JP2010071892A - Pattern inspection method and pattern inspection device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve a following problem with defect inspection on a circuit pattern using image processing: there is a risk that an error equating with one pixel occurs owing to displacement not larger than one pixel when an image is turned into a binarized image in image capturing and the error acts as a detection error when detecting whether any defect exists or not by comparison with standard data acquired by expanding/contracting image data serving as a reference. <P>SOLUTION: A pattern inspection device includes: a displacement measuring part 20 measuring, on the order of one pixel size or less, a positional error between image data on an inspected body and reference image data by using a specific domain of each of the respective groups of image data; and a rectifying part for rectifying the positioning of the yet-to-be-binarized image data on the order of sub-pixels. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a circuit pattern inspection method and apparatus for automatically inspecting printed circuit board circuit patterns and LSI circuit pattern defects by image processing.

  The demand for a circuit pattern in which a conductor pattern is formed on an insulating member is rapidly increasing with the spread of electronic devices such as mobile phones. Since the circuit pattern affects the performance and durability of an electronic device using the circuit pattern, the circuit pattern is not only electrically connected as designed, but also requires a predetermined dimensional accuracy.

  Therefore, all the circuit patterns are inspected after being manufactured, and it is confirmed that a predetermined specification is satisfied. By the way, since the types and number of circuit patterns have become enormous, it is necessary to complete the inspection in as short a time as possible.

  Conventionally, several methods have been proposed for this inspection. A method called a line width sub-pixel measurement method is a method in which the line width is measured for every fixed length, the line width between them is determined from several measurement points before and after, and the quality with respect to the reference width is judged. This method can be inspected with a resolution of about 1/10 of the pattern pitch.

  The image comparison method is a method of obtaining a difference image between the binarized imaging data and the non-defective product data and obtaining a difference portion having a certain area or more as a defect. Since this method binarizes data, the calculation speed can be increased.

In this image comparison method, it is necessary to acquire image data of an inspection object. For this purpose, the inspection apparatus has a two-dimensional imaging apparatus such as a CCD image sensor or an ITV camera (Patent Document 1). This imaging apparatus has an illumination unit that applies light from the camera side or from the back side of the inspection object in order to obtain a clear image of the inspection object.
JP 63-19541 A

  In the image comparison method, each pixel of the image data of the inspection object and the non-defective product is binarized and compared. Pixel binarization can shorten the time required for comparison by reducing the amount of information of each image data.

  However, pixel binarization may result in a quantization error in the vicinity of the threshold for binarization, in which the value of a pixel that should originally be 1 becomes zero or vice versa. is there. In particular, when comparing the image data of the inspection object and the non-defective product on a pixel basis, this error occurs due to a positional shift when the image data is used.

  This deviation will be described with reference to FIG. In FIG. 11A, the vertical axis represents luminance, and the horizontal axis represents the pixel position. Reference numeral 500 represents a partition of pixels, and 502 represents an actual image. In the image comparison method, since the image is converted into digital data, the luminance of the image is digitized for each pixel partition. For example, it is digitized by 8 bits or 10 bits (hereinafter, this numerical value is referred to as “luminance value” or “pixel value”). Reference numeral 504 represents a numerical point that is digitized for each pixel. Pixel binarization determines 1 or zero by comparing this number with a threshold.

  FIG. 11B is a diagram showing the state of binarization. Although the vertical axis represents the luminance value, in binarization, each pixel is given a value of 1 or zero, and may be considered to represent “1” or “0”. Reference numeral 506 is a threshold value. For example, since the pixel having the luminance value 508 is larger than the threshold value 506, the binarized data is 1. As a result, the binary image data has the code 510.

  Here, as shown in FIG. 11C, consider a case where a positional shift of one pixel or less occurs between the inspection object and the non-defective image data. As a result, the luminance value 512 in FIG. 11A becomes the luminance value denoted by reference numeral 514 in FIG. When this is binarized, the result is as shown in FIG. 11D, and the number of pixels having a pixel value of 1 is reduced by one pixel as compared with FIG. That is, the luminance value becomes smaller or larger than the threshold value at the boundary between the bright area and the dark area of the image.

  As a result, an error of plus or minus one pixel occurs between the binarized image data of the inspection object and the non-defective product. When inspecting for the presence or absence of a defect, the standard data obtained by expanding or reducing the binarized image data of the non-defective product is compared with the image data of the inspection object. Since expansion or reduction is performed at a predetermined magnification, the error of one pixel may be an error of several pixels, and there is a problem that the detection accuracy of defects is lowered.

  The pattern inspection method of the present invention has been conceived to solve the above problems. That is, before binarization, the positional deviation between the inspection object and the non-defective image data is measured with an accuracy of 1 pixel or less, and the positional relationship between the inspection object and the non-defective image data is corrected based on the deviation. . Then, binarize, expand and contract, and compare.

  In the present invention, the amount of deviation between the image data to be inspected and the non-defective image data is measured before binarization, and correction is performed based on this amount of deviation. For this reason, it is possible to suppress a decrease in defect detection accuracy due to a positional shift of one pixel or less and improve the accuracy of defect detection.

  Embodiments of the present invention will be described below, but the present invention is not limited to these embodiments, and can be modified and changed within the scope of the gist of the present invention.

(Embodiment 1)
FIG. 1 shows a configuration diagram of a pattern inspection apparatus 1 of the present embodiment. The pattern inspection apparatus 1 according to the present invention captures an object 90 and outputs inspection image data Vd, reference data Vsd that is non-defective image data, and inspection image data Vd are input, and the amount of deviation A deviation measurement unit 20 that measures ΔD, a correction unit 25 that corrects the inspection image data Vd based on the deviation amount ΔD, and outputs the corrected image data Ed. The binarization unit 30 that outputs the binary corrected image data BEd, the expansion / reduction unit 40 that receives the reference data Vsd and outputs the binarized expansion data Bsdx and the reduction data Bsds, and stores the respective data Storage unit 80, binary corrected image data BEd, expansion data and reduced data are compared, and respective differences Cdx and Cds are output. A determination unit 70 that determines the presence or absence of a defect and outputs the result is included. An output terminal 88 may also be included.

  The inspection object 90 is mainly a printed circuit board on which a circuit pattern is formed. The image capturing unit 10 includes a camera 11 and an illumination 12. The camera 11 is composed of an imaging element such as a CCD and a lens. As the image sensor, a linear sensor or a two-dimensional sensor can be used. In particular, in the present invention, it is preferable to use a two-dimensional area sensor. This is because the inspection image data Vd can be produced at high speed.

  The inspection object 90 is illuminated by the illumination 12. The camera 11 images the inspection object 90 and outputs inspection image data Vd. The inspection image data Vd is a set of digital data for one screen having a gradation value for each pixel.

  The components after the image capturing unit 10 can be configured in hardware, but most of them are realized by an arithmetic unit such as an MPU (Micro Processor Unit) and a memory, and by processing software. Therefore, in the present specification, these portions will be described using the units of processing executed by software in the computer as components. In addition, although these are demonstrated as one control apparatus 85, the control apparatus 85 may be comprised from the some calculator.

  An outline of the configuration and flow of the control device 85 will be described with reference to FIGS. 1 and 2, and each component will be described in detail later. When the control device 85 starts (S100), an end determination (S102) is first performed. The end determination may be a compulsory stop command due to the presence or absence of an inspection object or an artificial interruption. In the case of completion (Y branch in S102), the process proceeds to step S116 and the process is stopped (S116).

  When the process is continued (N branch in S102), the inspection image data Vd is input to the control device 85 (S104). The reference data Vsd is recorded in advance in the deviation measuring unit 20, and the amount of positional deviation between the inspection image data Vd and the reference data Vsd of the inspection object is measured (S106). In this case, the amount of deviation is measured in units of one pixel or less (hereinafter referred to as “sub-pixel”). The deviation amount is output as ΔD.

  The inspection image data Vd is also input to the correction unit 25. The correction unit 25 receives the shift amount ΔD and the inspection image data Vd, and corrects the position of the image data based on the shift amount ΔD (S108). The position-corrected image data is output as corrected image data Ed.

  The corrected image data Ed is input to the binarizing unit 30 and output as binary corrected image data BEd (S110). The binary corrected image data BEd is sent to the calculation unit 60.

  On the other hand, the reference image data Vsd is input to the expansion / reduction unit 40 in advance. The expansion / reduction unit 40 binarizes the reference image data Vsd, and then generates and outputs the expansion standard data Bsdx subjected to the expansion process and the reduced standard data Bsds subjected to the reduction process. These standard data are stored in the storage unit 80 of the pattern inspection apparatus.

  The arithmetic unit 60 receives the binary corrected image data BEd and the standard data from the storage unit 80, and outputs the differences Cdx and Cds between the standard data and the binary corrected image data (S112). The determination unit 70 outputs a result Con indicating the presence / absence of a defect based on the differences Cdx and Cds (S114). Thereafter, the process returns to S102 again to make an end determination.

  Next, a detailed processing flow of each unit in FIG. 1 will be described. The deviation measuring unit 20 receives the reference image data Vsd and the inspection image data Vd obtained by photographing the inspection object with the image photographing unit 10. Here, the inspection image data Vd and the reference image data Vsd may be slightly displaced from each other when the inspection image data Vd is acquired. As described above, when such a shift occurs, there is a possibility that the defect detection accuracy may be lowered.

In order to solve this problem, the positional deviation between the reference image data Vsd and the inspection image data Vd must be corrected. However, if these data are binarized, position correction in units of subpixels cannot be performed. In the present invention, a deviation amount is measured between the reference image data Vsd before binarization and the inspection image data Vd, and position correction is performed on the image data before binarization based on the deviation amount.

  Referring to FIG. 3, reference numeral 500 is a pixel delimiter, and reference numeral 502 is an image. Assume that the reference image data Vsd is divided into pixels as shown in FIG. If the inspection image data Vd of the inspection object is divided as shown in FIG. 3B, the inspection image data Vd becomes one pixel smaller than the standard data Vsd when binarized as shown in FIG. As shown. This correction cannot be corrected after binarization (FIG. 3D).

  Therefore, the displacement 520 in units of subpixels between FIGS. 3A and 3C is measured, and the inspection image data Vd before binarization is positioned in units of subpixels as shown in FIG. Correct the deviation.

  FIG. 4 shows a processing flow of the deviation measuring unit. This process is a process of deviation measurement (step S106) in FIG. The outline of the processing is as follows. Using specific areas of the reference image data Vsd and the inspection image data Vd, the degree of coincidence is measured while shifting each data by a predetermined distance determined in advance in units of pixels. Then, using the luminance value or all measured values in the vicinity of the amount of deviation with the highest degree of coincidence, the amount of deviation when the degree of coincidence becomes the highest is obtained in the order of subpixels.

  The image data is data in which pixels are two-dimensionally arranged. Accordingly, each pixel can be represented by the vertical and horizontal coordinates and the luminance value of the pixel. For example, when the pixel (x, y) has a luminance value b when the lower left corner of the screen is the origin, the horizontal direction is the x direction, and the vertical direction is the y direction, it may be expressed as (x, y, b) or the like. it can.

  Further, either the reference image data Vsd or the inspection image data Vd may be moved, but here, the description will be made assuming that the inspection image data Vd is moved.

  Here, the specific area may be an arbitrary area of the image data, but it is preferable to select an area including a relatively simple pattern provided at the side, corner, or central portion of the product. This is because there are few manufacturing difficulties and there is little possibility of occurrence of defects, which is suitable as a position for alignment. If the specific area is widened, the accuracy of alignment increases, but the amount of calculation increases. Therefore, it may be appropriately determined according to the processing time required for the inspection.

  In the following description, the specific area of the inspection image data Vd is R, the pixel of the coordinates (x, y) in the specific area R is R (x, y), the luminance value of the pixel is Rxy, and the specific area of the reference image data Vsd Is represented by Rsd, a pixel at coordinates (x, y) in the specific region Rsd is represented by Rsd (x, y), and a luminance value of the pixel is represented by Rsdxy. That is, the pixel in the specific region R is represented as (x, y, Rxy), and the pixel in the specific region Rsd is represented as (x, y, Rsdxy). For the sake of simplicity, the specific area will be described as a rectangular area.

  The specific areas R and Rsd are represented by the coordinates (a1, b1), (a2, b1), (a1, b2), (a2, b2) of the four corners of the rectangle on the inspection image data Vd and the reference image data Vsd, respectively. .

  The flow of FIG. 4 will be specifically described. When the process starts in the main step S106 (S200), an end determination is made (S202). Here, the end determination is whether or not the following processing has been performed by moving the region R or Rsd in units of pixels by a predetermined distance. Specifically, it is moved from x to x + me in the x direction and moved from y to y + ne in the y direction. When the following process is completed while moving all the distances, the process proceeds to a shift calculation process. The amount of movement in the x direction is represented by m. Note that me and ne are the maximum values of movement distances in the x and y directions.

  There is no particular limitation on how to move the area, and the following processing does not have to be performed for all coordinates in the planned movement range. That is, there may be skipped coordinates. Here, a description will be given of a procedure of sequentially moving in the x direction, and after moving in the x direction, moving one pixel in the y direction and moving in the x direction again. Accordingly, the end determination is made based on whether or not the movement amount n in the y direction has become ne.

  When continuing (N branch of step S202), the y direction (n) is incremented by one pixel, and the x direction (m) is set to an initial value. Next, the end determination in the x direction is performed. That is, the determination is made based on whether m becomes me. When the movement in the x direction ends, the process jumps to the step and receives the end determination in the y direction.

  If the movement in the x direction is not completed, the region R is moved in the x direction by m pixels. Specifically, the x coordinate of the pixel on the region R is rewritten to x + m. That is, when the coordinate of a pixel having a certain luminance value P is (x, y), the coordinate of the luminance value P is a new coordinate (x + m, y). Then, the degree of coincidence of the luminance values of the pixels at the same coordinates on the region R and the region Rsd is obtained. Specifically, for all u and v in the region, the degree of coincidence of luminance values is obtained between R (u, v) and Rsd (u, v).

  As long as the matching pixel has the same luminance value, any method may be used as long as the matching level is the highest. For example, a normalized correlation method can be suitably used. In the normalized correlation method, the evaluation value r approaches 1.0 if the degree of coincidence is high, and approaches -1.0 if the degree of coincidence is low. Specifically, the evaluation value r is obtained by the equation (1).

・・・・・（１）

(1)

  Here, i and j represent coordinates in the specific area. In the following description, the degree of coincidence will be described using r by the normalized correlation method.

  When the degree of coincidence is obtained, the movement distance (m, n) of the inspection image data Vd and the evaluation value r are recorded in association with each other (S214). Then, the x direction is incremented (S216), and the process returns to step S208.

  Thus, when the region R is moved, the pixel R (x, y) on the region R is moved on the inspection image data Vd. A normalized correlation value between different pixels of the inspection image data Vd and the reference image data Vsd is calculated for each movement.

  Returning to step S202, when the movement distance n in the y direction becomes ne (Y branch of S202), the amount of deviation is calculated using the calculated normalized correlation value r (S218). Since the image data is two-dimensional data, the deviation amount has values in both the x and y directions. The amount of deviation is calculated as follows.

  When the processing up to step S218 is completed, data in which the movement distance in the x and y directions and the normalized correlation value r corresponding thereto are associated remains. If the movement distance that maximizes r is obtained using this data, the movement distance is the amount of deviation. Specifically, considering the three-dimensional data (m, n, r) of the movement distance and the normalized correlation value, the x coordinate and the y coordinate for the vertex of the curved surface that complements the normalized correlation value r are obtained. As a complementing method, for example, a gradient-based method, an experimental design method, a response surface approximation method, or the like can be used. Alternatively, a method of complementing with splines in the x or y direction may be used. This is because the amount of calculation can be reduced.

  The amount of deviation is obtained on the order of one pixel or less (subpixel order). That is, the amount of deviation obtained by complementation is obtained on the order of at least one pixel. Therefore, the deviation amount ΔD is expressed as (m + Δm, n + Δn). Here, Δm and Δn are distances of one pixel or less. When the amount of deviation is output, the process returns to the main (S220).

  Returning to FIG. 2, the image data Vd is then moved by the amount of deviation (S108). In the present embodiment, this step corresponds to a step of performing correction based on the amount of deviation. In this process, the entire inspection image data Vd is moved by the amount of deviation (m, n) in pixel units, and the luminance value is interpolated in the sub-pixel order (Δm, Δn) to create a new luminance point.

  FIG. 5 shows the flow of this process. When this process starts (S230), the deviation amount ΔD is read (S232). The amount of deviation is represented by (m + Δm, n + Δn). Next, (m, n) is added to the coordinate values of all the pixels of the inspection image data Vd (S234). That is, the pixel represented by (x, y, b) on the inspection image data Vd is represented as (x + m, y + n, b). Note that although there are insufficient pixels or protruding pixels in the vicinity of the boundary of the inspection image data Vd, this excess and deficiency can be absorbed by taking a wide image in advance.

  Next, the luminance value of the point (Δm, Δn) is obtained for all the pixels in the image data Vd (S236). The method of obtaining the luminance value is not particularly limited, but a biquadratic interpolation method or a bicubic interpolation method can be preferably used. The biquadratic interpolation method (bilinear interpolation method) is a method for obtaining the luminance value of the point (Δm, Δn) based on the luminance values of the four points surrounding this point. The bi-cubic interpolation method (bicubic interpolation method) is a method for obtaining based on the luminance values of 16 points surrounding this point. It is calculated | required by (2) Formula and (3) Formula, respectively.

・・・・・（２）

(2)

Here, Cb (Δm, Δn) represents the luminance value at the point (Δm, Δn). The four coordinates around the point (Δm, Δn) are four points (X, Y), (X + 1, Y), (X + 1, Y + 1), and (X, Y + 1).

・・・・・（３）

(3)

  Note that f11 to f44 are 16 points around the point (Δm, Δn). The left side of consecutive numbers represents the x-axis coordinate, and the right side represents the y-axis coordinate. The point (Δm, Δn) is a point existing in a quadrangle surrounded by f22, f32, f23, and f33.

  X1 to y4 have the following relationship.

・・・・・（４）

(4)

  H (t) is an approximation of the sinc function by a cubic polynomial, and the general formula is expressed by formula (5).

・・・・・（５）

(5)

  Here, t is a variable of the function h (), and is x1 to x4, y1 to y4.

Next, the point of (Δm, Δn) obtained by complementation is set as a new coordinate (x, y) (S238). That is, the pixel of the new coordinate is represented by (x, y, Cb). This is the corrected image data Ed. Here, Cb is a new luminance value obtained by complementation. Then, the process returns to the main (S240). The process of setting the point (Δm, Δn) as a new coordinate (x, y) is a step of interpolating the distance of the inspection image data Vd in units of subpixels.

  By this processing, it is possible to obtain corrected image data Ed obtained by moving the inspection image data Vd by (m + Δm, n + Δn).

  Returning to FIGS. 1 and 2, the modified image data Ed is binarized by the binarization unit. FIG. 6 shows a flow of binarization processing. When step S110 in FIG. 1 starts (S250), the corrected image data Ed is read (S252), and an end determination is made (S254). Here, the end determination is made based on whether or not the following processing has been performed on all the pixels of the corrected image data Ed. If completed (Y branch of S252), the process returns to the main routine (S264).

  Otherwise (N branch in S254), the parameter indicating the pixel is incremented (S256), and all the pixels are compared with a predetermined threshold Thb (S258). If Ed (x, y) is large (Y branch of S258), the luminance value of the binary corrected image data BEd (x, y) is set to the maximum value Imax, and otherwise (N branch of S258). The luminance value of the binary corrected image data BEd (x, y) is set to the minimum value Imin. And it returns to step S254 again. Thus, the data obtained by binarizing the compensation image data Ed can be obtained as the binary corrected image data BEd. Although Imax and Imin can be set arbitrarily, Imax may be 1 and Imin may be zero.

  Returning to FIG. 1, the pattern inspection apparatus of the present invention has an expansion / reduction unit that reads standard image data Vsd, binarizes it, and creates standard data that has undergone expansion / reduction processing. The dilation processing is to change the luminance value in the range of one pixel around the pixel having the maximum luminance value to the maximum value Imax in the binarized image data, so that the pixel having the maximum luminance value and the luminance value are changed. This is a process of manipulating the boundary of the pixel having the minimum value Imin so that the pixel having the maximum luminance value spreads. The reduction process is an operation opposite to the expansion process and is performed so that pixels having a luminance value of Imin are expanded. The standard data thus produced is recorded in the storage unit 80.

  The arithmetic unit 60 receives the binary corrected image data BEd from the binarizing unit 30 and reads standard data from the storage unit 80. Then, calculation is performed between the binary corrected image data BEd and the standard data. Here, since the binary corrected image data is aligned with the reference image data in units of sub-pixels before being binarized, the binary corrected image data that is binarized image data and the reference image data Even if the standard data created based on the binarized image data is directly compared with each other, there is almost no error in quantizing the pixels. 7 and 8 show the flow of this process.

  Referring to FIG. 7, when the difference calculation (S112) in FIG. 2 is started (S270), first, binary corrected image data BEd and expansion standard data Bsdx are read (S272). Then, an end determination (S274) is performed. Here, the end determination is whether or not the following calculation has been performed for all the pixels of both image data. If completed (Y branch of S274), the process proceeds to step S290, which is processing with the next reduced standard data (S282). When continuing (N branch of S274), x and y which are parameters indicating pixels are incremented (S276).

  Next, a difference between the binary corrected image data BEd and the expansion standard data Bsdx is obtained (S278). Since the expansion standard data has a pattern portion larger than the binary corrected image data obtained by correcting the inspection image data and binarized, if this operation is performed, the values of all pixels should be zero or positive. is there. However, if there is a pixel whose result is negative among the pixels of the inspection image data, the pixel is considered to be a pixel constituting a portion that protrudes from the planned pattern.

  Therefore, if the result is negative (Y branch of S278), the difference Cdx is incremented (S280). If the result is positive (N branch in S278), no processing is performed and the process returns to the end determination in step S274. In the above calculation, the magnitudes of the brightness values of the binary corrected image data BEd and the expanded standard data Bsdx may be compared. In step S278 of FIG. 7, the calculation for obtaining the difference is shown as a size comparison.

  As a result, the difference Cdx counts the number of pixels in the portion of the inspection image data Vd that protrudes from the range of the expansion standard data Bsdx.

  Referring to FIG. 8, when the difference calculation with the expansion standard data is completed, the difference calculation with the reduced standard data is performed next. This flow is processing in which the expansion standard data Bsdx in FIG. 7 is changed to the reduced standard data Bsds. The difference lies in the comparison of the brightness values of the binary corrected image data BEd and the reduced standard data Bsds in step S298. Here, when the luminance value of the pixel of the reduced standard data is larger than the luminance value of the pixel of the binary corrected image data, the difference Cds is incremented (S300).

  Since the reduced standard data is created as a pattern smaller than the inspection image data, the luminance value of the pixel of the reduced standard data is larger than the luminance value of the pixel of the binary corrected image data. It means that there is a smaller part. As a result, the difference Cds counts the number of pixels in the inspection image data Vd that are smaller than the reduced standard data Bsds.

  As for the process for obtaining the difference between the expansion standard data and the reduction standard data, if there is a margin in the processing time, a so-called labeling process is performed on the difference image, the area of each defective portion is obtained, and a threshold with this area is set. If the part beyond the boundary is regarded as a defect, a more accurate inspection is possible.

  Returning to FIG. 1, the differences Cdx and Cds are input from the calculation unit 60 to the determination unit 70. The determination unit 70 compares the differences Cdx and Cds with a predetermined threshold value, and outputs a conclusion Con regarding whether or not it is a defect. As a result, Con may include differences Cdx and Cds, or may include information for specifying an inspection target.

  As described above, in the present embodiment, the displacement amount between the positions of the inspection image data Vd and the reference image data Vsd is measured in units of subpixels, and the reference image data Vsd is shifted before the inspection image data Vd is binarized. Since the position is corrected based on the amount, and then binarized data is compared and calculated, pattern inspection with very little quantization error when converting to image data can be performed.

(Embodiment 2)
In the first embodiment, the shift amount is corrected before the inspection image data is binarized. However, in this processing method, it is necessary to correct the entire inspection image data before binarization based on the shift amount. Since the amount of deviation is measured for a specific area of image data, a large amount of calculation is not required. However, in order to correct the entire inspection image data, it is necessary to obtain a new luminance point in one pixel by interpolation after rewriting the coordinates of the pixel (corresponding to moving the pixel). Performing this calculation for all points in the inspection image data every time a pattern inspection is performed on an individual inspection object means that the inspection takes time.

  Therefore, in the present embodiment, by creating standard data based on the reference image data that has been moved in advance by the sub-pixel, and selecting and calculating the standard data corresponding to the sub-pixel of the deviation amount, It reduces the calculation time of the entire inspection.

  A pattern inspection apparatus 2 according to the present embodiment is shown in FIG. Description of the same parts as those in FIG. 1 is omitted. In the present embodiment, the inspection image data Vd obtained from the image photographing unit 10 is output to the deviation measuring unit 20 and also output to the binarizing unit 30. The reference image data Vsd is input to the deviation measuring unit 20 in advance, and the amount of deviation ΔD with respect to the inspection image data Vd is measured in exactly the same manner as in the first embodiment.

  However, the process of correcting the inspection image data Vd based on the deviation amount is not performed, and the inspection image data Vd is directly converted into binary image data Bd.

  On the other hand, the preparation of standard data is different from that of the first embodiment. First, the reference image data Vsd is input to the correction unit 25. The correction unit 25 is set with a predetermined reference deviation amount ΔDsdk. The reference deviation amount is a deviation amount in a predetermined sub-pixel unit. That is, before pattern inspection is performed on an object to be inspected, reference image data whose position is shifted by a predetermined sub-pixel distance is created, and standard data is created based on the reference image data.

  The reference deviation amount ΔDsdk is not particularly limited as long as it is a distance corresponding to a subpixel. For example, a reference shift amount ΔDsd1 moved by 0.5 pixels in the x-axis direction, a reference shift amount ΔDsd2 moved by 0.5 pixels in the y direction, and a reference shift moved by 0.5 pixels in the x and y directions, respectively. An amount ΔDsd3 or the like can be considered. Note that the last k of ΔDsdk representing the amount of deviation is used as a variable representing a number.

  The corrected reference image data created for each reference deviation amount is corrected reference image data Esd1, Esd2, and Esd3 for ΔDsd1, 2, and 3, respectively. Of course, a reference deviation amount ΔDsk smaller than this may be set.

  The correction reference image data Esdk produced by the correction unit 25 is input to the expansion / reduction unit 40. Then, the expansion standard data BEsdxk and the reduced standard data BEsdsk are created for each correction reference image data and stored in the storage unit 80. In the above example, BEsdx1 to BEsdx3, BEsds1 to BEsds3, and the like. These standard data are called minute correction standard data.

  The alignment unit 50 receives the binary image data Vsd and the shift amount ΔD. The alignment unit 50 determines a reference deviation amount that is closest to the distance corresponding to the sub-pixel of the deviation amount ΔD, and obtains standard data created with the reference deviation amount from the storage unit 80. Then, the standard data or the binary image data is moved by a distance corresponding to the pixel unit distance of the shift amount ΔD, and the difference is performed by the calculation unit 60.

  The differences Cdx and Cds can be obtained in the same manner as in the first embodiment, and the result Con can be obtained based on the difference as in the first embodiment.

  FIG. 10 shows a flow of the present embodiment. When the process of the pattern inspection apparatus of the present embodiment starts (S400), an end determination is made (S402). The conditions for determining termination may be the same as in the first embodiment. Inspection image data Vd is captured by the image photographing unit 10 (S404), displacement measurement is performed (S406), and binarization is performed simultaneously (S408). Here, deviation measurement and binarization are described as parallel processing, but they may be performed in order.

  After measuring the amount of deviation, standard data is selected based on the amount of deviation ΔD. In this selection method, it is preferable to select the sub-pixel portion having the shift amount ΔD and the shortest distance between the reference shift amount ΔDsd.

  As described in the first embodiment, the deviation amount is expressed as (m + Δm, n + Δn). Here, m and n are the number of pixels or a distance corresponding to the number of pixels. Δm and Δn correspond to subpixels or distances corresponding to subpixels. The reference deviation amount ΔDsdk is now (Δxsd1, Δysd1),..., (Δxsdk, Δysdk). Δxsdk and Δysdk are subpixels or distances corresponding to subpixels, respectively. Then, (Δm, Δn) which is the sub-pixel portion of the deviation amount and the distance Leng of the reference deviation amount are expressed by Expression (6).

・・・・・（６）

(6)

  Standard data which is corrected and expanded / reduced by the reference shift amount at which the distance Leng is the smallest is selected.

  Next, the expansion standard data Bsdxk and the reduction standard data Bsdsk selected here are moved by the pixel value of the shift amount ΔD (S412). The movement of the pixel value corresponds to the movement of (−m, −n) when the amount of deviation is represented by (m + Δm, n + Δn) in the above example. In this process, standard data is created from reference image data that has been moved by a distance of subpixels in advance, and the subpixel misalignment between the inspection image data and the reference image data is preliminarily created in advance. The correction is to be made with standard data produced by moving the subpixels. In other words, in the present embodiment, steps S410 and S412 correspond to a process of performing correction based on the amount of deviation.

  Comparison of the selected standard data and binary image data and calculation of the difference are the same as in the first embodiment.

  In the present embodiment, it is not necessary to move the inspection image data Vd in units of subpixels as in the case of the first embodiment. Therefore, the calculation amount is smaller than that of the first embodiment, and the pattern inspection can be performed on the inspection object at high speed.

  The present invention can be suitably used for automatic defect inspection of circuit patterns.

It is a figure which shows the structure of the pattern inspection apparatus of this invention. It is a figure which shows the flow of a process of a pattern inspection apparatus. It is a figure explaining the principle of this invention which corrects the error for 1 pixel in the case of quantization of image data with the image data before binarization. It is a figure which shows the flow of the process which detects deviation | shift amount. It is a figure explaining the process which corrects the position of test | inspection image data Vd by the amount of deviation | shift. It is a figure explaining the flow of a binarization process. It is a figure which shows the flow of the process which calculates | requires the difference of expansion | swelling standard data and test | inspection image data. It is a figure which shows the flow of the process which calculates | requires the difference of reduction standard data and test | inspection image data. It is a figure which shows the structure of the pattern inspection apparatus of Embodiment 2 of this invention. It is a figure which shows the flow of a process of the pattern inspection apparatus of Embodiment 2. FIG. It is a figure explaining generation | occurrence | production of the error for 1 pixel in the case of quantization of image data.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1, 2 Pattern test | inspection apparatus 10 Image pick-up part 20 Deviation measuring part 30 Binarization part 40 Expansion / reduction part 50 Positioning part 60 Calculation part 70 Determination part

Claims (7)

A pattern inspection method for detecting a defect by comparing binary image data obtained by photographing an inspection object and binarizing the image data and standard data obtained by binarizing and expanding or reducing the reference image data,
Comparing the image data of the object to be inspected and a partial area of the reference image data, and detecting a shift amount between the image data with sub-pixel accuracy;
A pattern inspection method including a step of performing correction based on the shift amount. The step of detecting the shift amount includes moving either the image data of the inspection object or the reference image data in units of pixels, and the image data of the inspection object and the reference image data for each movement. Obtaining a normalized correlation coefficient of
The pattern inspection method according to claim 1, further comprising a step of interpolating the normalized correlation coefficient to obtain a deviation amount with an accuracy of 1 pixel or less. The step of performing the correction includes
The pattern inspection method according to claim 1, wherein the pattern inspection method is a step of moving image data of the inspection object by the amount of deviation. The moving step includes
Moving the image data of the inspected object by a distance in pixels;
The pattern inspection method according to claim 3, further comprising a step of interpolating a distance in sub-pixel units from the image data of the inspection object. 5. The pattern inspection method according to claim 4, wherein the step of interpolating the distance in units of subpixels is interpolation using a bicubic interpolation method. The standard data is
Interpolating the distance of sub-pixel unit of the reference image data, binarized, and then expanded or reduced micro-corrected standard data,
The step of performing the correction includes
A step of selecting the minute correction standard data corresponding to a shift amount of a sub-pixel among the shift amounts;
3. The pattern inspection method according to claim 1, wherein the pattern correction method is a step of moving the minute correction standard data by a distance corresponding to a shift amount in pixel units of the shift amount. A photographing unit for photographing the inspection object and outputting the image data;
A reference image data and a deviation measuring unit for measuring a deviation amount from the image data;
A binarization unit that outputs binary image data obtained by binarizing the image data;
A storage unit in which the reference image data is corrected by a distance corresponding to a predetermined sub-pixel, and the binary data is expanded or reduced, and then a minute correction standard data is recorded.
An alignment unit that selects the minute correction standard data that is corrected by a distance closest to the sub-pixel among the deviation amount, and moves the minute correction standard data by a distance corresponding to the pixel among the deviation amount; and
An arithmetic unit for comparing the moved minute correction standard data and the binary image data to obtain a difference;
The pattern inspection apparatus which has the determination part which determines the presence or absence of a defect based on the said difference.

Images