JP2000251069A - Pattern binarization method and wiring pattern inspection method using the same - Google Patents

Abstract

[PROBLEMS] To detect a wiring pattern formed on a base material with high precision. SOLUTION: A basic threshold value th1 and a shortage defect (for example,
A binary image based on three thresholds including a threshold thL for detecting pinhole defects) and a threshold thH for detecting excess system defects (for example, scattered defects), and secondary differential thresholds th2 and th3 for detecting hairline shorts and hairline breaks, for example. Second derivative by
By logically synthesizing the value image and the value image, a binary image of an accurate wiring pattern is detected. It is possible to individually set the detection conditions for under- and over-defects, so that these thresholds are optimally set, and stably detect minute pinhole defects, etc., which were difficult with the conventional technology. It is possible, and a fine pattern inspection can be realized.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for binarizing patterns in industrial image processing, and more particularly to a method for automatically inspecting a wiring pattern in an automatic appearance inspection apparatus for printed wiring boards, ceramic substrates, green sheets, and the like. The present invention relates to a pattern binarization method to be detected and a wiring pattern inspection method using the same.

[0002]

2. Description of the Related Art In a conventional inspection apparatus for printed circuit boards and the like, for example, an appearance inspection apparatus, a simple binarization, that is, a binarization process using a single binarization threshold is generally used as a pattern detection method. The method is described in
In the case of the binarization device described in Japanese Patent Publication No. 988, in order to stably detect a thin hairline-shaped disconnection defect occurring in a wiring pattern, a binary image with a fixed threshold and a binarized image of a second derivative are synthesized. The system has been adopted. According to this method, a wiring pattern to be detected is detected at a fixed threshold, and a hairline-shaped disconnection defect or a short-circuit defect that short-circuits between wiring patterns is detected by a second differential image.
It is configured to be able to detect actualization.

[0003] In addition, "Journal of the Japan Society for Precision Engineering JSPE-59-0"
6 ”1993 pp. 993-1000“ High-precision printed circuit board pattern inspection equipment by comparison with design patterns ”
The binarization method for extracting minute amplitude defects is described in detail. This method also detects minute defects by applying an operator based on differentiation.

[0004]

In the above-mentioned prior art, a thin hairline-like defect portion appears as a steep signal change by second-order differentiation of a detected grayscale image of a wiring pattern. This is to detect such a defect.

However, a pinhole-like defect generated inside the conductor pattern (hereinafter referred to as a pinhole defect)
Also, in the case of a semi-short defect in which the conductor interval is equal to or less than the reference value (design specification), although the short circuit is not actually occurring, the signal change in the grayscale image generally becomes gentle.
The effect of revealing such a defect by the second derivative processing is small, and the signal contrast of a minute pinhole defect is also low.
Not detectable.

On the other hand, as defects occurring in the wiring pattern, insufficient defects (also referred to as missing defects, such as defects such as pinholes, disconnections and half-disconnections) and excess defects (or protruding defects). , Which correspond to defects such as scattering, isolated points, semi-short and short), but it is difficult to treat these defect modes equally with a single binarization threshold.

SUMMARY OF THE INVENTION An object of the present invention is to solve such a problem and to stably extract a detection target from an image without depending on a difference in graphic properties due to the shape or area of the detection target or a change in target pattern density. It is an object of the present invention to provide a binarization method and a binarization threshold value determination method that can be performed.

[0008]

In order to achieve the above object, the present invention provides a method of photoelectrically converting a wiring pattern formed on an object to be inspected such as a printed circuit board or a ceramic substrate, and inputting a grayscale image thereof. The gray level image is converted into a second differential image by a binarizing means and binarized into a second differential binary image, and the gray level image is binarized into a binary image by three threshold values. The binarization of the gray image is performed by the second derivative 2
The value image is determined by logical combination of the binary image with the three thresholds, the binary image obtained by the logical combination is displayed on the image display means so that the output result can be confirmed, and the image is displayed on the image display means. The configuration is such that the setting of the inspection condition is supported by the image information, thereby facilitating the inspection of the wiring pattern.

In the above-mentioned binarizing means, binarization using three threshold values is performed by using a basic threshold value th1 and a low threshold value th for detecting an under-related defect.
L and a high threshold thH for surplus system defect detection, and the second derivative binarization is performed by a differential threshold th2 for detecting a short system defect and a differential threshold th3 for detecting a disconnection system defect.

Further, in the image display means, the image display selecting means includes three thresholds, ie, the reference threshold th1, the threshold thL for detecting an insufficient defect, and the threshold th for detecting a surplus defect.
thH and the binary differentiation binarization, that is, the logical synthesis of the binarization image by the thresholds th2 and th3, and the binarization by the three thresholds and the binarization by the secondary differentiation And an image is arbitrarily selected from the binarized images and displayed.

Further, the conditions set by the means for supporting the condition setting are configured from the binarized image of the pattern to-be-inspected and the settings and inspection conditions relating to the respective binarization thresholds.

Further, when performing the binarization, the thresholds th1, th2, th3, thL and thH are set using limit samples of a defect mode to be detected. Things.

Further, when performing the above-mentioned binarization, the switching of the binarization function is performed by the above-mentioned five thresholds th1, th2, th
3, a configuration determined by the set values of thL and thH.

[0014]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Printed circuit boards and ceramic substrates,
As shown in FIG. 4, a wiring pattern formed on a green sheet (hereinafter, represented by a printed circuit board) has a disconnection of the wiring pattern 1 or a half width of the wiring pattern 1 smaller than a reference value (design specification). Insufficient defects (also referred to as missing defects or concave defects) typified by defects such as disconnections and pinholes and surplus defects (projective defects or convex defects) typified by defects such as half shorts, shorts, and scattering A defect). A wiring pattern inspection device such as a visual inspection device inspects the presence or absence of such a defect.

FIG. 5 is a diagram showing a pattern connection relation comparison method which is an example of such a defect inspection algorithm in the wiring pattern inspection apparatus. Since the details of such an inspection algorithm are described in Japanese Patent No. 1589793, the principle will be briefly described here.

In FIG. 1, the connection relation comparison method binarizes a grayscale image of a detected wiring pattern and expands the binary pattern into a thick line to form a semi-short portion (interval between patterns is equal to or less than a reference value). ) Into an image pattern that converts the binary pattern into a short-circuit state, and an image pattern that contracts this binary pattern into a thin line to make a half-disconnected part (a part whose pattern width is equal to or less than a reference value) a disconnected state. The two types of processing of the conversion system and the contraction system are performed, and the image pattern obtained by the expansion system and the image pattern obtained by the contraction system are padded (circular portion at the end of the pattern).
It extracts a connection relationship between them and compares it with design information to detect a defect such as a semi-short circuit or a semi-disconnection.

This is based on the inspection rule that a defect occurs when the conductor interval is equal to or less than a specified interval, or a defect when the conductor width is equal to or less than the specified width. In the shrinkage system, the conductors are cut to eliminate the conductor width less than a specified value, and inspected for defects.
Therefore, when the expansion pattern is formed from the original wiring pattern, the expansion amount is the minimum conductor interval, and the contraction amount is the minimum pattern width.

Here, the connection relation extraction processing means that a number (pad number) is assigned to the center of a pad and the pad number is propagated along a pattern. When the pad number reaches another pad number, that pad number is This is a method of storing as if there was a connection with the pad number of the starting point.

According to FIG. 5, in the case of the expansion system, the connection relationship that the pad number is connected to the pad number is extracted, but according to the design information, the pad number and the pad number are different from each other. Since the number and the pad number are connected to each other, there is a defect such as a semi-short or short between the pad number and the pad number by comparing the extracted connection relationship with the design information. I understand.

Further, in the case of the contraction system, a connection relationship that only the pad number is connected is extracted in the same manner. According to the design information, the pad number,
Since the spaces are also connected, it can be seen that there is a defect such as a partial disconnection or disconnection between the pad numbers.

FIG. 6 is a diagram showing the entire configuration of the inspection algorithm of the wiring pattern inspection apparatus.

In FIG. 1, a pattern detector 100 photoelectrically converts a light image from a printed circuit board to obtain a signal of a grayscale image of the surface of the printed circuit board.
This grayscale image is sent to an expansion system and a contraction system.

In the dilation system, the grayscale image of the wiring pattern is binarized 101, and the obtained binary pattern is dilated 110a as described above. Data compression processing 1 to reduce and change
After performing 11a, the image is temporarily stored 112 in the image memory. The same applies to the reduction system. The grayscale image of the wiring pattern is binarized 102, the obtained binary pattern is reduced 120 as described above, and the reduced pattern is converted to an image while maintaining the connection relationship. Data compression processing 11 for reducing and changing
After performing 1b, the image data is temporarily stored 113 in the image memory.

Thereafter, the expansion pattern and the reduction pattern are read from the image memory to extract the respective connection relations.
4. Then, each extraction result is compared with the design information 115 to determine a defect 116, and based on the determination result, a connection-related portion including a defect in the expansion pattern and the contraction pattern is extracted 117 to specify the defect position. 118.

In the inspection of such a wiring pattern,
In the expansion system, short-circuits as shown in FIG.
Small surplus defects such as half shorts, narrow hairline shorts, fine scattering, and isolated points can be clearly detected as defects, and an expansion pattern that eliminates insufficient defects such as hairline breaks and pinholes can be obtained. It is also necessary for the contraction system to have not only a disconnection or a partial disconnection as shown in FIG.
Insufficient defects such as hairline breaks and fine pinholes with narrow widths can be clearly detected as defects, and it is necessary to obtain a shrinkage pattern that eliminates redundant defects such as hairline shorts, fine scattering, and isolated points. It is.

Conventionally, as the binarization processes 101 and 102, the detected grayscale image is binarized by a simple binarization method using a single binarization threshold, and the binarization is performed by an expansion system and an erosion system. However, it is not possible to set a binarization threshold value capable of detecting all of the above-mentioned surplus and deficiency defects. On the other hand, the present invention enables binarization 101 and 102 so that all of the above-mentioned defects can be detected by processing in the expansion system and the contraction system.

An embodiment of the pattern binarizing method according to the present invention will be described below. In this embodiment, binarization using three threshold values and binarization using secondary differentiation are used together, and five types of threshold values for the binarization can be appropriately set. First, the five types of thresholds will be described with reference to FIG.

FIG. 1A shows a specific example of a wiring pattern on a printed circuit board.
2 is a hairline disconnection, 3 is a pinhole, 4 is a hairline short, 5 is scattered, and 6 is a substrate portion of a printed circuit board.

In FIG. 1A, it is assumed that the surface of the base portion 6 of the printed circuit board has high brightness and is bright, and the wiring pattern 1 having low brightness and dark is formed on this surface.
Here, hairline breaks 2 and pinholes 3 are generated in the wiring pattern 1, and a hairline short 4 is formed between the wiring pattern 1 and another wiring pattern. It is assumed that scattering 5 has occurred.

The gray-scale image signal 10 read along the line AA on the printed board is, as shown in FIG.
, And the level change portion at the boundary between the low level portion and the high level portion becomes the edge portion of the wiring pattern 1.

Here, the presence of the hairline disconnection 2 in the dark wiring pattern 1 is brighter than the wiring pattern 1 portion, but is darker and thinner than the base portion 6 and approaches the brightness of the wiring pattern 1 as it becomes thinner. . Therefore, the grayscale image signal 1
At 0, the level of the hairline break 2 is slightly higher than the level of the wiring pattern 1 as shown in FIG. The pinhole 3 also has the brightness between the wiring pattern 1 and the base member 6 as in the case of the hairline disconnection 2, but has a larger area than the hairline disconnection 2. As shown in FIG. 1B, the level is higher than that of the hairline break 2.

Also, the hairline short 4 and the scattering 5
Since the periphery is surrounded by the base member 6, it has brightness close to that of the base member 6, and the brightness approaches the base member 6 as the width or size decreases. Therefore, in the grayscale image signal 10, as shown in FIG.
The level of the hairline short 4 and the scatter 5 is lower than the level of the base 6, but higher than the level of the hairline disconnection 2 and the pinhole 3. The larger the difference in brightness (contrast) from the wiring pattern 1 becomes, the larger the difference becomes. Again,
Since the scattering 5 has an area with respect to the hairline short 4, the level of the scattering 5 is lower than the level of the hairline short 4 in the grayscale image signal 10.

As described above, the signal level of the grayscale image signal 10 is significantly different between the shortage defect such as the hairline disconnection 2 and the pinhole 3 and the surplus defect such as the hairline short 4 and the scattering 5. In a conventional binarization method using a single binarization threshold, a binarization pattern including all of these defects cannot be obtained.

In this embodiment, as shown in FIG. 1B, for the grayscale image signal 10, th1: basic threshold value thH: high threshold value for scattering detection thL: low threshold value for pinhole detection .

The detection of the hairline disconnection 2 and the hairline short 4 uses a binarization method of the secondary differential signal of the grayscale image signal 10 together. FIG. 1 (c) shows this 2
This shows the next differential signal 11. Since the signal change in the portion of the hairline disconnection 2 in the grayscale image signal 10 is steeper than that in the portion of the pinhole 3, as shown in FIG.
In the second derivative signal 11, the hairline break 2 has a larger amplitude than the pinhole 3. Similarly,
In the second derivative signal 11, the hairline short 4 has a larger amplitude than the scattering 5. From this,
For this secondary differential signal 11, th2: a threshold for detecting a hairline short-circuit for the secondary differential image, and th3: a threshold for detecting a hairline disconnection for the secondary differential image.

FIG. 1 (d), there is shown a binary pattern A ₁ obtained by binarizing the basic threshold th1 which is properly set the gray-scale image signal 10 in FIG. 1 (b), the the grayscale image When the brightness (shade value) of the signal is I,

[0037]

(Equation 1)

Is as follows. FIG. _1H shows a binary pattern A _H obtained by binarizing the grayscale image signal 10 in FIG. 1B with an appropriately set scattering detection threshold thH,

[0039]

(Equation 2)

Is as follows. Figure 1 (j), there is shown a binary pattern A _L obtained by binarizing pinhole detection threshold thL that is properly configured grayscale image signal 10 in FIG. 1 (b), the

[0041]

(Equation 3)

Is as follows.

[0043] Further, FIG. 1 (i) is intended to indicate a binary pattern A ₂ obtained by binarizing with hairline short detection threshold th2 that is properly configured secondary differential signal 11 in FIG. 1 (c) Then, the second derivative values are expressed as I " _a , I" _b , I " _c , I" _d
As

[0044]

(Equation 4)

Is as follows. Here, the second derivative I " _a is shown in FIG.
The grayscale image signal 10 shown in (b) are those obtained by the action of second derivative operator of 5 × 5 pixel units shown in FIG. 2 (a), below, the secondary differential value I _"b , I " _c , I" _d are shown in FIG.
This is obtained by applying a 5 × 5 second derivative operator in pixel units shown in (b), (c) and (d). FIG. 1 (k) shows a properly set hairline short detection threshold for the second derivative signal 11 in FIG. 1 (c).
13 shows a binary pattern A ₃ obtained by binarization at th3,

[0046]

(Equation 5)

Is as follows.

Note that the second derivative operator shown in FIG. 2A makes the pattern edge along the horizontal direction obvious, and the second derivative operator shown in FIG. 2B shows the pattern edge along the upward right direction. FIG. 2
The second derivative operator shown in FIG. 2C makes the pattern edge along the vertical direction obvious, and the second derivative operator shown in FIG. 2D makes the pattern edge along the downward right direction appear. . That is, these secondary differential operators make the edge portion of the input grayscale image obvious, and the steeper the edge portion, the higher the secondary differential value I " _a ,
I " _b , I" _c , and I " _d are large, and therefore, as shown in FIG.
The second derivative value of the part of the part is the value of 2
It is larger than the second derivative.

By the way, when the binarization by the above equations _{1 to 5} is performed, the binary pattern A ₁ based on the reference threshold value becomes
As shown in (d), the value becomes "1" at the wiring pattern 1 portion and "0" at the base material portion 6 at the pattern edge. That is, it is a signal whose level is inverted at the pattern edge.

As shown in FIG. 1 (h), the binary pattern A _H based on the scattering detection threshold thH of the grayscale image 10 is
Becomes “1” for a defect such as a hairline short 4 or a scatter 5 existing in the scatter, and by appropriately setting the scatter detection threshold thH, the signal component for the scatter 5 becomes a time corresponding to the actual dimension of the width of the scatter 5. Included as width signal.
However, the binary pattern A _H may include not only the signal component for the scattering 5 but also the signal component for the hairline short circuit 4, and is always “1” in the wiring pattern 1. .

[0051] Further, binary pattern A _L by pinhole detection threshold thL grayscale image 10, as shown in FIG. 1 (j), such as hairline disconnection 2 and pinholes 3 existing in the wiring pattern 1 against defects By setting the threshold thL for pinhole detection appropriately, the signal component for the pinhole 3 is included as a signal having a time width corresponding to the actual dimension of the width of the pinhole 3. In the portion of the base member 6, the binary pattern _AL is always "0".

The differential binary pattern A ₂ of the secondary differential pattern 11 based on the hairline short detection threshold th2 is shown in FIG.
As shown in (i), hairline short detection threshold th
By setting 2 properly, hairline short 4
Is included as a signal of "1" having a time width corresponding to the actual size of the width of the defect, but depending on the steepness of the edge portion, the edge portion of the pinhole 3 or the scattering 5 or the wiring pattern 1 and the like.

The differential binary pattern A ₂ of the secondary differential pattern 11 based on the hairline disconnection detection threshold value th3 is shown in FIG.
As shown in (2), by appropriately setting the hairline disconnection detection threshold value th3, the signal component for the hairline disconnection 2 is included as a signal of “0” having a time width corresponding to the actual dimension of the width of the defect. Depending on the steepness of the edge portion, the edge portion such as the pinhole 3 or the wiring pattern 1 is also included.

Now, in the binary pattern as described above, the binary pattern A _L , A ₃ is subjected to the logical product process, so that the time corresponding to the actual size of the hairline disconnection 2 and the defect of the pinhole 3 is obtained. A binary pattern ( _{AL.A 3} ) containing a signal component of a width is obtained. Similarly, 2
By performing a logical AND operation on the value patterns A _H and A ₂ , a binary pattern (A _H + A ₂ ) including signal components of a time width corresponding to the actual size of each defect of the hairline short 4 and the scattering 5 is obtained. can get. Here, * represents a logical product and + represents a logical sum.

[0055] By the way, the binary pattern (A _L · A ₃₎
As for, because hair line disconnection 2 and pin holes 3 is a defect existing in the area of the wiring pattern 1, in order to obtain the signal components of these defects, of the binary pattern (A _L · A ₃₎ It is necessary to extract only the area of the wiring pattern 1. Similarly, looking at the binary pattern (A _H + A ₂ ), the hairline short 4 and the scatter 5 are defects existing in the region of the base material 6,
In order to obtain the signal components of these defects, it is necessary to extract only the part of the area of the base material part 6 from the binary pattern (A _H + A ₂ ). The edge of the wiring pattern 1 is
It is detected by binarization using the thresholds th1, thH, thL, th2, and th3, but is not a defect. Also, thresholds th1, th
In the binarization using H, thL, th2, and th3, the edge of the wiring pattern 1 is not correctly detected. That is, the basic threshold value th1 is for detecting the edge portion of the wiring pattern 1, and by appropriately setting the basic threshold value th1, the edge portion of the wiring pattern 1 is detected at a correct position. The level inversion portion of the binary pattern A ₁ shown in FIG. 1D represents the edge of the wiring pattern 1.

[0056] In the foregoing, the hair line disconnection 2 and pin holes 3 binary pattern (A _L · A _3), hairline Short 4 and scatters 5 binary pattern (A _H +
From A ₂ ), it is necessary that the edge of the wiring pattern 1 be obtained from the binary pattern A ₁ .

Therefore, in this embodiment, in the signal time axis direction, the region of the wiring pattern 1 excluding the edge portion, the edge region including the edge portion of the wiring pattern 1 and the region of the base portion 6 are further divided. Divide the area. FIG. 1E shows a binary pattern A _NL representing a region of the wiring pattern 1 (this region of the wiring pattern 1 is also referred to as a wiring pattern region A _NL ).
FIG. 1F shows a binary pattern A _NM representing an edge area of the wiring pattern 1 (the edge area is also referred to as an edge area A _NM ), and FIG. binary pattern a _NH representing the area (Incidentally, this region of the base portion 6 also called base region a _NH) represents respectively a. Such regions A _NL, A _NM, A _NH is, N the number of effective pixels surrounding the "1" of the target pixel in the later-described region division operator, the determination value as described below N _L, as N _H, the following It is represented by Equation 6.

[0058]

(Equation 6)

However, A _NL + A _NM + A _NH = 1.

[0060] Then, in order to detect hairline disconnection 2 and pin holes 3 in the wiring pattern area, binary pattern (A _L · A _3), calculation of A _{NL, i.e., A NL · (A L ·} A ₃ ) in order to detect the hairline 4 and the scattering 5 in the region of the base material portion 6, the binary pattern (A _H +
A _2), calculation of A _NH, i.e., performs _{A NH · (A H + A} 2). Further, in order to detect the edge portion of the wiring pattern 1 in the edge region, performs processing of _{A NM · (A 1 · A} 3 + A 1 · A 2). In the above equation, A ₁ · A ₃
Is the level in the edge region and A in the wiring pattern region.
_{NL · (A L · A 3} ) is the level of an operation for aligning the, also, levels and groups in the edge region when A ₁ · A ₂ has a binary pattern A1 is level inversion and illustrated This is an operation for matching the level of A _NH · (A _H + A ₂ ) in the material region.

Therefore, the binarized output A _R obtained by the above binarization processing is:

[0062]

(Equation 7)

Is obtained. This output _AR is shown in FIG.

Next, the above-described area division will be described.

Here, the division into the wiring pattern area A _NL , the edge area A _NM , and the base material area A _NH is performed by dividing the area into a binary pattern A ₁ based on the basic threshold th1 shown in FIG. The operator operates to determine which of the areas A _NL , A _NM , and A _NH the pixel of interest in the binary pattern A ₁ belongs to. This region division operator acts on a two-dimensional array of pixels in the binary pattern A ₁ , and depends on how the target pixel is surrounded by surrounding effective pixels serving as a determination reference. This determines whether the pixel of interest belongs to the area A _NL , A _NM or A _NH .

FIG. 3 shows a specific example of the area dividing operator. In this specific example, the effective pixels are pixels B on the circumference around the pixel of interest P, and it is determined whether or not these effective pixels B are “1”. The number N of the effective pixels B is counted, and it is determined according to the count value N which of the areas A _NL , A _NM , and A _NH the pixel of interest P belongs.

Here, when the diameter of the circumference where the peripheral pixel B exists is n pixels in pixel units, this area dividing operator is called an operator of size n × n. The operator size is set in accordance with the line width of the wiring pattern 1 and the edge width of the wiring pattern. In the case of an inspection apparatus for inspecting a printed circuit board having a wiring pattern with a different line width or an edge width, A region dividing operator having an optimum size is provided in advance for each printed circuit board, and the region dividing operator to be used is made different depending on the printed circuit board to be inspected.

FIG. 3 shows that the sizes are 7 × 7, 9 × 9 and 1
A 1 × 11 area division operator is shown. Size 7
In the × 7 region division operator, the number of effective pixels B is 16 pixels, and two judgment values N _L = 14,
N _H = 2 is set. In the area division operator having a size of 9 × 9, the effective pixels B are 24 pixels, and
Two determination values N _L = 22 and N _H = 2 are set for the number of pixels, and the effective pixel B is 28 pixels in the area division operator of size 11 × 11. Determination values N _L = 25 and N _H = 3 are set.

The determination value N of such an area dividing operator
_{Based on L} , N _H and the number N of the effective pixels B of “1”, the areas A _NL , A _NM ,
A _{NH is} to be determined.

By the way, if the above-mentioned defect does not occur, when all the effective pixels B are "1", it can be determined that the pixel of interest P is in the area of the wiring pattern. When “0”, it can be determined that the pixel of interest P is in the region of the base material portion. Further, when some effective pixels B are “1” and the effective pixels B of stiffness are “0”, the pixel of interest P Can be determined to belong to the region of the edge portion of the wiring pattern. For this reason, for example, it is conceivable that N _L = 16 and N _H = 0 in a region division operator of size 7 × 7.

[0071] However, in this case, for example, when the region segmentation operator size 7 × 7, a circumference lined effective pixel B in the wiring pattern area A _NL is hairline disconnection of the width of one pixel or less across, 2 Effective pixels are “0”
Thus, the number N of effective pixels B of “1” becomes N = 14, and N
Since <N _L , an erroneous determination is made as the area A _{NM at} the edge of the wiring pattern despite the wiring pattern area A _NL . Therefore, in FIG.
_L is set to a value smaller than the total number of effective pixels B by the width of the hairline break. Similarly, also the determination value N _H, to have a value of width of the hairline short, even with hairline short, thereby, the area A _NH substrates portion
Is not erroneously determined as the area _{ANM at} the edge of the wiring pattern.

By setting the determination values N _L and N _H in this manner, the respective regions of the wiring pattern, the edge portion, and the base portion can be correctly determined, so that the hairline disconnection in the wiring pattern region A _NL can be prevented. Further, it becomes possible to detect a hairline short-circuit in the area _ANH of the substrate portion.

The width of the hairline disconnection or the hairline short may be made different depending on the size of the region dividing operator, and the width of such a defect in the wiring pattern using the region dividing operator is determined. It is determined accordingly.

Here, the effective pixel B is replaced with the pixel of interest P
Although it was arranged on a circumference that is isotropically arranged with respect to, but is not limited thereto, for example,
The range of all pixels in a circle centered on the pixel of interest P and all pixels in a square centered on the pixel of interest P can be changed as appropriate according to the graphic properties and specifications of the inspection object.

Next, the relationship between the area dividing operator and the line width of the wiring pattern will be described with reference to FIG.

FIG. 7A shows an area dividing operator (here, the operator size is 5 × 5) whose size is equal to or less than the line width of the wiring pattern 1 (here, this line width is 5 pixels). The drawing shows an operation of crossing the wiring pattern 1 in the right direction on the drawing, and it is assumed that the position of the area dividing operator with respect to the wiring pattern 1 changes in the order of (i), (ii), and (iii).

In the state (i), all the effective pixels B around the target pixel P are in the area of the wiring pattern 1, and all the effective pixels B at this time are “1”. Here, all effective pixels B
Assuming that the ratio of the number N of effective pixels B of "1" to the number of pixels is α, in this state (i), α = 12/12. Next, in the state of (ii) in which the area dividing operator has moved rightward by two pixels with respect to the wiring pattern 1, the area dividing operator 5 on the right side
Effective pixels B protrude from wiring pattern 1 and are “0”
In this case, the ratio α = 7/12. Further, when the area dividing operator moves rightward by one pixel with respect to the wiring pattern 1 (iii), the right seven effective pixels B protrude from the wiring pattern 1 and become “0”. , The ratio α = 5/12.

As described above, when the operator size is smaller than the line width of the wiring pattern 1, when the area dividing operator is within the wiring pattern 1, the ratio α = 1, which is the maximum, and the area dividing operator becomes the wiring pattern 1 Α becomes smaller as the distance from the distance increases.

On the other hand, FIG. 7B shows a case where the operator size is larger than the line width of the wiring pattern 1 (here, this line width is set to 5 pixels) (here, the operator size is 7 × 7). ). Also in this case, similarly to the case of FIG. 7A, the operation of the area dividing operator crossing the wiring pattern 1 in the right direction in the drawing is shown, and the position of the area dividing operator with respect to the wiring pattern 1 is (a). , (B) and (c) in this order.

In the state (a), the target pixel P is the wiring pattern 1
At the center in the width direction. In this case, the operator size is 7 × for a line width of 5 pixels.
7, three effective pixels B on both the left and right sides protrude from the area of the wiring pattern 1 and are “0”. Therefore,
In the state (a), α = 10/12. Next, in the state of (b) in which the region dividing operator has moved rightward by one pixel with respect to the wiring pattern 1, the five effective pixels B on the right side
Only the portion protruding from the wiring pattern 1 becomes “0”, and in this case, the ratio α becomes 11/12. Further, when the area dividing operator moves rightward by one pixel with respect to the wiring pattern 1 (C), the seven effective pixels B on the right are protruded from the wiring pattern 1 and become “0”. , The ratio α = 9/12.

As described above, when the operator size is larger than the line width of the wiring pattern 1, the region dividing operator is slightly shifted from the wiring pattern 1 to one side rather than the pixel of interest P is located at the center in the width direction of the wiring pattern 1. Then, the ratio α increases, and thereafter, as the area dividing operator departs from the wiring pattern 1, the ratio α decreases.

From the above, when the operator size is equal to or less than the line width of the wiring pattern 1, when the area dividing operator is within the wiring pattern 1, the ratio α = 1 and the maximum, and the area operator Α monotonously decreases as the pattern departs from pattern 1. However, when the operator size is larger than the line width of wiring pattern 1, α is smaller than that of pixel P of interest at the center of wiring pattern 1 in the width direction. Is slightly shifted, α becomes large, and the relationship between α and the positional relationship between the area dividing operator and the wiring pattern becomes unnatural.

Therefore, the size of the area dividing operator must be smaller than the minimum line width of the wiring pattern 1 on the printed circuit board. As a result, the maximum size of the usable area dividing operator is determined according to the minimum line width of the wiring pattern 1 on the printed circuit board to be inspected.

FIG. 8 is a diagram for explaining the minimum size of the usable area dividing operator for the wiring pattern 1 on the printed circuit board to be inspected.

As shown in FIG. 8, the edge portion of the grayscale image signal 10 of the wiring pattern 1
It has a finite period of an inclined waveform in which the brightness (shading value) changes continuously. Here, in this edge portion, the number of pixels of brightness included in the brightness range from the brightness of the pinhole detection threshold thL to the brightness of the scatter detection threshold thH is hereinafter referred to as the number of brightness change pixels.

FIG. 8A shows a binary pattern near an edge when the operator size> the number of brightness change pixels. In this case, the area A of the edge determined by the area dividing operator is shown. because _NM, including the entire edge portion of the grayscale image signal 10, so that the the edge region a _NM in binary pattern a ₁ edge E ₀ obtained in the basic threshold value th1 only is obtained.

On the other hand, as shown in FIG. 8B, when the operator size <the number of brightness change pixels, the edge area _ANM determined by the area division operator is the edge area of the grayscale image signal 10. becomes narrower than the range, the region a _NH area a _NL also base portion 6 of the wiring pattern 1 is determined by the area division operator also come enters the range of the edge portion. Therefore, of course the area A _NM at the edge E ₀ of the binary pattern A ₁ obtained in the basic threshold value th1 is obtained, this region A _NL even binary obtained pinhole detection threshold thL pattern A _L The edge E _{L corresponding} to the edge and the edge E _{H corresponding} to the edge in the binary pattern A _H obtained by the scattering detection threshold thH also occur in the area A _NH .

As described above, when the operator size is smaller than the number of brightness change pixels, the unnatural situation that the same edge portion in the wiring pattern 1 is erroneously detected in all of the areas A _NL , A _NM , and A _NH. Will happen. Therefore, the size of the region dividing operator needs to be equal to or larger than the width of the edge of the wiring pattern 1.

The area dividing operator sets the size to 7 ×
Although it can only take discrete values of 7, 9, 9 and 11, 11
By providing the thresholds N _H and N _L for the count of the number of effective pixels, it is possible to realize an area division operator having a size between such discrete actual sizes. That is, when the thresholds N _H and N _L are made closer to each other, the same effect as the reduction in the operator size can be obtained.

In this embodiment, as described above, a pinhole defect is manifested at a low threshold value thL, and a scattering defect is manifested at a high threshold value thH. Hairline breaks and hairline shorts are exclusively detected by the second derivative threshold. As a result, the binarized output A _R expressed by the above equation (7) is obtained, and the threshold values th1, thH, thL, th2, th
In the case of the defect inspection apparatus shown in FIG. 6, the setting of 3 is performed by the binarizing units 101 and 102 and the image display units 131 and 1 for displaying the binarized image obtained by the binarization and expressed by the above equation (7).
32.

Incidentally, the binarizing means 101 and 102 can switch the binarizing function by appropriately setting the threshold value. That is, in order to perform binarization using second-order differentiation, the threshold value may be set to be th1 = thH = thL. As a result, A ₁ = A _H = A _L is obtained from Expressions ₁ , 2, and 3, so that the binarized output signal A _R shown in Expression 7 is A _R = (A _NL · A ₁ · A ₃ ) + _{{A NM · (A 1 ·} A 3 + A 1 · A 2)} + {A NH · (A 1 + A 2)} = A NL · A 1 · A 3 + A NM · A 1 · A 3 + A NM · A _{1 · A 2 + A NH ·} A 1 + A NH · A 2 = A 1 · A 3 · (A NL + A NM) + A 2 · (A NM · A 1 + A NH) + A NH · A 1 wherein, A ₁ ⊂ By (A _NL + A _NM ), A ₁ · (A _NL + A _NM ) = A
The _{1 A NH ⊂A 1, A NM} · A 1 + A NH = A NM · A 1 + A NH
A ₁ , so A _R = A ₁ · A ₃ + A ₂ · A ₁ · (A _NM + A _NH ) + A _NH · A ₁ Further, A ₁ （(A _NM + A _NH ) gives A ₁ · ( A _NM + A _NH ) = A
_Since A ₁ , A _R = A ₁ · A ₃ + A ₂ · A ₁ + A _NH · A ₁ , and because A _NH ⊂A ₁ , A _NH · A ₁ = 0 and A _R = A ₁ · A ₃ + A ₁ · A ₂ , and the binarization using the conventional second derivative can be realized. However, in the above equation (7), A _NM · A ₁ · A ₂ represents a second-order differential binary image based on the threshold value th2 in the area where A _NM = 1 (pattern edge area) and A ₁ = 0. Yes, and A _NH · A ₃ · A ₂ is the area of A _NH = 1,
In this region, A ₃ = 1, which is equivalent to A ₁ = 1. Therefore, in this case, (A _NM · A ₁ · A ₂ + A _NH
A ₃ · A ₂ ) becomes A ₁ · A ₂ , and the above-mentioned binarized output signal A _R is obtained.

On the other hand, in order to function as a three-threshold binarization without using the second-order differential binarization, the thresholds of the second-order differential image I ″ may be set to th2 = ∞ and th3 = −∞. Since A ₂ = 0 according to Equation 4 and A ₃ = 1 according to Equation 5, when this is substituted into Equation 7, the binary output signal A
_R would _{A R = A NL · A L} + A NM · A 1 + A NH · A H , and the three threshold binary pattern obtained.

Further, when functioning as simple binarization (binarization using only the threshold th1), th1 = thH = th
L, th2 = ∞, th3 = −∞. This allows
_{A 1 = A H = A L} , A 2 = 0, A 3 = 1 , and the and A _NL + A
_{Since NM} + A _NH = 1, if this is substituted into the above equation 7, the binarized output signal A _R becomes A _R = A ₁ .

As described above, in this embodiment, it is possible to cope with the detection of various kinds of defects and to switch the functions flexibly.

Next, a method of setting the binarization thresholds th1, th2, th3, thH, and thL for accurately detecting a pattern will be described.

For this setting, a limit defect sample of each defect mode (that is, a type of defect such as a broken hairline or a short hairline) is used as a reference. The reference sample may be a real defect or a simulated defect. That is, there are hairline breaks and hairline shorts having the minimum width to be detected, semi-shorts having the minimum conductor spacing, half-break defects having the minimum line width, pinholes and scattering defects having the minimum detection size, and the like.

FIG. 9 shows a specific example of the criteria for setting each binarization threshold in the expansion system and the contraction system of FIG.

When setting each threshold in the expansion system of FIG.
The grayscale image from the pattern detection means 100 with a sample of the supplied to the binarization unit 101 to display the binary pattern A _R obtained here in the image display unit 131, by adjusting the threshold, optimum This is set as a threshold. Hereinafter, the setting method will be described with reference to FIG.

In the case of an expansion system, a reference threshold value for pattern detection
th1 is a detection target which is a short-circuit with the edge of the pattern (a defect in which the conductor pattern interval is narrower than the minimum interval: see FIG. 4), and is obtained by this reference threshold th1. As described above, when the binary pattern is expanded by a specified width, a threshold value is set so that the half short-circuit is always short-circuited.
This is a reference when determining the threshold value thH for scattering detection. Therefore, it is necessary to set the threshold value th1 so that the minimum conductor interval (not a defect) defined in the binarized pattern obtained by binarizing the grayscale image with the reference threshold value th1 is detected to the actual size. is there. This means that if there is no half short, even if the space between the conductor patterns is the prescribed minimum, no short circuit will occur when inflated. In order to set the reference threshold th1 in this way,
th1 = thH = thL, th2 = ∞, th3 = a -∞, and using the gray-scale image of the minimum conductor spacing as a sample, the binarized output signal A _R in Formula 7 obtained from the sample A _R = adjusting the reference threshold th1 while displaying on the image display unit 131 as a _1.

In the case of the dilation system, the threshold value th2 is used to binarize the second derivative image, and a hairline short or an isolated point as a surplus system defect is set as a defect to be inspected. The size of the defect is set so as to be detected at the actual size. For this purpose, a grayscale image of a sample having the size of the defect is obtained, and this is binary-coded with the reference threshold th1 as described above. To obtain a binarized pattern A ₁ , obtain a second derivative image of the grayscale image, binarize this with a threshold th 2, and detect such a defect having the minimum size in the actual size. Is to adjust. In this case, the binarized pattern A of the binary image based on the threshold value th2 of the secondary differential image is set so that the defect to be inspected is extracted and displayed on the image display unit 131.
Only the “0” area of ₁ is extracted and supplied to the image display means 131. Therefore, th1 = thH = thL, th
By setting 3 = −∞, A ₁ = A _H = A _L = 0 (accordingly, A ₁ = 1) and A ₃ = 1, and the binarized output signal A _R
From the above equation 7, A _R = A _NM · A ₁ · A ₂ + A _NH · A ₂ = (A _NM · A ₁ +
A _NH ) · A ₂ . The area (A _NM · A ₁ + A _NH ) is shown in FIG.
As is clear from FIG. 5, the region where A ₁ = 1.

In the case of the expansion system, the threshold value thH is 2
The value is used to detect scattering as a surplus system defect by making a value, and the defect having the minimum size to be detected in the binarized image is set to be detected in the actual size. In this case, only the “0” region of the binarized pattern A ₁ of the binary image based on the threshold thH of the grayscale image is extracted so that the defect to be inspected is extracted and displayed on the image display unit 131. It is supplied to the image display means 131. Therefore, by setting th1 = thL, th2 = ∞, th3 = −∞, A ₁ = A _L = A ₂ = 0 (accordingly, A ₁ = 1) and A ₃
= 1, and the binarized output signal A _R is A _R = A _NH · A _H from the above equation (7). As is clear from FIG. 1, the area A _NH is an area where A ₁ = 1 excluding the pattern edge portion.

The threshold value th3 is for detecting hairline breakage, which is an insufficient defect in the conductor pattern.
Since thL is for detecting a pinhole which is a shortage defect in the conductor pattern, these thresholds th3 and thL are not used in the expansion system, and are set to the same values as those in the contraction system as an example.

Next, when setting each threshold in the contraction system of FIG. 6, a grayscale image having a sample is supplied from the pattern detecting means 100 to the binarizing means 102, and the binarized pattern A _R obtained here is supplied. Is displayed on the image display means 132, and the threshold value is adjusted to set the optimum threshold value. Hereinafter, the setting method will be described with reference to FIG.

In the case of a contraction system, a reference threshold value for pattern detection
The th1 is a detection target which is an edge portion of the pattern and a semi-disconnection (a defect in which the width of the conductor pattern is narrower than the minimum width defined: see FIG. 4).
As described above, when the binarized pattern obtained at th1 is reduced by the specified width, the threshold value is such that this half-disconnection is always disconnected. Is used as a criterion when deciding.
Therefore, it is necessary to set the threshold value th1 so that the minimum conductor width (not a defect) defined in the binarized pattern obtained by binarizing the grayscale image with the reference threshold value th1 is detected to the actual size. is there. This means that if there is no half-break,
This means that even when the width of the conductor pattern is the prescribed minimum, no disconnection occurs when contracted. In order to set the reference threshold th1 in this manner, th1 = thH = thL, th1
2 = ∞, th3 = −∞, and using the grayscale image of the minimum conductor width as a sample, the binarized output signal A _R of Equation 7 obtained from this sample is set as A _R = A ₁ , and image display means The reference threshold value th1 is adjusted while displaying at 131.

In the case of the contraction system, the threshold value th3 is used to binarize the secondary differential image, the hairline breakage as an insufficient system defect is used as the defect to be inspected, and the threshold value th3 has a minimum size to be detected. The defect is set so as to be detected at the actual size. For this purpose, a gray image of a sample having a defect of such a size is obtained, and is binarized by the reference threshold th1 as described above. In addition to obtaining the binarized pattern A ₁ , a second derivative image of the gray image is obtained and
This threshold value th3 is adjusted so that such a defect having the minimum size binarized by th3 is detected in the actual size. In this case, only the “1” area of the binarized pattern A ₁ of the binary image based on the threshold th3 of the secondary differential image is extracted so that the inspection target defect is extracted and displayed on the image display unit 132. Then, it is supplied to the image display means 132. Therefore, by setting th1 = thH = thL and th2 = ∞, A ₁ = A _H = A _L = 1 (accordingly, A ₁ = 0) and A ₂ = 0, and the binarized output signal A _{From R} , A _R = A _NL · A ₃ + A _NM · A ₃ = (A _NL + A _NM ) · A ₃ (where A ₁ = A _H = 0 in the region of A _NH = 1) From, A
_NH · A ₃ · A _H = 0). Area (A _NL + A
_NM ) is an area where A ₁ = 1, as is clear from FIG.

In the case of the contraction system, the threshold value thL is 2
This is for detecting a pinhole as a defect due to deficiency and is set so that a defect having a minimum size to be detected in the binarized image is detected in an actual size. In this case, the inspection target defect is extracted and the image display unit 132
, Only the area of “1” of the binarized pattern A _{1 in} the binary image based on the threshold value thL of the grayscale image is extracted and supplied to the image display unit 132. For this purpose, by setting th1 = thH, th2 = th, th3 = −∞, A ₁ = A _H = 1 (therefore, A ₁ = 0) and A ₂ =
0, and A ₃ = 1, the binarized output signal A _R, from Equation _{7, A R = A NL ·} A L + A NM · A 1 ( however, in the region of _{A NH = 1 A 1 = A} H = 0, so A
_NH · A ₃ · A _H = 0). As is clear from FIG. 1, the area A _NL is A ₁ excluding the pattern edge portion.
= 1, and A ₁ = 1 in the divided area A _NM (that is, the pattern edge portion).

The threshold value th2 is for detecting a hairline short which is a surplus defect in the conductor pattern.
The threshold value thH is for detecting scattering, which is a surplus system defect in the conductor pattern.
th2 and thH are not used, and are set to the same value as the expansion system as an example.

As described above, in the expansion system shown in FIG.
The threshold value th1 and the threshold value th for detecting a hairline short / isolated point are determined by using the value converting means 101 and the image display means 131.
6 and a threshold value thH for scattering detection, and in the contraction system of FIG. 6, a reference threshold value th1 and a threshold value for hairline disconnection detection are set using the binarizing unit 102 and the image display unit 132.
th2 and a threshold thL for pinhole detection are set.

In the above setting, the relationship between the detection dimensions of the target defect may be obtained using the binarization threshold as a parameter. FIG. 10 shows a setting example of the threshold thL for detecting pinholes and the threshold thH for detecting scattering.

FIG. 10A shows the detection size of a pinhole defect when the threshold thL is changed for a pinhole defect having a diameter of 25 μm. As is apparent from this detection result, the detection size becomes smaller as the threshold thL is increased. From the relationship shown in the drawing, in order for pinhole defects to be accurately detected, the threshold th is required.
It can be seen that L should be set to 50 gradations. Here, the reason why the detected defect size changes depending on the threshold value is due to the modulation of the pattern waveform in the grayscale image. As the reference threshold th1, a value obtained from the minimum conductor interval by the above procedure is used.

FIG. 10B shows the detected size of a scattered defect having a diameter of 28 μm when the threshold value thH is changed. As is apparent from the detection results, the detection size increases as the threshold value thH is increased. From the relationship shown in the drawing, the threshold value thH is set to 90 to 1 in order to accurately detect a scattered defect.
It can be seen that the range should be 10 gradations, and especially, 100 gradations is optimal.

In determining the above-described binarization threshold value, the display of each binarized image and the counting of data relating to the detection size of the reference defect when the threshold value is used as a parameter are performed by the image display means 131 and FIG. 132 is performed. These image display means 131 and 132 can be constituted by, for example, a personal computer and a monitor.

FIG. 11 shows such image display means 131 and 13.
2 is a diagram showing a specific example of the display screen 200 in FIG.

In the figure, image display means 131, 13
The two functions are a storage function of the binary image 201 when each threshold value is changed and a display function on the display screen 200, and a binary image based on the five threshold values th1, th2, th3, thL, and thH. Function to select and display at 200, original image 20
It is also possible to configure so that a comparison display between a binary image and a binary image can be performed on the display screen 200. Also, the display screen 200
It is also possible to configure so that an arbitrary local area 203 of the binary image is enlarged and displayed at a predetermined magnification. Furthermore, it is also possible to configure so that the binarization threshold and the data regarding the defect detection size are totaled and the graph 204 as described in FIG. 10 is displayed on the display screen 200. In addition, it is possible to automate the procedure of setting the above-mentioned binarization threshold value by teaching the measurement position of the dimension in advance in the binary image of the object when summing up such data. is there. When executing the above functions, it is convenient to collect them in the menu 205 by the graphic interface.

Note that, in the configuration of the inspection algorithm of this embodiment, it is assumed that the basic threshold th1 for detecting the minimum line width and the minimum conductor interval according to the actual size is different from each other, and as shown in FIG. Binarization means 1 separately for contraction system
01 and 102 were provided. However, under the condition that the basic threshold value th1 in these expansion system and contraction system is the same or the difference between them is negligible, the binarization of the expansion system and the contraction system may be shared.

In detecting a pattern, the parameters can be standardized by setting a binarization threshold after performing shading correction and dark level correction. That is, this has an effect of minimizing the influence of a change in brightness that depends on a change in the material or the characteristics of the detection system.

[0117]

As described above, according to the present invention,
The pattern input by the predetermined image input means can be binarized and detected with high accuracy as the actual size, so that defects on the pattern can be inspected more stably and precisely. By using the limit value sample of the defect to be detected, the binarization threshold can be easily set. As a result, an excellent effect is obtained in that it is possible to detect a minute defect that has been difficult to cope with in the conventional inspection.

[Brief description of the drawings]

FIG. 1 is a diagram showing an embodiment of a pattern binarization method according to the present invention.

FIG. 2 is a diagram showing a specific example of a second derivative operator for obtaining a second derivative image in FIG. 1;

FIG. 3 is a diagram showing a specific example of an area division operator for area division in FIG. 1;

FIG. 4 is a diagram showing types of defects to be inspected on a wiring pattern.

FIG. 5 is a diagram showing a pattern connection relation extraction process for detecting a defect;

FIG. 6 is a diagram showing an entire configuration of an inspection algorithm using a pattern binarization method according to the present invention.

FIG. 7 is a diagram illustrating a relationship between a minimum width of a pattern and a size of an area dividing operator.

FIG. 8 is a diagram illustrating a relationship between a pattern edge width and a size of an area dividing operator.

FIG. 9 is a diagram showing setting criteria of each binarization threshold shown in FIG. 1;

FIG. 10 is a diagram showing a relationship between a set threshold value and a detection size of a pinhole defect and a scattering defect.

11 is a diagram showing a specific example of a display screen on the image display means shown in FIG.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 wiring pattern 2 hairline disconnection 3 pinhole defect 4 hairline short 5 scattering defect 6 base material section 10 grayscale image 11 secondary differential image 100 pattern detection / image input means 101,102 binarization means 131,132 image display means th1 reference Threshold thH threshold for scattering detection thL threshold for pinhole detection th2 threshold for hairline short detection th3 threshold for hairline disconnection detection

 ──────────────────────────────────────────────────続 き Continuing from the front page (72) Inventor Mitsunori Isobe 292, Yoshida-cho, Totsuka-ku, Yokohama-shi, Kanagawa Prefecture Inside of Hitachi, Ltd. Address F-term in Hitachi, Ltd. Production Engineering Laboratory (Reference) 5B057 AA03 BA23 BA29 CE08 CF01 CF02 CH08 DA03 DB02 DB08 DC16 DC22

Claims (8)

[Claims] 1. A grayscale image of a wiring pattern formed on an object to be inspected such as a printed board or a ceramic substrate is input by an image input means, and the grayscale image is converted into a secondary differential image by a secondary differential processing means. The second derivative binarizing means binarizes the secondary differential image with a predetermined threshold value, and the three-threshold binarizing means converts the grayscale image into two using three threshold values.
Binarizing the grayscale image by logically combining the secondary differential image binarized by the secondary differential binarization unit and the grayscale image binarized by the three-threshold binarization unit; A pattern binarization method characterized by obtaining an image. 2. A gradation image of a wiring pattern formed on an object to be inspected such as a printed board or a ceramic substrate is inputted by an image input means, and the gradation image is converted into a secondary differentiation image by a secondary differentiation processing means. The second derivative binarizing means binarizes the secondary differential image with a predetermined threshold value, and the three-threshold binarizing means converts the grayscale image into two using three threshold values.
Binarizing the grayscale image by logically combining the secondary differential image binarized by the secondary differential binarization unit and the grayscale image binarized by the three-threshold binarization unit; An image is displayed, and the binarized image of the grayscale image obtained by logically synthesizing the binarized image obtained by the binarizing unit and the binarized image of the grayscale image obtained by the logical synthesis are displayed on the image display unit. A threshold value used for obtaining a binarized image can be confirmed; and a condition setting support unit supports setting of inspection conditions from image information displayed on the image display unit. . 3. The image display means according to claim 1, wherein said three-threshold / binarization means includes a reference threshold th1, a threshold thL for detecting a defect in an insufficient system, and a threshold thH for detecting a defect in a surplus system. A binarized image, a binarized image by the differential thresholds th2 and th3 in the secondary differential binarizing means, a binarized image by the tri-threshold binarizing means, and a binary image by the secondary differential binarizing means A pattern binarization method comprising a display image selecting means for selecting and displaying an arbitrary binarized image from a binarized image obtained by logical synthesis with the binarized image. 4. The setting and inspection relating to each of the binarization thresholds according to claim 1, 2 or 3, wherein the setting condition of the condition setting support means is based on a binarized image of the wiring pattern of the inspection object. A pattern binarization method characterized by including a condition. 5. The method according to claim 1, wherein the second differential binarization unit and the three-threshold binarization unit set a threshold for binarization, respectively. A pattern binarization method characterized by being performed using a limit sample of a defect mode. 6. The binarization unit according to claim 1, wherein the binarization is performed by the second-order differential binarization unit and the third threshold binarization unit. By appropriately setting the threshold value in the threshold value binarizing means,
A pattern binarization method, wherein a binarization function of the gray image is switchable. 7. A printed wiring board product, wherein a binarized image of a grayscale image of a wiring pattern is obtained by the pattern binarizing method according to claim 1. 8. A wiring pattern, wherein a defect of the wiring pattern formed on the object to be inspected is inspected using a binarized image obtained by the pattern binarizing method according to claim 1. Inspection methods.