JP2006284471A - Pattern inspection method, pattern inspection device and pattern inspecting program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a pattern inspection method which enables the accurate measurement of the shape or outer shape dimension of a flaw even if there is illumination irregularity. <P>SOLUTION: This pattern inspection method of a semiconductor circuit or the like includes a first step S1 for applying differential processing to an image to be inspected to form a differential image, a second step S2 for comparing the differential image with the mask image preliminarily formed from a good product to form a difference image, a third step S3 for calculating a small region containing pixels having brightness exceeding a predetermined value in a case that the brightness value of the difference image exceeds the predetermined value, a fourth step S4 for calculating image feature quantity in the small region from the image to be inspected and a fifth step S5 for judging the quality of the small region from the value of the image feature quantity. The image processing in this pattern inspection method can be performed by software using a personal computer. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a technique for inspecting a defect in a circuit pattern of a semiconductor or a liquid crystal, for example, a foreign object, a scratch, a pattern defect, etc. by using an image processing technique.

  As a conventional method of pattern inspection while eliminating the influence of illumination unevenness called shading, the image of the good pattern and the image of the pattern to be inspected are differentiated to extract the pattern outline, There is one that detects a minute defect with a difference (see, for example, Patent Document 1).

JP 7-77495 A

  However, there are the following problems in applying the conventional inspection method.

  For example, when this conventional method is applied when a black foreign substance defect exists on the pattern, only the outline of the foreign substance can be detected.

  Further, it cannot be determined whether the foreign matter is a ring-shaped foreign matter having the outline or the foreign matter on the lump.

  Furthermore, in practice, the brightness of the portion corresponding to the contour of the foreign material is often thin. In this case, only the partial contour can be obtained, so that the external dimensions of the defect cannot be obtained correctly. There is a point.

  The present invention has been made to solve the various problems as described above, and to obtain a pattern inspection method or a pattern inspection apparatus capable of correctly detecting the shape of a defect stably and at high speed. The main purpose.

  A pattern inspection method according to the subject of the present invention is a method for inspecting an image to be inspected obtained using an illuminating means and an imaging means, wherein a first differential image is generated by differentially processing the image to be inspected. A second step of creating a differential image by comparing the differential image with a mask image previously created from a non-defective product, and when the luminance value of the differential image exceeds a predetermined value, the predetermined value A third step of calculating a small region including pixels having a luminance exceeding the threshold, a fourth step of calculating an image feature amount in the small region from the image to be inspected, and the small region from the value of the image feature amount And a fifth step of determining whether the quality is good or bad.

  Further, the pattern inspection apparatus according to the subject of the present invention compares the differential image with a mask image prepared in advance from a non-defective product, the first means for differentially processing the image to be inspected to create a differential image. A second means for creating a difference image, a third means for calculating a small area including pixels having a luminance exceeding the predetermined value when a luminance value of the difference image exceeds a predetermined value, and the inspection A fourth means for calculating an image feature amount in the small area from a target image, and a fifth means for determining the quality of the small area from the value of the image feature amount are provided.

  Hereinafter, various embodiments of the subject of the present invention will be described in detail along with the effects and advantages thereof with reference to the accompanying drawings.

  According to the subject matter of the present invention, it is possible to determine a defect to be inspected with high accuracy at high speed while eliminating the influence of uneven illumination such as shading.

  Further, the subject of the present invention is that a dead zone region is provided in advance in the mask image, and a configuration is made so as not to create a difference image in the dead zone region. Can be performed at higher speed.

  In addition, the subject of the present invention is that it is easy to observe, for example, electroerosion failure of a glass substrate in particular by configuring the illumination means by combining any one of a plurality of coaxial incident illuminations, oblique illuminations, transmission illuminations, or a combination thereof. Inspection accuracy can be further improved.

  The feature point of the pattern inspection method (image processing method) according to the present invention is that, first, a differential image obtained by performing differential processing on an image to be inspected to eliminate the influence of illumination unevenness is used to compare with a good image prepared in advance. A rough inspection of the defect location is performed by the difference detection processing, and then the inspection is performed by calculating the image feature amount from the original image, limited to the portion (small region) that is found to be defective and found by the rough inspection. The point to implement, in other words, is that the gray scale image processing is partially performed on the place found by the rough inspection and the fine inspection is performed in stages. Hereinafter, the contents of each specific embodiment will be described with reference to the drawings.

(Embodiment 1)
FIG. 1 is a flowchart showing a pattern inspection method according to the present embodiment. FIG. 2 is a block diagram schematically showing the configuration of a pattern inspection system for realizing the pattern inspection method according to the present embodiment. FIG. 3 is a flowchart in which an intermediate image in the middle of the inspection is shown in FIG.

  In FIG. 2, a pattern inspection system that realizes the pattern inspection method according to the present embodiment includes a CCD camera 1 (an example of an imaging unit) that images the inspection object 3, a coaxial incident illumination 2a that illuminates the inspection object 3, and an oblique direction. Illumination 2b and transmitted illumination 2c (an example of illumination means), a table 4 holding inspection object 3 and transmitted illumination 2c, and an X stage 5 and a Y stage 6 for moving the table 4 to predetermined positions are provided. Further, the pattern inspection system includes an image memory 7 (first storage means; for example, a memory inside a personal computer or an expansion board memory connected to the outside of the personal computer) and an image forming the core of the present pattern inspection method execution means. Processing means 8 (for example, the means 8 is composed of a personal computer or a hardware circuit such as a CPU or DSP and a combination of a hard disk and a hardware circuit). Here, when the image memory 7 of FIG. 2 is a memory inside a personal computer, the portion composed of the image processing means 8 and the image memory 7 constitutes a “pattern inspection device” composed of, for example, a personal computer. . On the other hand, when the image memory 7 is a memory of an external expansion board, the image processing means 8 itself comprising a personal computer, for example, constitutes a “pattern inspection device”. When both internal and external memories are used as the image memory 7, the portion composed of the image processing means 8 and the image memory 7 forms a “pattern inspection device”. In this example, the “pattern inspection apparatus” composed of a personal computer is controlled in software by a program for executing execution of a series of image processes schematically displayed in FIGS. 1 and 3.

  In this apparatus, an image captured by the CCD camera 1 is stored in a storage device including an image memory 7. The image of the inspection object 3 stored in the image memory 7 is subjected to image processing to be described later by the image processing means 8 that performs the pattern inspection method according to the present embodiment. As a result, the image processing means 8 A determination result 9 indicating the quality of the image is output.

  As an example of the inspection target, an object on which a circuit pattern is formed, such as a liquid crystal or an IC, can be cited. Instead of the CCD camera 1, a line sensor camera, a CMOS camera, a vidicon camera or the like having an equivalent function as an imaging unit may be used as the “imaging unit”.

  Next, the operation of the pattern inspection system for realizing the pattern inspection method according to the present embodiment will be described in detail with reference to FIG. First, the inspection object 3 is illuminated with the coaxial incident illumination 2 a, an image of the inspection object 3 is captured by the CCD camera 1, and the image data of the imaged inspection object 3 is input to the image memory 7. In the configuration of FIG. 2, input to the image memory 7 is input from the CCD camera 1. However, for example, an image data file that has been captured in advance may be read from the medium and input to the image memory 7. Note that, when an image of the inspection object 3 is picked up, a filter that allows only a specific wavelength region to pass between the inspection object 3 and the coaxial incident illumination 2a for the purpose of increasing the contrast of the image obtained from the inspection object 3. (Not shown) may be arranged.

  The image of the inspection object 3 input / stored in the image memory 7 is subjected to image processing in the image processing means 8 for performing the pattern inspection method described in detail below, and the image processing means 8 outputs a pass / fail judgment result 9. . Below, based on the flowchart of FIG.1 and FIG.3 which shows typically the operation | movement in the image processing means 8, this pattern inspection method implemented by the image processing means 8 is described.

  First, in step S1 (corresponding to the first step or the first means), the image processing means 8 (for example, CPU) performs a differentiation process on the image of the inspection object 3 input and stored in the image memory 7 of FIG. Thus, a differential image of the inspection object image is created. Step S1 is internally decomposed into step S11 for performing differentiation and step S12 for removing minute noise. In step S11, an image processing method for outputting a differential in the XY direction of the image is performed. For example, a Sobel filter or a Prewitt filter, which are known techniques, may be implemented as step S11, or an average of an output result of a filter that obtains a derivative in the X direction and an output result of a filter that obtains a derivative in the Y direction. It does not matter as the output of S11. Since the differential image output in step S11 often contains fine noise components, the noise components are removed in step S12. In step S12, for example, a minute noise is removed by a contraction / expansion process in which the 3 × 3 minimum value filter, which is a known technique, is applied a plurality of times and then the 3 × 3 maximum value filter is applied the same time. As other known techniques that can be used in step S12, a 3 × 3 median filter that removes sesame salt noise or an average value filter may be applied. By performing these steps S11 and S12, a differential image of the captured image of the inspection object 3 is obtained. The differential image is an image in which the edge portion of the inspection object 3 (the portion with a large luminance difference) is emphasized. FIG. 3 shows a differential image of the inspection target image. In the differential image, the edge portion is originally light and the background portion is dark. On the other hand, in the illustration of FIG. 3, for the sake of easy understanding, a diagram in which the light and dark are reversed (the edge portion is dark). The background part is bright. In this differential image example, the pattern of the inspection target image appears as an edge, and at the same time, the periphery of a linear defect and a circular defect also appears as an edge.

  Here, creation of a mask image will be described. Creation of the mask image is performed in step S0 of FIGS. The operation in step S0 may be performed simultaneously with the main pattern inspection, or may be performed off-line prior to the execution of the main inspection. Hereinafter, creation of a mask image will be described as a method to be performed offline.

  Step S0 includes steps S01 to S04. First, prior to step S01, an image of a non-defective target object for comparison with the inspection target 3 is input and stored in the image memory 7. Thereafter, in step S01, differentiation processing similar to that in step S11 is performed. In the next step S02, noise removal processing similar to that in step S12 is performed to obtain a differential image of a non-defective target object. In step S03, an image obtained by expanding the differential image of the non-defective target object is obtained by applying the aforementioned 3 × 3 maximum value filter or the like. In the present invention, the image after the expansion process is referred to as a “mask image”. The expansion process here is applied in order to eliminate the influence of the alignment error and the pattern error between the inspection object and the non-defective product when the alignment process and the difference process are executed later in step S2. FIG. 3 shows an example of the mask image obtained in step S03. In this mask image, since the expansion process is performed, the edge portion of the pattern is a thicker line than the edge portion obtained in the normal differential image. For convenience, the mask image in FIG. 3 is also shown with the light and dark reversed. In step S04, the obtained mask image and non-defective image are stored in the image database D1. Here, the image database D1 constitutes second storage means composed of, for example, a hard disk in a personal computer.

  In the above description, the procedure for generating a mask image from a quality image that was actually captured was described for the creation of a mask image. However, for the creation of a mask image, an image of an edge portion is generated based on pattern design data. And you may employ | adopt the method of obtaining by expanding.

  In the above description, a method of creating a mask image offline is described. However, when step S0 is performed simultaneously with the inspection, the following may be performed. That is, two cameras are prepared, one camera images the inspection object 3, the other camera images the non-defective product, and the CPU executes step S1 for the inspection object 3 and performs the step for the non-defective product. S0 is executed in parallel. If these steps S0 and S1 are executed at the same time, the following step S2 and subsequent steps can be executed without delay.

  Further, when the inspection object continues to flow, the previous inspection is regarded as the above-mentioned non-defective product, and the image of the previous inspection is processed in step S0 and currently inspected. A method may be employed in which the target image is processed in step S1, and then step S2 and subsequent steps are performed. Assuming that defects do not continuously occur at the same location, defective products can also be detected by this method.

  Step S2 (corresponding to the second step or the second means) is composed of three steps S21 to S23 as shown in FIG. 1, and the CPU and the like are the differential image of the inspection object 3 created in step S1. A difference image is created from the comparison with the non-defective mask image created in step S0. In the first step S21, the CPU or the like selects a mask image corresponding to the product type from the mask images created in step S0 and stored in the image database D1 according to the product type of the inspection object 3. select. Next, in step S22, the CPU or the like performs alignment between the selected mask image and the differential image of the inspection object 3. The reason why this alignment is necessary is that there is usually a slight misalignment between the differential image of the inspection object 3 and the non-defective image, and a correct image cannot be obtained even if the difference processing is performed as it is. . Here, as an example of the alignment method, alignment is performed by measuring X and Y positions of characteristic portions common to the differential image and the mask image. Taking FIG. 3 as an example, in the differential image and mask image of FIG. 3, the center position of the rectangular pattern in the upper left corner that is shown together may be measured by image processing. As an alignment method, for example, the edge of the differential image may be used to obtain the center position of the above-described square pattern. That is, the CPU or the like obtains the edge position of each side of the quadrangle, performs linear regression to extract the four sides, obtains four vertices that are the intersections of the four sides, and uses the average value of the four vertices as the center of the pattern. To do. In addition to the differential image, an original image obtained at the time of image input may be used. If the pattern matching method using normalized cross-correlation that is a known technique is used, the position of a rectangular pattern taught in advance is detected from both the original image to be inspected and the original image of the non-defective image. Can be used for alignment.

  In this way, if the center position (Xd, Yd) of the quadrangular pattern in the differential image of the inspection object 3 and the center position (Xm, Ym) of the quadrangular pattern in the non-defective mask image can be calculated, the CPU, etc. The positional deviation amounts (Xd−Xm, Yd−Ym) can also be calculated. If this positional deviation amount is corrected and the difference described below is taken, the edge portion of the pattern common to both images can be removed.

  In step S23, the CPU or the like performs difference processing. As the difference processing here, various methods can be considered as described below.

  After the alignment correction, if the luminance of one arbitrary point of the differential image is defined as Gd, the luminance of one point on the corresponding mask image is defined as Gm, and the luminance value of one point on the corresponding differential image is defined as Gs, When a simple difference is applied, the luminance value Gs is

Given in.

  On the other hand, in the case of a method of removing all the portions of the mask image whose luminance is a predetermined value (T) or more from the differential image, the luminance value Gs is:

Given in.

  Further, when the transmitted illumination 2c is used, the difference may be taken after logarithmic conversion. In this case, the luminance value Gs is

Given in. This is because the image obtained using transmitted light reflects the light absorption rate of the object, and logarithmic transformation is more appropriate when obtaining this difference image. Based on the physical basis to say.

  Here, the effect of the expansion process in step S03 will be described. When the difference process of step S23 is executed without the expansion process of step S03, the edge portion of the pattern is completely removed due to the difference between the minute positional deviation remaining after the alignment correction and the pattern due to the individual difference of the target. Cannot be performed and a noise component is generated. In particular, since the edge is a part where the luminance change is large in the first place, the magnitude of the noise component due to the difference between them may not be negligible. Therefore, if the expansion processing described above is executed in step S03, these noises can be completely eliminated.

  Next, the effect of using the differential image created in step S1 in creating the difference image in step S2 will be described. In the case of illuminating a wide range, uneven illumination is unavoidable. For this reason, the brightness of the original image to be inspected varies depending on the location (shading). Similarly, since the brightness of a good original image used for creating a difference image also varies, when the difference between these original images is calculated, an illumination unevenness is emphasized rather than a defect. Of course, this problem occurs when the illumination when the non-defective product is imaged differs from the illumination at the time of inspection. However, even if the illumination is exactly the same, the misalignment between the non-defective product and the inspection object 3 is corrected. Uneven lighting effects will appear at each stage. Moreover, the problem of uneven illumination also occurs due to deterioration of illumination.

  In such a case, the above-mentioned differential image is created and the difference between them is solved. This is because the illumination unevenness is a gradual change in brightness, and if the differential calculation, which is a local calculation, is performed, the influence of the illumination unevenness is eliminated. For the above reason, if a differential image is created by creating a differential image as described above, the influence of uneven illumination can be eliminated.

  The next third step (corresponding to the third means) is a step of calculating a small region including pixels having luminance exceeding the predetermined value when the luminance value of the difference image exceeds the predetermined value. That is, in step S3, the CPU or the like calculates the position of the defect from the difference image generated in step S2, and calculates a small area including the defect. First, in step S31, the CPU or the like binarizes the difference image with a predetermined threshold value to obtain a binary image. In step S32, the CPU or the like calculates binary feature values such as the area of the binary block (binary region), the position of the center of gravity, and the coordinates of the circumscribed quadrilateral from the binary image. In step S33, the CPU or the like determines whether or not a defect exists in the binary image by using the obtained binary feature amount with a predetermined determination threshold value. For example, when an area is used as a binary feature amount, the CPU or the like determines that a defect exists if the area is equal to or greater than a predetermined value. Alternatively, when using the coordinates of the circumscribed quadrilateral as the binary feature amount, the CPU determines that a defect exists if the length of the side is equal to or greater than a predetermined value. In step S <b> 34, the CPU or the like defines a range having a predetermined distance around the center of gravity of the binary block that is regarded as a defect as a “small region”. It should be noted that there may be a plurality of binary feature amounts calculated in step S32, and when the pass / fail determination is made in step S33, the determination may be made using a plurality of binary feature amounts.

  Here, an important point in the determination step S33 is that the determination threshold value at this stage is not so severe. As described above, in the case of a black foreign object or the like, the difference image can only be obtained from the contour part of the foreign object, so the binary feature value obtained in step S32 is not very large. For this reason, the determination threshold value in step S33 is set to be lower (relaxed). In the first place, since step S3 plays a role of rough search for defects, if a strict determination is made here, the true defect is overlooked.

  In step S4 (corresponding to the fourth step or the fourth means), the CPU or the like reads the inspection target image of the inspection target 3 stored in the image memory 7, and the image feature in the small region from the read inspection target image. Calculate the quantity. At this point, since the measurement range is limited to a small area, the influence of illumination unevenness is also small. In this step S4, for example, a difference image between the original image of the inspection object 3 (inspection object image) and the original image of the non-defective image is calculated and binarized with a predetermined threshold value. In this way, since the shape of the black foreign substance appears clearly in addition to its outline, the CPU or the like calculates the area or the length of the side of the circumscribed quadrilateral as the image feature amount.

  In step S5 (corresponding to the fifth step or the fifth means), the CPU or the like compares the above-described image feature amount with a determination threshold value, and determines the quality of the small region. At this stage, unlike step S3, since the shape of the defect is clear, the determination threshold value is set to an appropriate value for the original inspection.

  In step S4, as a method other than the above-described measurement method, for example, the original image is binarized by a known automatic binarization method or a fixed binarization method, and the area or circularity (roundness ratio) of the binary block is obtained. Alternatively, a circumscribed quadrilateral of a binary block may be obtained, and the length of the side may be obtained as an image feature amount. Further, the luminance average in a small area of the original image may be obtained without binarization, or the luminance dispersion value may be obtained as an image feature amount. In these cases, the method is effective particularly when the edge of the pattern is not included in the small area. Of course, in this case, since a good original image is unnecessary, the calculation speed is also increased. It should be noted that there may be a plurality of image feature amounts calculated in step S4, and even when the quality is determined in step S5, determination may be made using a plurality of image feature amounts.

  Since the pattern inspection method is constructed as described above, it is possible to accurately obtain the outer shape of foreign matter while eliminating the effects of uneven illumination due to shading, illumination deterioration, etc. High inspection is possible. Also, instead of searching for defects from the beginning, the difference image of the differential image is first used to detect a rough defect portion and calculate a highly reliable feature amount only for a small region that is a defect candidate. As a result, high-speed inspection is possible as a whole.

(Embodiment 2)
FIG. 4 is a flowchart showing the pattern inspection method according to the present embodiment. Note that FIG. 2 is used in this embodiment. In FIG. 4, the same step name is assigned to the portion that performs the same processing as FIG. 1 cited in the first embodiment. Since the basic flow is substantially the same as that of FIG. 1 of the first embodiment, common portions are omitted, and only different points will be described below.

  The difference between the inspection method illustrated in FIG. 4 and the inspection method illustrated in FIG. 1 is that step S13 is added in step S1 of creating a differential image. That is, immediately after the differentiation process is performed on the inspection target image, in step S13, the CPU or the like performs a binarization process with a predetermined threshold value. By doing in this way, the effect which emphasizes the edge emphasized by a differential image arises. In subsequent step S12, since the image to be handled is a binary image, in addition to the noise removal method described in the first embodiment, the binary feature amount calculation is performed to calculate the area of the binary block, and then a predetermined value is obtained. It is also possible to remove a binary block having an area smaller than that as noise. This noise removal method has a stronger noise removal effect than the contraction / expansion processing, the median filter, and the average value filter. Further, the addition of step S05 when creating the mask image is also for the same reason as described above.

  With the addition of step S13 and step S05, the differential image has become a binary image. Therefore, with respect to the difference image calculation method in step S23, in addition to the method described in the first embodiment, the difference is also obtained by performing XOR (eXclusive OR), which is a logical calculation, on the differential image and the mask image. Images can be calculated. In this case, there is an advantage that the calculation can be performed even faster than the simple difference calculation.

  In step S3 of the second embodiment, since the difference image is already a binary image at the input stage, step S31 for performing the binarization process described in the first embodiment is not necessary.

  When the pattern inspection method is configured as illustrated in FIG. 4, the effect obtained in the first embodiment described above can be obtained, and the binary processing is performed at an early stage to increase the contrast of the differential image and the mask image. The inspection accuracy can be improved by this and the difference image calculation using XOR can be performed at higher speed.

(Embodiment 3)
In the first embodiment and the second embodiment described above, when calculating the difference image between the differential image and the mask image, weighting is not performed on a specific portion of the image, and the difference is simply calculated and the inspection is performed. Yes.

  However, in actual inspection, for example, if the pattern is a circuit pattern, scratches and foreign matter on the circuit pattern are defects that should be detected in the inspection, but even if there are scratches or foreign matters in the portion outside the pattern, it is a problem. It may not be possible. This point will be described with reference to FIG. 3. If the white portion of the inspection target image in FIG. 3 is a circuit pattern portion, foreign matter and scratches on the white circuit pattern are problematic, but the portion of the inspection target image in FIG. This means that foreign matter and scratches in the black part, which is the background, will not be a problem. In such a case, in step S2 for calculating the difference image, it is not necessary to make a difference for the entire image, and it is only necessary to calculate the difference image only for the pattern portion necessary for the inspection.

  In this embodiment, when creating a mask image, a non-pattern portion is stored in advance in the image database D1 as a dead zone region. An appropriate binarization threshold value applied to a good original image and a good original image can be substituted for the data in the dead zone area. Thereafter, when a difference image is created in step S2 at the time of inspection, a difference image is created only at a pattern portion that is not a dead zone, and no difference image is created in the dead zone.

  According to the present embodiment, it is possible to greatly reduce the processing of steps S3 to S5 subsequent to step S2, and it is possible to realize further increase in the inspection time.

(Embodiment 4)
In the first embodiment, the second embodiment, and the third embodiment described above, the coaxial incident illumination 2a shown in FIG. 2 is mainly described as the illumination unit. Here, the coaxial epi-illumination 2a is the same as the illumination when a human visually inspects with a microscope, and the obtained image is almost the same as when the human observes, so that the level of the inspection standard with the human can be easily adjusted. It becomes.

  On the other hand, the coaxial epi-illumination 2a may not provide the image contrast necessary for image processing. For example, there is a case where a clear contrast cannot always be obtained with coaxial illumination with respect to the electroerosion failure of the transparent electrode of the liquid crystal.

  At this time, the oblique illumination 2b and the transmitted illumination 2c of FIG. 2 may be used. When the oblique illumination 2b and / or the transmitted illumination 2c is used, a translucent electric corrosion failure that is not completely metallized can be clearly observed, so that the illumination is effective. Further, in the inspection in which the inspection object is transparent like a liquid crystal glass substrate and the defect is opaque such as a foreign substance, it is particularly suitable to use the transmitted illumination 2c that can detect the defect with good contrast. Even a semiconductor that is seemingly opaque, such as a semiconductor, can be inspected by using infrared illumination as transmitted illumination because the semiconductor transmits infrared.

  In this way, it is desirable to apply the illumination that can detect each defect stably by appropriately using the coaxial incident illumination 2a, the oblique illumination 2b, and the transmitted illumination 2c according to the type of defect to be detected. Of course, when there are a plurality of defects to be detected, a plurality of these illuminations 2a, 2b, 2c may be used in an appropriate combination.

(Appendix)
While the embodiments of the present invention have been disclosed and described in detail above, the above description exemplifies aspects to which the present invention can be applied, and the present invention is not limited thereto. In other words, various modifications and variations to the described aspects can be considered without departing from the scope of the present invention.

  The pattern inspection apparatus or method according to the present invention is preferably applied to a manufacturing apparatus or manufacturing process for a device having a circuit pattern such as a semiconductor or a liquid crystal.

It is a flowchart which shows the flow of the pattern inspection method which concerns on Embodiment 1 of this invention. It is a block diagram which shows typically the structural example of the pattern inspection system which concerns on Embodiment 1 of this invention. It is the flowchart which added the example of the image under image processing to the flowchart of the pattern inspection method by Embodiment 1 of this invention. It is a flowchart which shows the flow of the pattern inspection method which concerns on Embodiment 2 of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 CCD camera, 2a Coaxial incident illumination, 2b Oblique illumination, 2c Transmitted illumination, 3 Inspection object, 4 Table, 5 X stage, 6 Y stage, 7 Image memory, 8 Image processing means, 9 Determination result.

Claims (6)

In a method for inspecting an image to be inspected obtained using an illuminating means and an imaging means,
A first step of differentially processing the image to be inspected to create a differential image;
A second step of creating a differential image by comparing the differential image with a mask image previously created from a non-defective product;
A third step of calculating a small region including pixels having a luminance exceeding the predetermined value when the luminance value of the difference image exceeds a predetermined value;
A fourth step of calculating an image feature amount in the small region from the image to be inspected;
A fifth step of determining pass / fail of the small area from the value of the image feature amount,
Pattern inspection method. The pattern inspection method according to claim 1,
In the second step, a dead zone is provided in advance in the mask image, and the differential image is not created in the dead zone.
Pattern inspection method. The pattern inspection method according to claim 1 or 2,
At least one of coaxial epi-illumination, oblique illumination and transmitted illumination is used for the illumination means,
Pattern inspection method. An apparatus for inspecting an image to be inspected obtained using an illuminating means and an imaging means,
First means for differentially processing the image to be inspected to create a differential image;
A second means for creating a differential image by comparing the differential image with a mask image previously created from a non-defective product;
A third means for calculating a small area including pixels having luminance exceeding the predetermined value when the luminance value of the difference image exceeds a predetermined value;
A fourth means for calculating an image feature amount in the small area from the image to be inspected;
And a fifth means for determining the quality of the small area from the value of the image feature amount,
Pattern inspection device. The pattern inspection apparatus according to claim 4,
The second means includes
In the dead zone area provided in advance in the mask image, the difference image is not created,
Pattern inspection device. 3. A method for instructing an electronic computer to perform each of the image processing steps from the first step to the fifth step in the pattern inspection method according to claim 1 or 2 to cause the electronic computer to sequentially execute the image processing steps. of,
Program for pattern inspection.

Images