JP2002236908A - Master registering method for searching, and object retrieving method and testing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a master registering method capable of eliminating noise even in the case where the condition of illumination or the like is not good. SOLUTION: In registering a searching master, an image S of a photographed master M is divided into six small areas S1 to S6, and edge point sequence constituting the edge in the respective small areas is detected to obtain the density change angle in the direction of changing the image density in each detected edge point and obtain frequency distribution graphs a1 to a6 of the density change angle. The edge point (indicated by a solid line in b1 to b6) having the same density change angle as the density change angle making the frequency distribution maximum is registered as a feature point of the master.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a search master for comparing positioning at the time of component mounting and alignment of a printed circuit board, a liquid crystal panel, a PDP (plasma display panel) and the like by using image processing. And a method of searching for and inspecting an object.

[0002]

2. Description of the Related Art Conventionally, when registering a search master,
An edge point is detected within the registration range of the master, and an edge point at which the intensity of the detected edge point is equal to or greater than a predetermined threshold value is registered as a feature point of the master.

[0003]

However, the conditions such as the illumination of the article registered as the master are not always the best, and a high contrast area and a low contrast area are mixed. Therefore, when registering an edge point of a low contrast area as a feature point, if an edge of a noise component existing in the weak area is not detected, an edge point of a high contrast area cannot be detected, Further, when an edge point of a region having a high contrast is detected, there is a problem that an edge point of a noise component is registered as a feature point.

Accordingly, an object of the present invention is to provide a method of registering a master, a method of searching for an object, and a method of inspecting an object capable of removing noise even when conditions such as lighting are not good.

[0005]

In order to solve the above-mentioned problems, a search master registration method according to a first aspect of the present invention uses a pre-registered pattern (hereinafter, referred to as a master).
Is a method of registering a master in a search method for searching from an input image, wherein an image obtained by capturing the master is divided into a plurality of small areas, and an edge constituting an edge in each of the small areas. A point sequence is detected, and a density change angle in a direction in which the image density changes at each of the detected edge points is obtained, and a frequency distribution of the density change angle is obtained. In this method, an edge point having a density change angle is registered as a feature point of a master.

According to a second aspect of the present invention, there is provided a method for registering a master in a search method in which a previously registered pattern (hereinafter, referred to as a master) is searched from an input image. In addition, the image obtained by capturing the master is divided into a plurality of small areas, and a sequence of edge points constituting edges in each of the small areas is detected, and the image density at each of the detected edge points changes. And the frequency distribution of the density change angle is obtained, and an edge point having the same density change angle as the density change angle at which the frequency distribution exceeds a predetermined threshold value is registered as a feature point of the master. How to

According to the search master registration method of the present invention, an edge point is detected in each small area of an image obtained by photographing the search master, and a density at the detected edge point is detected. Find the change angle,
An edge point at which the frequency distribution of the density change angle is maximum or exceeds a predetermined threshold value is defined as a feature point of the master.
That is, since the data is registered as the external shape data, the edge portion can be satisfactorily extracted regardless of the illumination state with respect to the master, and noise can be reliably removed.

According to a third aspect of the present invention, there is provided a method for registering a master in a search method in which a previously registered pattern (hereinafter, referred to as a master) is searched from an input image. In addition, the image obtained by capturing the master is divided into a plurality of small areas, and a sequence of edge points constituting edges in each of the small areas is detected, and the image density at each of the detected edge points changes. And the frequency distribution of the density change angle is obtained, and the edge points having the same density change angle as the density change angle at which the frequency distribution becomes maximum are extracted. If the distance between them is within a predetermined threshold range, the area is determined to be the same area. If you have a method of registering the edge points present in the partial area as the feature points of the master.

According to a fourth aspect of the present invention, there is provided a method for registering a master for search in a search method for searching a previously registered pattern (hereinafter, referred to as a master) from an input image. In addition, the image obtained by capturing the master is divided into a plurality of small areas, and a sequence of edge points constituting edges in each of the small areas is detected, and the image density at each of the detected edge points changes. And a frequency distribution of the density change angle is obtained, and an edge point having the same density change angle as the density change angle at which the frequency distribution exceeds a predetermined threshold value is extracted. If the distance between the specified edge points is within a predetermined threshold range, the area is determined to be the same area. If more than the constant number is a method for registering the edge points present in the partial area as the feature points of the master.

According to the search master registration method of the present invention, an edge point is detected in each small area of an image obtained by photographing the search master, and the density at the detected edge point is detected. Find the change angle,
An edge point whose frequency distribution of the density change angle is maximum or exceeds a predetermined threshold value is extracted, and an edge point whose distance between the edge points is within a predetermined threshold value is determined to be in the same area. In addition, the total number of edge points is determined for each of the partial areas determined to be within the same area, and it is determined whether the number is greater than the set number that can be determined to be an edge portion. Regardless of the state, the edge portion can be satisfactorily extracted and noise can be reliably removed.

According to a fifth aspect of the present invention, there is provided a method for retrieving an object, comprising: registering a pattern as a master using the registration method according to any one of the first to fourth aspects; A search step of searching for an object having the same shape as the master.

According to the object search method of the present invention, since the object is searched for using the registered pattern, the object search can be performed accurately. Further, according to a method for inspecting an object according to a sixth aspect of the present invention, a step of registering a pattern as a master using the registration method according to any one of the first to fourth aspects, and a step of registering the master registered by the registration step A search step for searching for an object having the same shape as the above, and an inspection for checking the state of the object by comparing a master registered in the registration step with an object having the same shape as the master searched in the search step. This is an inspection method including steps.

According to a seventh aspect of the present invention, an object to be inspected according to the sixth aspect is a printed circuit board, a liquid crystal panel, or a plasma display panel.

According to the method for inspecting an object according to the sixth and seventh aspects, the object is inspected by using the registered pattern without fail, so that the object, for example, a printed circuit board, a liquid crystal panel, or a plasma display panel is inspected. Can be performed accurately.

[0015]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A search master registration method according to a first embodiment of the present invention will be described below with reference to FIGS.

In this master registration method, a master to be compared with a search is photographed, and the photographed image is subjected to image processing to extract feature points. For example, the outline data is used as pattern data of a search master. How to register. As shown in FIG. 1, as the comparison object, that is, the master M, in the present embodiment, for example, a cross-shaped plan view will be described. Noise N
It will be explained as if there is.

First, a schematic configuration of a registration device used for the master registration method will be described. The master registration device includes a photographing camera (for example, a CCD camera) for photographing the master, and an image photographed by the photographing camera. An image processing unit that detects the point sequence as a point sequence, an angle detection unit that calculates a density change angle in a direction in which the image density at each edge point detected by the image processing unit changes maximum, and an angle detection unit that obtains A frequency calculation unit for obtaining a frequency distribution of the density change angle at each of the obtained edge points, and an edge point having the same density change angle as the density change angle at which the frequency distribution obtained by the frequency calculation unit is the maximum is extracted. And a feature registration unit that registers the edge points obtained by the edge point extraction unit as feature points of the master, that is, as shape data. In addition,
The image processing unit, the detection unit, the calculation unit, the extraction unit, and the registration unit are realized by hardware and software provided in a computer device.

Next, a method of registering a master using the above-described master registration device will be described with reference to the flowcharts of FIGS. 2, 3 and 4. First, image data obtained by photographing a master with a CCD camera is taken in by an image processing unit and divided into a plurality of small areas (step 1), and an edge portion is detected as a point sequence for each of the small areas. At the same time, the angle change section determines the density change angle at the edge point detected in each of the small areas (step 2).

The image S of the master M in the first embodiment has a cross shape as shown in FIG. 1, and the entire screen is divided into, for example, six small areas. . In order to identify a small area, the areas are referred to as first to sixth small areas (S1 to S6) in the order of upper left → middle left → lower left → upper right → middle right → lower right.

As for the density change angle, for example,
As shown in FIG. 2, when the density at the edge point in each small area changes from left to right, the angle changes from 0 to 90 degrees, from top to bottom changes from 90 degrees, and changes from right to left. The angle in the case is 180 degrees and the angle in the case of changing from bottom to top is 270 degrees.

Next, the frequency distribution of the density change angle at each of the detected edge points is obtained by the frequency calculation unit (step 3), and the edge point of the density change angle having the largest frequency distribution is extracted. Then, the extracted edge points are registered as characteristic points of the master, that is, as outline data in the characteristic registration unit (step 4).

For example, regarding the first small area S1, FIG.
As shown in the frequency distribution graph of (a1), the maximum value is at the point of 90 degrees, and therefore, (b1) of FIG.
As shown by a solid line, an edge in the horizontal direction is registered as a feature point.

In the second small area S2, (a
As shown in the frequency distribution graph of 2), the location at 0 degrees has the maximum value, and therefore, as shown by the solid line in (b2), the vertical edge portion is registered as a feature point.

For the second small area S2, (b2)
As shown in the figure, noise is shown on the lower left side (indicated by a small circle)
Although N1 exists, since the frequency distribution of this noise is absorbed by other edge portions and only the portion at 0 ° has the maximum value, the noise is not registered as a feature point. That is, the noise has been removed.

In the third small area S3, as in the second small area S2, a noise N2 exists at the lower left.
Again, the frequency distribution of this noise is also absorbed by other edge parts, and only the point at 0 degree has the maximum value,
Therefore, as shown by the solid line in (b3), only the vertical edge portion is registered as a feature point.

The same applies to the fourth to sixth small areas (S4 to S6). The frequency distributions are shown in (a4) to (a6), and the edge portions extracted in (b4) to (b6) are Show. As described above, the edge points are detected in each small region of the image obtained by capturing the search master, the density change angle at each of the detected edge points is determined, and the edge point at which the frequency distribution of the density change angle is the maximum is obtained. Are registered as the feature points of the master, that is, as the outline data, so that a predetermined edge portion can be satisfactorily extracted irrespective of the lighting state of the master, and noise can be reliably removed.

Next, a method of registering a master for search according to the second embodiment of the present invention will be described with reference to FIGS. In the first embodiment, when the edge point is detected, the one corresponding to the density change angle at which the frequency distribution is the maximum is detected. However, in the second embodiment, the frequency distribution is determined by a predetermined value. This is to detect a threshold value or more, and the basic configuration and registration method are the same as those described in the first embodiment. See Figure 6 for the method.
Based on the flow chart of FIG.

That is, as in the first embodiment, first, image data obtained by photographing a master with a CCD camera is taken in by an image processing unit and divided into a plurality of small areas (step 11). ), An edge portion is detected as a point sequence for each small area, and a density change angle at an edge point detected in each of the small areas is obtained by an angle detection unit (step 12). The frequency distribution of the density change angle at the point is obtained by the frequency calculation unit (step 13).

Next, as shown in FIG. 5, in the frequency distribution of the density change angle at the detected edge point, an edge point having a density change angle exceeding a predetermined threshold value (set value) L is determined. The feature points are extracted and registered as feature points (step 14).

For example, regarding the first small area S1, FIG.
As shown in the frequency distribution graph of (a1), when the angle exceeding the threshold value L is seen, the portions at 0 ° and 90 ° correspond to each other. Therefore, as shown by the solid line in (b1) of FIG. The vertical and horizontal edges are registered as feature points.

In the second small area S2, (a
As shown in the frequency distribution graph of 2), 0 degree and 27 degree
The position at 0 ° corresponds to the position, and therefore, as shown by the solid line in (b2), the vertical and horizontal edges are registered as feature points.

Hereinafter, the third small area S3 to the sixth small area S62
The same applies to. As described above, by extracting the edge points having the density change angle whose frequency distribution exceeds the predetermined threshold value, the number of the edge points is increased as compared with the case where the frequency distribution having the maximum density change angle is extracted. Since extraction can be performed and many feature points can be registered, search accuracy as a search master is improved.

Next, a method of registering a master for search according to the third embodiment of the present invention will be described with reference to FIGS. This master registration method basically captures a master serving as a search comparison object and performs image processing on the captured image to detect feature points, similarly to the first embodiment. For example, there is a method of registering the outline data as search master data. Therefore, the comparison object described in the third embodiment is the same as that described in the first embodiment, and a small round noise is placed in the middle left and lower left portions of the captured image. It is what contains.

First, a schematic configuration of a registration device used for the master registration method will be described. The master registration device includes a photographing camera (for example, a CCD camera) for photographing the master, and an image photographed by the photographing camera. An image processing unit that detects the point sequence as a point sequence, an angle detection unit that obtains a density change angle in a direction in which the image density at each edge point detected by the image processing unit changes maximum, and an angle detection unit A frequency calculation unit for obtaining a frequency distribution of the density change angle at each of the obtained edge points, and an edge point having the same density change angle as the density change angle at which the frequency distribution obtained by the frequency calculation unit is the maximum is extracted. An edge point extracting unit, a region determining unit that determines that the distance between these extracted edge points is within a predetermined threshold value (set distance) as the same region, and an area determining unit Number calculating section for calculating the number of edge points in the partial area determined to be the same area, and determining whether or not the total number obtained by the number calculating section is greater than a predetermined threshold value (set number). A number judging section and a feature registering section for registering, as a feature point of the master, an edge point constituting the partial area when the number judging section exceeds a predetermined threshold value, that is, as shape data. .

By the way, the above-mentioned area judging section judges whether or not the distance between the extracted edge points is within a predetermined threshold value because the extracted edge points are in the same area. (Partial area), that is, to determine whether or not it is a series of connected edge points.
Also, in the number determination section, whether the total number of edge points in each partial area is noise compared to a predetermined set number, that is, a minimum number that can be determined to be an edge part, It is determined whether or not.

The image processing unit, the detecting unit, the calculating unit,
The extraction unit, each of the determination units, and the registration unit are realized by software provided in a computer device.

Next, a method of registering a master using the master registration device will be described with reference to the flowcharts of FIGS. First, image data S obtained by photographing the master M with a CCD camera is taken in by an image processing unit and divided into a plurality of small areas (step 21). At the same time, the density change angle at the edge point detected in each of these small areas is obtained by the angle detection unit (step 22).

Of course, as in the first embodiment, the master image data is in the shape of a cross, and the entire screen is divided into six small areas. The first to sixth small areas (S1 to S6) are referred to in the order of → lower left → upper right → middle right → lower right.

The density change angle is also determined by the first
Similarly to the embodiment, the direction of each edge point sequence is expressed by an angle (direction) of a normal with respect to the direction of the edge point sequence (when the point sequence is represented by a straight line, the straight line). And
In addition, when the density at the edge point in each small region changes from left to right, the angle is 0 degree, when the density changes from top to bottom is 90 degrees, and when the density changes from right to left is 180 degrees.
The degree and the angle when changing from bottom to top are 270 degrees.

Next, the frequency distribution of the density change angle at each of the detected edge points is obtained by the frequency calculation unit (step 23), and the density distribution having the maximum frequency distribution is obtained by the edge point extraction unit. Edge points at the changing angles are extracted (step 24).

Next, the area judging section judges whether or not the distance between the extracted edge points is within a predetermined threshold value. Points are considered to be the same area (determined)
(Step 25).

Next, the total number of edge points is obtained and compared with the set number for each partial area determined to be the same area by the number calculating section. It is determined that the partial area is the edge to be sought, and these edge points are registered as master feature points (step 26).

In this embodiment, the second embodiment shown in FIG.
The small area S2 will be described. That is, the second small area S
As shown in the frequency distribution graph of (a2) in FIG. 7, the maximum value is obtained at the position of 0 degrees, and the partial region in which the distance between the edge points is within the predetermined threshold value is shown in FIG. , The first partial region R1, the second partial region R2, and the third partial region R3, and the total number of edge points in each of these partial regions is larger than the set number, the first partial region R1
And the second partial region R2.
1 and R2 are registered as feature points.

Therefore, as shown by the solid line in (b2) of the figure, the vertical edge portion is registered as a feature point. The third partial region R3 smaller than the set number is the noise N1.
Is not registered as a feature point. That is, the noise has been removed.

As for the third small area S3, similarly to the second small area S2, the fourth partial area R4 and the fifth partial area R5 are obtained, and the fourth partial area R4 and the fifth partial area R5 are determined based on the total number of each edge point. Only R4 is registered as a feature point.

The same applies to the first area and the fourth to sixth small areas (S1, S4 to S6).
1), (a4) to (a6) show frequency distributions, and (b)
1) shows edges registered in (b4) to (b6).

As described above, the edge points are detected in each small area of the image obtained by photographing the search master, the density change angle at these detected edge points is obtained, and the frequency distribution of the density change angle is maximized. Edge points which are within a predetermined threshold range are determined to be within the same area, and the total of the edge points is determined for each of the partial areas determined to be within the same area. Since the number is determined and it is determined whether the number is larger than a set number that can be determined to be an edge portion, a predetermined edge portion can be satisfactorily extracted irrespective of an illumination state with respect to the master, and noise can be reliably reduced. Can be removed.

Next, a search master registration method according to the fourth embodiment of the present invention will be described with reference to FIGS. In the third embodiment, when the edge point is detected, the one corresponding to the density change angle having the maximum frequency distribution is detected. However, in the fourth embodiment,
This is to detect the frequency distribution of which is equal to or more than a predetermined threshold value. The basic configuration and registration method are the same as those described in the third embodiment. In addition to the description, the registration method will be described focusing on different points based on the flowcharts of FIGS. 9 and 10.

That is, first, as in the third embodiment, image data S obtained by photographing the master M with a CCD camera is used.
Is captured by the image processing unit and is divided into a plurality of small areas (step 31). Then, an edge portion is detected as a point sequence for each of the small areas, and detected in each of these small areas. After the density change angle at the edge point is obtained by the angle detection unit (step 32), the frequency distribution of the density change angle at each edge point is obtained by the frequency calculation unit (step 33).

Then, as shown in FIG. 9, in the frequency distribution of the density change angle at the detected edge point, an edge point having a density change angle exceeding a predetermined threshold value (set value) L is determined. This is extracted (step 34).

For example, regarding the second small area S1, FIG.
As shown in the frequency distribution graph of (a2), when the threshold value L is exceeded, the points at 0 degree and 270 degree correspond to each other, and therefore, as shown by the solid line in (b2) of FIG. The first to fourth partial areas in the vertical and horizontal directions
The partial regions (R1 to R4) are extracted, and the fourth partial region R4 is determined to be noise by comparing the total number of edge points in each of the partial regions (R1 to R4).

In the third small area S3, (a)
As shown in the frequency distribution graph of 3), the points at 0 ° and 270 ° still apply, and by comparing the total number of edge points, the fifth point is obtained as shown by the solid line in (b3).
The partial region R5 and the sixth partial region R6 are registered as feature points. Note that the seventh partial region R7 is determined to be noise. Remaining small areas S1, S4 to S6
The same applies to.

After the edge points are extracted, as in the third embodiment, the area determining unit determines whether the distance between the extracted edge points is within a predetermined threshold value. Is determined, and for those falling within a predetermined threshold, each edge point is regarded as the same area (step 35).
Next, the total number of edge points is calculated and compared with the set number for each partial area determined to be the same area by the number calculation unit, and if the number is larger than the set number, the partial area is calculated. Edges are determined, and these edge points are registered as master feature points (step 3).
6).

As described above, by extracting the edge points having the density change angle at which the frequency distribution exceeds the predetermined threshold value, the one having the density change angle at which the frequency distribution becomes the maximum is extracted. Compared to the case of the form, more edge points can be extracted, and thus more feature points can be registered, so that the accuracy of the search as a search master is improved.

Here, an embodiment according to the object search method of the present invention will be briefly described. That is, this search method includes a step of registering a predetermined pattern as a master using the master registration method according to each of the above-described embodiments, and a part having the same outer shape (shape) as the master registered in this step. And a search step of searching for an (object). According to this search method, a part can be searched for using a registered pattern without fail, so that the part can be searched accurately.

Further, an embodiment of the object inspection method of the present invention will be briefly described. That is, this inspection method includes a step of registering a predetermined pattern as a master using the master registration method according to each of the above-described embodiments, and a search for an object having the same shape as the master registered in the registration step. A search step, and an inspection step of inspecting the state of the object by comparing the master registered in the registration step with the object having the same shape as the master searched in the search step,
Objects to be inspected include printed circuit boards, liquid crystal panels, and plasma display panels.

According to this inspection method, the component is inspected using the registered pattern without fail.
Inspection of components such as a liquid crystal panel and a plasma display panel can be performed accurately.

[0058]

As described above, according to the method for registering a master for search according to the first or second aspect of the present invention, edge points are detected in each small area of an image obtained by photographing the master for search. A density change angle at the detected edge point is obtained, and an edge point at which the frequency distribution of the density change angle is the maximum or exceeds a predetermined threshold value is registered as a master feature point, that is, as external shape data. As a result, the edge portion can be satisfactorily extracted irrespective of the illumination state of the master, and noise can be reliably removed.

According to the search master registration method of the present invention, an edge point is detected in each small area of an image obtained by photographing the search master, and the detected edge is detected. While obtaining the density change angle at the point, extract the edge point where the frequency distribution of this density change angle is the maximum or exceeds a predetermined threshold,
Further, it is determined that the distance between the edge points is within a predetermined threshold value range within the same region, and the total number of edge points is determined for each of the partial regions determined to be within the same region. Since it is determined whether or not the number is larger than the set number that can be determined as, the edge portion can be satisfactorily extracted regardless of the lighting state with respect to the master,
Noise can be reliably removed.

Further, according to the object search method of the fifth aspect of the present invention, the object can be searched using the registered pattern without fail, so that the object can be searched accurately. Furthermore, according to the object inspection method according to the sixth and seventh aspects of the present invention, since the object is inspected using the registered pattern, the object inspection such as a printed circuit board, a liquid crystal panel, and a plasma display panel is performed. Can be performed accurately.

[Brief description of the drawings]

FIG. 1 shows an image of a master for explaining a master registration method according to a first embodiment of the present invention.

FIG. 2 is a diagram showing a density change angle in a master registration method according to the first embodiment.

FIG. 3 is a diagram illustrating an extraction state of an edge point in the master registration method according to the first embodiment.

FIG. 4 is a flowchart showing a master registration method according to the first embodiment.

FIG. 5 is a diagram illustrating an extraction state of an edge point in a master registration method according to a second embodiment of the present invention.

FIG. 6 is a flowchart showing a master registration method according to the second embodiment.

FIG. 7 is a diagram illustrating an extraction state of an edge point in a master registration method according to a third embodiment of the present invention.

FIG. 8 is a flowchart showing a master registration method according to the third embodiment.

FIG. 9 is a diagram illustrating an extraction state of an edge point in a master registration method according to a fourth embodiment of the present invention.

FIG. 10 is a flowchart showing a master registration method according to the fourth embodiment.

[Explanation of symbols]

 M master S image S1 to S6 Small area R1 to R6 Partial area

 ────────────────────────────────────────────────── ─── Continuing from the front page (72) Noriaki Yukawa, Inventor 1006 Odakadoma, Kadoma City, Osaka Prefecture F-term in Matsushita Electric Industrial Co., Ltd. QQ32 QQ34 QQ51 RR09 SS03 UU05 5B057 AA03 CA08 CA12 CB12 CB20 CC02 CE11 CG09 DA03 DB02 DB09 DC03 DC08 DC16 DC19 DC22 5L096 AA06 BA03 FA03 FA35 FA66 FA67 FA72 GA51

Claims (7)

[Claims] 1. A method of registering a master in a search method for searching a previously registered pattern (hereinafter referred to as a master) from an input image, wherein an image obtained by capturing the master is stored in a plurality of small areas. In addition to the division, the edge point sequence constituting the edge in each of these small areas is detected, the density change angle in the direction in which the image density changes at each of the detected edge points is obtained, and the frequency of this density change angle is calculated. A search master registration method, wherein a distribution is obtained, and an edge point having the same density change angle as the density change angle at which the frequency distribution is maximum is registered as a feature point of the master. 2. A method for registering a master in a search method for searching a pre-registered pattern (hereinafter referred to as a master) from an input image, wherein an image obtained by capturing the master is stored in a plurality of small areas. In addition to dividing, the edge point sequence constituting the edge in each of the small areas is detected, the density change angle in the direction in which the image density changes at each of the detected edge points is obtained, and the frequency of the density change angle is calculated. A search master registration method comprising: obtaining a distribution; and registering, as feature points of a master, edge points having the same density change angle as that of which the frequency distribution exceeds a predetermined threshold value. 3. A method for registering a master in a search method for searching a previously registered pattern (hereinafter, referred to as a master) from an input image, wherein an image obtained by capturing the master is stored in a plurality of small areas. In addition to dividing, the edge point sequence constituting the edge in each of the small areas is detected, the density change angle in the direction in which the image density changes at each of the detected edge points is obtained, and the frequency of the density change angle is calculated. A distribution is obtained, and an edge point having the same density change angle as the density change angle at which the frequency distribution becomes maximum is extracted, and the distance between the extracted edge points within a predetermined threshold range is determined. If the total number of edge points in the partial area determined to be the same area is larger than a predetermined number, the edge points existing in the partial area are determined as master features. A search master registration method characterized by registering as a mark. 4. A method of registering a master in a search method for searching a previously registered pattern (hereinafter, referred to as a master) from an input image, wherein an image obtained by capturing the master is stored in a plurality of small areas. In addition to dividing, the edge point sequence constituting the edge in each of the small areas is detected, the density change angle in the direction in which the image density changes at each of the detected edge points is obtained, and the frequency of the density change angle is calculated. A distribution is obtained, and edge points having the same density change angle as this frequency distribution exceeds a predetermined threshold value are extracted, and the distance between the extracted edge points is within a predetermined threshold range. Are determined to be the same area, and if the total number of edge points in the partial area determined to be the same is larger than a predetermined number, the edge points existing in the partial area A master registration method for searching, characterized by registering as a feature point of a master. 5. A step of registering a pattern as a master using the registration method according to claim 1, and a search step of searching for an object having the same shape as the master registered in the registration step. An object retrieval method characterized by comprising: 6. A step of registering a pattern as a master using the registration method according to claim 1, and a search step of searching for an object having the same shape as the master registered in the registration step. An inspection step of inspecting the state of the object by comparing the master registered in the registration step with an object having the same shape as the master searched in the search step. Method. 7. The method according to claim 6, wherein the object is a printed circuit board, a liquid crystal panel, or a plasma display panel.