure in the present embodiment;
FIG. 13 is a flowchart showing an outline of a processing procedure from input of inspection object image data to pass / fail judgment of a periodic pattern in the present embodiment.
FIG. 14 is a flowchart showing details of a procedure of noise region extraction processing in the present embodiment.
FIG. 15 is a schematic diagram showing characteristics of noise area extraction processing;
FIG. 16 is a flowchart showing a processing procedure centering on noise area measurement and electrical noise exclusion in the present embodiment;
FIG. 17 is a schematic diagram showing an image of the characteristics of the processing procedure of FIG.
18 is another schematic diagram showing an image of the characteristics of the processing procedure of FIG.
FIG. 19 is a flowchart showing a procedure for removing a residual noise area;
FIG. 20 is a schematic diagram showing an example of a sample image after residual noise removal according to the present embodiment.
[Explanation of symbols]
1 ... Imaging device
2 ... Inspection device body
3. Display device
4 ... Inspection object (sample)
5 ... Inspection stage
6. Light diffusing plate
7. Illumination light source
8 ... Stabilized light source

Claims (7)

A noise region extraction process for creating a noise region extraction image data by extracting a noise region from the inspection target image data obtained by imaging the inspection target;
An electrical noise exclusion process of creating residual noise area image data consisting of noise areas of two or more pixels by excluding electrical noise generated only in one pixel unit from the noise area extraction image data;
Residual noise region-removed image data is created by replacing the inspection target image data corresponding to the residual noise region in the residual noise region image data excluding electrical noise with a representative value of pixels around the residual noise region. Noise area removal process,
An electrical noise removal process for removing the electrical noise that has been excluded from the residual noise region removal image data,
A pattern defect extraction process for extracting defects to be inspected based on the residual noise area removed image data from which electrical noise has been removed , and
The electrical noise exclusion process is a noise area measurement process for measuring the area of each noise area extracted in the noise area extraction image data,
An electrical noise determination process for determining electrical noise because the measured area is for one pixel, and a periodic pattern unevenness inspection method, comprising: In claim 1,
A periodic pattern unevenness inspection method, wherein the noise region is caused by fine foreign matter or the like adhering to an inspection target. In claim 1,
The noise region extraction process includes:
A smoothing process for creating a smoothed image data by performing a smoothing process on the inspection target image data;
An image subtraction process for creating subtracted image data by performing a process of subtracting a pixel value of a corresponding pixel of the inspection target image data from a pixel value of each pixel of the smoothed image data;
And a binarization process for generating binarized image data by performing a process of binarizing the subtracted image data. In claim 1 ,
The noise area extraction image data is binarized image data;
A dilation process is performed in which dilation processing is performed on the electrical noise excluded image data created by excluding the determined electrical noise and the binarized residual noise area image data is created. Inspection method for periodic patterns. In claim 1,
The residual noise area removing process includes:
A labeling process for assigning a number for identifying an isolated region corresponding to a noise region of two or more pixels in the residual noise region image data is performed to create a labeled image in which each isolated region is numbered. Labeling process,
A label N extraction process for creating label N image data by performing a process of leaving only a label N region which is an Nth isolated region of the labeled image data;
A second expansion process for expanding the label N image data to create label N expanded image data;
A second image subtraction process for creating a label N subtraction image data by performing a process of subtracting a pixel value of a corresponding pixel of the label N image data from a pixel value of each pixel of the label N expansion image data;
A representative value calculation step of extracting a pixel value of a pixel corresponding to the label N subtraction image data from the inspection target image data, and performing a process of calculating a representative value from the pixel value;
A pixel value replacement step of creating a label N area removed image data by performing a process of replacing a pixel value of a pixel corresponding to the label N image data from the inspection target image data with the representative value;
Repeating the above process for all label N regions, and performing a process of removing all label N regions from the inspection target image data to create the residual noise region removed image data. A method for inspecting unevenness of a periodic pattern as a feature.   The periodic pattern unevenness inspection method according to claim 1, wherein the representative value is one of an average value, a median value, and a mode value. A noise region extraction means for extracting a noise region from the inspection target image data obtained by imaging the inspection target and creating noise region extraction image data;
An electrical noise exclusion means for excluding electrical noise generated only in one pixel unit from the noise region extracted image data and creating residual noise region image data composed of noise regions of two or more pixels ;
Residual noise region-removed image data is created by replacing the inspection target image data corresponding to the residual noise region in the residual noise region image data excluding electrical noise with a representative value of pixels around the residual noise region. Noise area removing means;
Electrical noise removing means for removing the electrical noise that has been excluded from the residual noise area removed image data;
Pattern defect extraction means for extracting a defect to be inspected based on the residual noise area removed image data from which electrical noise has been removed , and
The electrical noise exclusion means measures noise area area measuring means for measuring the area of each noise area extracted in the noise area extraction image data;
An apparatus for inspecting unevenness of a periodic pattern, comprising: electrical noise determination means for determining electrical noise because the measured area is for one pixel .

Images