JPH0797410B2 - Image processing method - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to an image processing method of an appearance inspection apparatus for inspecting an inspection object for defects such as chips, cracks, and stains.

[Prior Art] Conventionally, a defect inspection apparatus for inspecting defects such as cracks, cracks, and stains on an inspected object is image data obtained by A / D converting an image signal from a TV camera for photographing the inspected object. Is stored as original image data in the original image memory, the data read from the original image memory is appropriately arithmetically processed to recognize the contour of the inspection object, and defects such as chipping, cracks, and stains of the inspection object are detected. To recognize.

That is, when a TV camera captures an object to be inspected and captures it as image data, the preset position of the object to be inspected and the actual position of the object to be inspected are often displaced. In such a case, in order to correct the displacement of the inspected object, the defect inspection area is recognized by accurately grasping the outline of the inspected object (or the contour of the reference position or the irregularities of the inspected object). Then, it is determined whether or not a defect exists in this defect inspection area.

For example, as shown in JP-A-62-88946, an image processing method for surely performing contour extraction using a plurality of illumination means has been proposed, and an object to be inspected is illuminated by a plurality of illumination means ( Bright illumination, dark illumination), and logical operation (logical sum) of multiple inspected object images taken by the TV camera in each illumination state to recognize the contour of the inspected object and set the defect inspection area It was like this.

[Problems to be Solved by the Invention] However, in the case of recognizing the contour of the object to be inspected by the conventional image processing method as described above, in the periphery of the edge part of the image of the object to be inspected taken by the TV camera, However, no consideration is given to the presence of foreign matter or noise, and if foreign matter or noise is present around the contour, the contour of the object to be inspected will be accurate no matter how the lighting means and logical operations are changed. There was a problem that it could not be extracted.

The present invention has been made in view of the above points, and an object thereof is to detect the presence of foreign matter or noise around the contour portion of the original image of the inspection object photographed by the TV camera. In addition, the contour line of the inspection object can be searched and the inspection object can be accurately recognized, and the reference point for contour line search of the inspection object can be accurately and easily recognized with one search line. An object of the present invention is to provide a possible image processing method.

[Means for Solving the Problems] According to the image processing method of claim 1 of the present invention, a differential absolute value image is obtained by performing arithmetic processing on an original image of an inspection object photographed by a TV camera,
In the image processing method in which the differential direction value image and the edge image are obtained, the contour line of the inspection object is searched based on each image to recognize the contour of the inspection object, an arbitrary search start point on the edge image From the above, a search line is set in the direction of the object to be inspected, a point where an edge flag exists is determined as a contour inspection target point, and an appropriate local parallel window consisting of N × M pixels is set around the contour inspection target point. The presence or absence of an edge flag is checked for each row of the local parallel window, and when all rows have an edge flag, the local parallel window is set as a contour line inspection target area, and each edge flag of the contour line inspection target area is set. Extract the corresponding differential absolute value from the differential image, and for the extracted differential absolute value that is greater than the preset differential threshold value. When the sum of the differential absolute values is calculated, the contour inspection target point is recognized as a point on the contour line of the inspection object when the sum of the differential absolute values is larger than a preset sum threshold value. It is the one.

Further, the image processing method according to claim 2 sets a search line in the direction of the inspection object from an arbitrary search start point on the differential absolute value image to set a point at which the differential absolute value is larger than a preset differential threshold value. When the contour inspection target point is obtained and the differential direction value corresponding to the contour inspection target point is extracted from the differential direction value image, and the extracted differential direction value exists within a preset differential direction threshold range. The contour inspection target point is recognized as a point on the contour line of the inspection object.

Furthermore, in the image processing method according to claim 3, a search line is set in the direction of the inspection object from an arbitrary search start point on the edge image, a point where an edge flag exists is determined as a contour inspection target point, and The differential absolute value corresponding to the contour inspection target point is extracted from the differential absolute value image, and when the extracted differential absolute value is larger than a preset differential threshold value, the differential direction value corresponding to the contour inspection target point Is extracted from the differential direction value image, and when the differential direction value is within a preset differential direction value range, the contour inspection target point is recognized as a point on the contour line of the inspection object. .

Furthermore, the image processing method according to claim 4 sets a search line in the direction of the inspection object from an arbitrary search start point on the edge image, obtains a point where an edge flag exists, and sets it as a contour inspection target point. The differential absolute value corresponding to the inspection target point is extracted from the differential absolute image value image, and the differential direction value corresponding to the contour inspection target point when the extracted differential absolute value is larger than a preset differential threshold value. Is extracted from the differential direction value image, and the density inspection points are the original density values of the density inspection points at the points distant from the point corresponding to the contour inspection target point on the original image by n pixels in the direction perpendicular to the differential direction value. The contour inspection target point extracted from the image, when the density of one density inspection point is higher than the density threshold for the inspection object and the density of the other density inspection point is less than the density threshold for the background. To It is obtained so as to recognize a point on the contour line of the test object.

[Operation] The present invention is configured as described above, and in the image processing method according to claim 1, image processing for searching a point on the contour line of the inspection object to recognize the contour of the inspection object. In the method, the point where the edge flag exists is found by the search line set in the direction of the object to be inspected on the edge image, and the presence or absence of the edge flag is checked for each row of the local parallel window set around this contour inspection target point. However, if all the rows have edge flags, the local parallel window is set as the contour line inspection target area, and the preset differential of the differential absolute value corresponding to each edge flag of the contour line inspection target area is performed. If the sum of differential absolute values is obtained by adding those larger than a threshold value, the contour inspection is performed when the sum of the differential absolute values is larger than a preset total threshold value. The elephant point has to recognize a point on the contour line of the object to be inspected, even when the foreign matter or noise is present in the contour portion surrounding the original image of the object taken by the TV camera,
It is possible to accurately recognize an object to be inspected by searching the contour line of the object to be inspected, and also to accurately and easily recognize a reference point for contour line search of the object to be inspected with one search line. It has become.

Further, the image processing method according to claim 2 obtains a point whose differential absolute value is larger than a preset differential threshold value by a search line set on the differential absolute value image, and differentiates the contour inspection target point. When the direction value is within a preset differential direction threshold value range, the contour inspection target point is recognized as a point on the contour line of the inspection object. The reference point can be recognized accurately and easily.

Furthermore, the image processing method according to claim 3 obtains a point where an edge flag exists by a search line set on the edge image, and a differential absolute value corresponding to the contour inspection target point is a preset differential threshold. When the differential direction value corresponding to the point is larger than the value and is within the preset differential direction value range, the contour inspection target point is recognized as a point on the contour line of the inspection object. Therefore, the reference point for contour line search of the object to be inspected can be accurately and easily recognized.

Furthermore, the image processing method according to claim 4 sets a search line in the direction of the object to be inspected from an arbitrary search start point on the edge image, obtains a point where an edge flag exists, and sets it as a contour inspection target point. The differential absolute value corresponding to the target point is extracted from the differential absolute value image, and the differential direction value corresponding to the contour inspection target point is differentiated when the extracted differential absolute value is larger than a preset differential threshold value. From the original image, the density of both density inspection points is extracted from the direction value image, and the density inspection points are points separated from the point corresponding to the contour inspection target point on the original image by n pixels in the direction perpendicular to the differential direction value. When the density of one of the density inspection points is larger than the density threshold for the object to be inspected and the density of the other density inspection point is smaller than the density threshold for the background, the contour inspection target point is extracted. Inspection It is obtained by to recognize the point on the contour line of the object, thereby making it possible to accurately recognize and easily a reference point of the contour line searching of the object.

[Embodiment] FIGS. 1 and 2 show a schematic structure of an appearance inspection apparatus according to the present invention. A TV camera 1 for photographing an object to be inspected O illuminated obliquely by an illumination lamp La, A / D converter 2 for A / D converting the image signal output from the TV camera 1
A pre-processing circuit 3 for pre-processing A / D converted image data, an original image memory 4 for storing the pre-processed image data as an original image of the inspection object O, and each pixel of the original image Differential absolute value image memory 5 for storing the calculated differential absolute value, differential direction value image memory 6 for storing the calculated differential direction value, and a change point of the density (brightness) of the original image as a line drawing An edge image memory 7 that stores the edge flag, and an original image that is arithmetically processed to store predetermined data in each of the memories 5 to 7, and an object to be inspected based on the data stored in each of the memories 4 to 7. It is formed by a microprocessor (or a microcomputer) 8 that searches the contour line of O, recognizes the contour of the inspection object O, and checks whether the inspection object O has a defect.

Hereinafter, the outline recognition and the defect detection of the inspection object O of the above-described appearance inspection device will be specifically described.

3 and 4 are operation explanatory views. First, the microprocessor 8 calculates an original image f ₁ of the inspection object O as shown in FIG. 3 (a) taken by the television camera TV1. It is necessary to process and convert it into an edge image f ₄ as shown in FIG. 3 (b), and this processing is performed as follows.

First, the original image f ₁ obtained by imaging the spatial region including the inspection object O is a grayscale image, and as shown in FIG. 3A, the inspection object O, the defect X ₁ , and the foreign matter X ₁ . _The image includes ₂ . Here, the density of each pixel is represented by, for example, 8 bits and is set to 256 gradations. The process of extracting the edge such as the outline of the inspection object O from the grayscale image is based on the idea that "the edge portion corresponds to a portion where the density change is large". Therefore, it is general to extract the edge by differentiating the density. The differentiation processing is performed by dividing the grayscale image into local parallel windows W of 3 × 3 pixels, as shown in FIG. That is, the pixel E of interest and the eight pixels A to D, F to I around the pixel E
Form a local parallel window W with, and a vertical density change ΔV and a horizontal density change H of the density of the pixels A to I in the local parallel window W are obtained by the following equation, and ΔV = (A + B + C) − ( G + H + I) ΔH = (A + D + G)-(C + F + I) Further, the differential absolute value | e _E | and the differential direction value ∠e _E are obtained by the following equations.

However, A to I indicate the densities of the corresponding pixels. By performing the above calculation for all the pixels of the original image f ₁ , the portion where the density change is large such that the contour of the object to be inspected O, the defect X ₁ , the foreign material X _{2 and the} like are present, and the direction of the change Can be extracted, and the differential absolute value image f ₂ (6 bits),
The differential direction value image f ₃ (4 bits) is stored in the differential absolute value image memory 5 and the differential direction value image memory 6, respectively.

Next, thinning processing is performed. The thinning process is performed by paying attention to the fact that the larger the differential absolute value, the larger the density change. That is, an edge having a width of one pixel is extracted by comparing the differential absolute value of each pixel with the differential absolute value of surrounding pixels and connecting those that are larger than the surrounding pixels. . That is, the position of each pixel on the screen is represented by XY coordinates, and the differential absolute value is Z.
If the axis is taken, a curved surface representing the differential absolute value will be formed, and the thinning processing corresponds to obtaining the ridge line on this curved surface. At this stage, an edge due to noise or the like is also included, so a threshold value is appropriately set, and only a value equal to or higher than the threshold value is used to remove the noise component.

The edge image f ₄ obtained by the thinning process is likely to be a discontinuous line when the contrast of the original image f ₁ is insufficient or when there is a lot of noise. Therefore, edge extension processing is performed. The edge extension processing starts from the end point of the discontinuity line, compares the pixel of interest with the surrounding pixels, extends the edge in the direction in which the evaluation function f (e _J ) represented by This continues until it hits the end of the line.

Here, e _O is the differential data of the central pixel, e _J is the differential data of the adjacent pixel, and J = 1, 2, ...

With the above processing, as shown in FIG. 3 (b), 1-bit object to be inspected O, defect X _1, outline l ₀ to l ₂ such foreign matter X ₂ is a pattern of closed curve ( "1" Is referred to as an edge flag), and an edge image f ₄ is obtained and stored in the edge image memory 7.

By the way, in the above-described appearance inspection apparatus, it is necessary to accurately set the reference point Q that is the starting point of the contour line search as described above. Hereinafter, the image processing method for setting the reference point Q according to the present invention will be described. Will be described.

FIG. 5 is a flow chart of the image processing method according to claim 1,
6 to 9 are explanatory diagrams of the operation, FIG. 6 is an original image photographed by the TV camera 1, and each image memory 4 to 7 has pixels (xn) as shown in FIG. , yn) corresponding to the original image data f ₁ (xn, yn) and the differential absolute value image data f ₂ (xn, yn)
yn), differential direction value image data f ₃ (xn, yn), and edge image data f ₄ (xn, yn) are stored in correspondence (with the same address code).

First, any search starting point on the edge image f ₄ P (x
p, yp) and set the search start point P (xp, y
From p), a search line a is set in the direction of the object to be inspected, and a point at which an edge flag (f ₄ (x, y) = 1) exists is obtained and set as a contour inspection target point Q (xq, yq). Next, as shown in FIG. 8, centering on the contour inspection target point Q (xq, yq), N ×
Sets the appropriate local parallel window W formed of (7 × 5 pixels in the example) M pixels, as shown in FIG. 9, the scan for each line of the local parallel window W b ₁ ~
b ₇ Check for edge flags. here,
As shown in FIG. 8A, when all the rows have edge flags, the local parallel window W is set as the contour line inspection target area. As shown in FIG. 8 (b), even if there is an edge flag on the search line a but there is no edge flag in all the rows of the local parallel window W, the above processing is repeated to repeat the contour line. Search for the area to be inspected.

Next, the differential absolute value f ₂ (xq, yq) corresponding to each edge flag of the contour line inspection target area obtained as described above
Is extracted from the differential absolute value image f ₂ and the extracted differential absolute value f ₂ (xq, yq) is larger than a preset differential threshold L ₂ (f ₂ (xq, yq)> L ₂ ) is added to obtain the sum S of differential absolute values.

Next, when the sum S of the differential absolute values is larger than a preset sum threshold L ₂ s (S> L ₂ s), the contour inspection target point Q (xq, yq) is inspected. It is determined to be a point on the contour line l _{0 of} O. The sum S of the differential absolute values is the sum threshold L ₂
If it is smaller than s, it is determined that the contour inspection target point Q (xq, yq) is not a point on the contour line of the inspection object O, and the above-described processing is repeated to search for the reference point. If there is no point satisfying the reference point condition within the preset search range, it is determined that the inspection object O does not exist.

Here, in the present embodiment, in order to accurately extract the reference point on the contour line l ₀ , firstly, focusing on the feature that the edge of the inspection object O is linear and continuous, It is determined by the local parallel window W whether or not the edge flag exists in all the rows, and the contour line inspection area based on the edge of the foreign matter and noise X ₂ is not set. Next, the differential absolute value of the edge of the foreign matter and noise X _{2 is} the object O to be inspected.
Focusing on a point that is generally smaller than the differential absolute value of the edge of, the differential threshold L ₂ is set in the middle, so that the differential absolute value of the edge of the foreign matter and noise X ₂ is ignored. There is. Furthermore, focusing on the fact that the number of differential absolute values of the edge of the foreign matter and noise X ₂ exceeds the differential threshold L ₂ is small, and the sum S thereof is small, the differential absolute value larger than the differential threshold L ₂ is obtained. The condition that the total sum S of the values must be larger than the total sum threshold L ₂ s is added. By determining the points satisfying the above three conditions as the points on the contour line of the inspection object O, The reference point for contour line search can be accurately and easily recognized with one search line a without being affected by foreign matter and noise X ₂ .

FIG. 10 is a flowchart of the image processing method of claim 2,
FIG. 11 is an operation explanatory view of the same, as shown in FIG. 11, any search starting point on the differential absolute value image f ₂ P (f ₂
(Xp, yp)), a search line a is set in the direction of the inspection object, and as shown in FIG. 15, each pixel Ri on the search line a is set.
At (i = 1,2,3, ...), a point whose differential absolute value f ₂ (xp, yp) is larger than a preset differential threshold L ₂ is obtained, and the contour inspection target point Q (xq, yq). Next, contour inspection point Q (xq, yq) differential direction value f ₃ (xq, yq) corresponding to extracts from differentiated direction value image f _3, extracted differentiated direction value f ₃ (xq, y
q) is within a preset differential direction threshold [L ₃ a, L ₃ b] range (L ₃ a <f ₃ (xq, yq) <L
₃ b) Then, when this judgment condition is satisfied, it is judged that this contour inspection target point Q (xq, yq) is a point on the contour line l ₀ of the inspection object O (the address code is stored). If the above determination conditions are not satisfied, the contour inspection target point Q (xq, yq)
Is not a point on the contour line l ₀ , the above-described reference point search processing is repeated for the next pixel Ri on the next search line a, and the determination condition is satisfied within the preset search range. If there is no point, it is determined that the inspection object O does not exist. The differential direction threshold range is set by accurately examining the differential direction value of the contour line of the inspection object O and narrowing down the differential direction threshold range.

FIG. 12 is a flow chart showing the operation of the image processing method of claim 3. First, the search line a is set from the arbitrary search start point P on the edge image f ₄ toward the inspected object, and the edge flag (f A point at which ₄ (x, y) = 1) exists is determined as a contour inspection target point Q (xq, yq). Next, the differential absolute value f ₂ (xq, yq) corresponding to the contour inspection target point Q (xq, yq) is extracted from the differential absolute value image f ₂ , and the extracted differential absolute value f ₂ (xq, yq) is larger than a preset differential threshold L ₂ and the differential direction value f ₃ (xq, yq) corresponding to the contour inspection target point Q (xq, yq) is extracted from the differential direction value image f _3. Then, when the differential direction value f ₃ (xq, yq) is within the preset differential direction value range (L ₃ a> f ₃ (xq, yq)> L ₃ b),
This contour inspection target point Q (xq, yq) is the contour line l of the object O to be inspected.
Judge as a point above ₀ (store address code). In addition,
If the above two determination conditions are not satisfied at the same time, it is determined that the point is not on the contour line l ₀ , the above-described search processing is repeated, and there must be a point that satisfies the determination condition within the preset search range. For example, it is determined that the inspection object O does not exist.

FIG. 13 (a) is a diagram for specifically explaining the operation of the above-described embodiment, in which a white object to be inspected O exists between the black backgrounds B ₁ and B ₂ , and foreign matter or foreign matter is present on the background B ₂ side. There are many noises X ₂ ,
When recognizing the contour line l ₀₂ by searching, if the search point is set to the B ₂ side, it is easy to misrecognize the foreign matter and the noise X ₂ , so it is desirable to set the search start point P to the background B ₁ side. . In this way, the search start point is set on the background B ₁ side, and the edges Q ₁ (xq ₁ , yq ₁ ) and Q ₂ (xq ₂ , y of the object to be inspected O are further set.
The differential absolute value of q ₂ ) is sufficiently larger than the differential absolute value of foreign matter and noise X ₂ , and exceeds the differential threshold L ₂ , and the first judgment condition is satisfied. Now, the differential direction value f ₃ (xq ₁ , yq ₁ ) corresponding to the point Q ₁ (xq ₁ , yq ₁ ) on the contour line l ₀₁
Is “4” and the differential direction value f ₃ (xq ₂ , yq ₂ ) corresponding to the point Q ₂ (xq ₂ , yq ₂ ) on the contour line l ₀₂ is “12”, the second
If the differential direction value range that is the judgment condition of is set to “10 to 14”, the point Q ₁ (xq ₁ , yq ₁ ) on the contour line l ₀₁ does not satisfy the second judgment condition. Contour line l that satisfies the judgment condition
The point Q ₂ (xq ₂ , yq ₂ ) on ₀₂ is determined as the reference point for contour line search of the inspection object O, and an appropriate reference point can be recognized by considering the differential direction value. In this example,
As shown in FIG. 13 (b), it is also effective when recognizing a point on the inner contour line l ₀₂ .

FIG. 14 is a flowchart of the image processing method of claim 4, and similarly to the case of claim 3, the differential direction value f ₃ (xq of the point Q (xq, yq) where the edge flag exists on the search line a. ,
yq) is extracted, and as shown in FIG. 15, from the point corresponding to the contour inspection target point Q (xq, yq) on the original image f ₁ , the differential direction value f ₃
Density inspection points R ₁ (xq−n, yq−n) and R ₂ (xq + n, xq + n) are points that are n pixels apart from each other in the vertical direction with respect to (xq, yq).
, And these density inspection points R ₁ (xq−n, yq−
n), the concentration of R ₂ (xq ＋ n, xq ＋ n) is extracted, and one of the concentration inspection points R ₁ (xq－n, yq－n) or R ₂ (xq ＋ n, xq ＋ n)
Is greater than the concentration threshold L ₁₀ for the object to be inspected, and the concentration of the other concentration inspection point R ₂ (xq + n, yq + n) or R ₁ (xq-n, xq-n) is the concentration for the background B. When it is smaller than the threshold value L _1B , the contour inspection target point Q (xq,
yq) is determined as a point on the contour line, the address code is stored, the contour line search is started by recognizing it as a reference point for the contour line search. In addition, when the above determination condition is not satisfied, it is determined that the point is not on the contour line l ₀ , the reference point search process is repeated, and when there is no point within the search range that satisfies the determination condition, the object to be inspected It is determined that O does not exist.

Now, assuming that the density of the object O to be inspected is sufficiently higher than the density of the background B and the density of the foreign matter and the noise X ₂ is in the middle thereof, the density threshold L _1O is set sufficiently high, and
If the density threshold L _1B is set to be sufficiently small, the edge of the foreign matter and the noise X ₂ will not be erroneously determined as the contour line of the inspection object O, and the reference point on the contour line of the inspection object O will be accurately recognized. You can do it. Since the density of the foreign matter and the noise X ₂ is generally sufficiently lower than the density of the inspection object O and sufficiently higher than the density of the background B, both density thresholds L _1O and L _1B can be relatively easily set. It can be set.

EFFECTS OF THE INVENTION The present invention is configured as described above, and even when foreign matter or noise is present around the contour portion of the original image of the inspection object captured by the TV camera, There is an effect that a contour line can be searched for and an object to be inspected can be accurately recognized, and a reference point for contour line search of an object to be inspected can be accurately and easily recognized with one search line.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram of an appearance inspection device according to the present invention, and FIG.
FIG. 3 is a block circuit diagram of the above, FIGS. 3 and 4 are operation explanatory diagrams of the same, and FIGS. 3A to 3C are original images.
An edge image, an explanatory view showing an example of an edge image of a defective portion,
FIG. 4 is an explanatory view showing a local parallel window, and FIGS. 5 to 9 are explanatory views of the operation of the invention of claim 1, FIG. 10 and FIG.
FIG. 11 is an operation explanatory view showing one embodiment of the invention of claim 2,
12 and 13 are explanatory views of the operation of the invention of claim 3, and FIG.
FIG. 15 and FIG. 15 are operation explanatory views of the invention of claim 4. 1 is a TV camera, 2 is an A / D converter, 3 is a pre-processing circuit, 4 is an original image memory, 5 is a differential absolute value image memory, 6 is a differential direction value image memory, 7 is an edge image memory, and 8 is a micro. It is a processor.

Continuation of the front page (56) References JP 61-9769 (JP, A) JP 61-126455 (JP, A) JP 61-9775 (JP, A)

Claims (4)

[Claims] 1. An original image of an object to be inspected taken by a TV camera is arithmetically processed to obtain a differential absolute value image, a differential direction value image and an edge image, and a contour line of the object to be inspected is searched based on each image. In the image processing method for recognizing the contour of the inspected object, a search line is set in the direction of the inspected object from an arbitrary search start point on the edge image to obtain a point where the edge flag exists, and the contour inspection is performed. As a target point, an appropriate local parallel window consisting of N × M pixels centering on the contour inspection target point is set, and the presence or absence of an edge flag is checked for each row of the local parallel window to check the edge flag for all rows. If there is, the local parallel window is set as the contour inspection target area, and the differential absolute value corresponding to each edge flag of the contour inspection target area is extracted from the differential image and extracted. Of the differential absolute values that have been issued, those that are larger than the preset differential threshold value are added to obtain the total sum of differential absolute values, and the total sum of the differential absolute values is greater than the preset total threshold value. At this time, the contour inspection target point is recognized as a point on the contour line of the object to be inspected. 2. A differential absolute value image, a differential direction value image, and an edge image are obtained by arithmetically processing an original image of the inspection object photographed by a TV camera, and a contour line of the inspection object is searched based on each image. In the image processing method in which the contour of the inspected object is recognized, a differential absolute value is preset by setting a search line in the direction of the inspected object from an arbitrary search start point on the differential absolute value image. A point larger than the threshold value is obtained as a contour inspection target point, a differential direction value corresponding to the contour inspection target point is extracted from the differential direction value image, and the extracted differential direction value is a preset differential direction threshold value. An image processing method, characterized in that, when it exists within a range, the contour inspection target point is recognized as a point on the contour line of the inspection object. 3. An original image of an object to be inspected taken by a TV camera is arithmetically processed to obtain a differential absolute value image, a differential direction value image and an edge image, and a contour line of the object to be inspected is searched based on each image. In the image processing method for recognizing the contour of the inspected object, a search line is set in the direction of the inspected object from an arbitrary search start point on the edge image to obtain a point where the edge flag exists, and the contour inspection is performed. As the target point, the differential absolute value corresponding to the contour inspection target point is extracted from the differential absolute value image, and when the extracted differential absolute value is larger than a preset differential threshold value, the contour inspection target point is set. A corresponding differential direction value is extracted from the differential direction value image, and when the differential direction value is within a preset differential direction value range, the contour inspection target point is recognized as a point on the contour line of the inspection object. Characterized by Image processing method for. 4. A differential absolute value image, a differential direction value image, and an edge image are obtained by arithmetically processing an original image of an object to be inspected taken by a TV camera, and a contour line of the object to be inspected is searched based on each image. In the image processing method for recognizing the contour of the inspected object, a search line is set in the direction of the inspected object from an arbitrary search start point on the edge image to obtain a point where the edge flag exists, and the contour inspection is performed. As the target point, the differential absolute value corresponding to the contour inspection target point is extracted from the differential absolute value image, and when the extracted differential absolute value is larger than a preset differential threshold value, the contour inspection target point is set. A corresponding differential direction value is extracted from the differential direction value image, and both density inspections are performed by using a point that is n pixels away from the point corresponding to the contour inspection target point on the original image in the direction perpendicular to the differential direction value as the density inspection point. Based on the density of points When the density of one density inspection point is larger than the density threshold for the object to be inspected and the density of the other density inspection point is smaller than the density threshold for the background, the contour inspection target point is extracted from the image. An image processing method characterized in that it is recognized as a point on a contour line of an object to be inspected.