JPH06174658A - Image filtering method - Google Patents

Abstract

(57) [Summary] [Purpose] Even if the direction of the processing mark changes in the inspection target image area, for each area where the direction changes,
It is an object of the present invention to provide an image filtering method for automatically selecting an effective filtering operator. In a filtering method for removing noise and emphasizing a required image in image processing, a small area cutting step 10 for dividing an inspection object image area obtained by imaging an inspection object 2 into small areas, and the division. An image direction component calculation step 11 for calculating an image direction component of the image of each of the small areas that have been calculated from the history of the inspection object, and an image component of a specific direction to be emphasized or erased is selected from the image direction component calculation result. To emphasize the image component in the selected specific direction, or
An operator selecting step 12 for selecting a filtering operator to be deleted, and the selected filtering
A filtering step 14 for filtering using an operator.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to image filtering which is a pre-process at the time of scratch inspection for inspecting scratches or defects of inspection objects such as various parts and products by image processing.
Filtering according to the directionality of the image of the inspection object
The present invention relates to an image filtering method for automatically selecting an operator.

[0002]

2. Description of the Related Art In recent years, automation of exterior inspection of basic mechanical parts, products, etc. has progressed, and non-contact flaw inspection devices automated by image processing using a TV camera etc. are widely used.
In a flaw inspection using this flaw inspection device, image filtering for enhancing or erasing an image component in a specific direction is performed.

Next, a conventional image filtering method will be described with reference to FIGS. 8 and 9.

In FIG. 8, an illumination device 23 illuminates an inspection object 22 placed on a positioning table 21,
The TV camera 24 supported by the movable TV camera support portion 25 images the inspection object 22. The television camera 24 is controlled by a television camera control circuit 26.

The image signal from the television camera 24 is A /
The data is A / D converted by the D conversion circuit 27, digitized into image data of 0 to 255 (256 gradations) according to the image density, and input to the flaw inspection apparatus.

In the flaw inspection device, a judgment control circuit 28 for controlling the entire device by receiving a command from a main controller or an operation panel, and a window frame control circuit 29 for setting a range to be processed for an input image as a window frame. A filtering circuit 30 for performing image filtering for emphasizing or erasing an image component in a specific direction of the input image data; a binarization circuit 31 for extracting and inspecting a scratch as a binary image;
A pass / fail judgment circuit 32 for judging whether or not the binary image is scratched based on the area of the value image or the like.

FIG. 9 is a flow chart showing the operation of the conventional flaw inspection apparatus for performing image filtering. In the image input process of step # 1, the inspection object placed on the positioning table 21 as shown in FIG. The object 22 and the lighting device 2
3 is illuminated, the television camera 24 supported by the movable television camera support portion 25 takes an image, and the image signal thereof is input to the A / D conversion circuit 27, and the A / D conversion circuit 27 sends the received image. The signal is A / D converted and digitized into image data of 256 gradations according to the image density, and the process proceeds to step # 2.

In the window frame setting step of step # 2, the window frame control circuit 29 shown in FIG. 8 converts the image data sent from the A / D conversion circuit 27 to the filtering circuit 30 into the image data.
Set the window frame and proceed to step # 3.

In the filtering process of step # 3, the filtering circuit 30 shown in FIG. 8 erases or emphasizes an image in a specific direction, as shown in FIG. 7, by various filtering operators M ₁ and M. Image filtering is performed by using a filtering operator designated in advance from ₂ or the like, and the filtering is completed, and thereafter, the binarization circuit 31 and the quality determination circuit 32 inspect.

[0010]

However, as described above, in the configuration of the conventional example in which the image filtering is performed by using the pre-specified filtering operator, the machining mark of the finish machining is left on the machined metal surface. Yes, there is no problem if the processing trace is in a fixed direction in the entire inspection target image area, but if the processing trace changes depending on the region in the inspection target image area, the entire inspection target image area is valid. However, there is a problem that it is difficult to perform image filtering.

The present invention solves the above problems and, even when the direction of a processing mark changes in the image area to be inspected, an effective filtering operator is effective for each area in which the direction changes. It is an object to provide an image filtering method for automatically selecting.

[0012]

In order to solve the above-mentioned problems, an image filtering method of the present invention is a filtering method for removing noise and enhancing a necessary image in image processing, by imaging an inspection object. A small area cutting step of dividing the obtained inspection object image area into small areas; an image direction component calculating step of calculating an image direction component of the image of each of the divided small areas from a history of the inspection object; An operator selecting step of selecting an image component in a specific direction to be emphasized or erased from the direction component calculation result, and selecting a filtering operator to emphasize or erase the image component in the selected specific direction; And a filtering step of filtering using a different filtering operator.

Further, the image filtering method of the present invention is
In order to solve the above-mentioned problem, the image density difference between pixels arranged on the same straight line of each straight line that faces the same direction for each small region between the small region cutting step and the image direction component calculation step. Has a bounded fluctuation amount calculating step of calculating the aggregated image directional strength, and the image direction component calculating step uses the calculated bounded fluctuation amount to specify a specific value to be emphasized or erased. It is preferable to select the directional image component.

[0014]

In the image filtering method of the present invention, the processing history of the inspection object determines the direction of the processing marks in the image area of the inspection object. It is characterized in that the filtering operator is selected for each region in which the directions of the processing marks are different by utilizing the sameness, and has the following configuration and action.

The procedure up to setting the inspection object image area by imaging the inspection object is the same as in the conventional example, but the small area cutout means divides the inspection object image area into small areas and cuts them out. In this way, when the area is divided into small areas, the difference in the direction of the processing mark becomes clear, and it becomes easier to select the filtering operator.

The image direction component calculating step calculates the image direction component of the image of each of the divided small areas from the history of the inspection object, that is, the processing history. Since the image direction components of the image for each small area obtained by this are all the same for the inspection object of the same shape, a filtering operator that erases or emphasizes the image component in that direction according to this image direction component is used. If selected, the entire inspection target image area can be effectively subjected to image filtering.

Therefore, in the next operator selecting step, an appropriate filtering operator is selected based on the calculated image direction component, and effective image filtering is performed in the filtering step.

Further, the image filtering method of the present invention is
A bounded variation amount calculation step is inserted between the small region cutting step and the image direction component calculation step, and in this bounded variation amount calculation step, the same direction of each straight line that faces the same direction for each small region Image density difference between pixels arranged on a straight line is calculated. If the image directional strength is obtained by calculating the aggregated image directional strength, the image directional strength is calculated from the processing history. Instead of the directional component, it can be used as described above.

[0019]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the image filtering method of the present invention will be described with reference to FIGS.

In FIG. 1, an illumination device 3 illuminates an inspection object 2 placed on a positioning table 1, and a television camera 5 supported by a movable television camera support 4 images the inspection object 2. To do. The television camera 5 is controlled by the television camera control circuit 6.

The image signal from the TV camera 5 is A / D.
The data is A / D converted by the conversion circuit 7, digitized into image data of 0 to 255 (256 gradations) according to the image density, and input to the flaw inspection apparatus.

The flaw inspection apparatus sets an image area to be inspected as a judgment control circuit 8 which receives a command from the main controller or an operation panel to control the entire apparatus and a window frame of a range to be processed for an input image. A window frame control circuit 9, a small area cutout circuit 10 for cutting out the inspection target image area into a predetermined small area, and an image direction for calculating an image direction component of an image of each of the divided small areas from a history of the inspection target. From the component calculation circuit 11 and the image direction component calculation result, an image component in a specific direction to be emphasized or erased is selected, and a filtering operator for enhancing or deleting the image component in the selected specific direction is selected. The operator selection circuit 12 and various filtering operators for emphasizing or erasing image components in each direction are stored, and the operator selection circuit 2. An operator storage circuit 13 for sending the filtering operator selected by 2, a filtering circuit 14 for image filtering using the sent filtering operator, and a binary image by binarizing the image filtered image data. Between the binarizing circuit 15, the pass / fail judgment circuit 16 for determining pass / fail based on the binary image, and the pixels arranged on the same straight line in the same direction for each small region. And a bounded fluctuation amount calculation circuit 17 for calculating the image directional strength by calculating and summing the image density differences.

FIG. 2 is a flow chart showing the operation of this embodiment. In the image input step of step # 1, the inspection object 2 placed on the positioning table 1 as shown in FIG. 3 illuminates and movable TV camera support 4
The television camera 5 supported by the camera picks up an image, inputs the image signal into the A / D conversion circuit 7, and the A / D conversion circuit 7 performs A / D conversion on the sent image signal to obtain an image density. By 2
Digitize into image data of 56 gradations, and proceed to step # 2.

In the window frame setting step of step # 2, the window frame control circuit 9 shown in FIG. 1 adds the image data to be inspected to the image data sent from the A / D conversion circuit 7 to the small area cutting circuit 10. Set the window frame of and proceed to step # 3.

In the small area cutting step of step # 3, the small area cutting circuit 10 shown in FIG.
The inspection target image area 40 in the window frame is divided into small areas 41.

By this division, the inspection target image area 40 having the processing mark 42 and the scratch 43 is divided into the small areas 41, and the small area 41 having only the processing mark 42 and the small area having the processing mark 42 and the scratch 43. It is divided into a region 41. If the processing marks 42 of both are erased, the flaw 43 can be easily determined, and even if the flaw 43 is emphasized, the flaw 43 can be easily determined. Then, according to an instruction in advance, the process proceeds to step # 4 or step # 5.

In the bounded fluctuation amount calculation step of step # 4, the bounded fluctuation amount calculation circuit 17 shown in FIG. 1 causes the bounded fluctuation amount calculation circuit 17 shown in FIG. The image directionality intensity obtained by calculating and summing the image density differences between the pixels arranged on the same straight line, that is, the bounded variation amount is calculated.

The bounded fluctuation amount is calculated as follows. As shown in FIG. 5, eight directions centering on the own pixel 44, θ ₀ , θ ₁ , θ ₂ , θ ₃ , θ ₄ , θ ₅ , θ ₆ , and θ _7.
Regarding the pixel density of the own pixel 44 and the own pixel 4
4, the image density difference from the pixel density of the pixel separated by a predetermined distance, in this embodiment, the pixel density of every other pixel is calculated, and the image density difference in the same direction is added as shown in Expression (1). . In this case, the number of direction divisions is adjusted according to the distance between the own pixel 44 and surrounding pixels.

[0029]

[Equation 1]

N is the number of small region pixels in the i direction. M is the number of small region pixels in the j direction.

In the image component direction calculation step of step # 5,
In the first case, the image component direction calculation circuit 11 shown in FIG. 1 calculates the image direction component of the image of each small region divided in step # 3 from the history of the inspection object in the first case, according to a previous instruction. The processing trace 42 becomes the processing trace 42 shown in FIG. 3 in the inspection target image area 40 and the processing trace 42 shown in FIG. 4 in the small region 41, but the processing principle, the processing procedure, and the position of the small region 41. For example, the direction of the processing trace 42 is obtained from the expression of the spiral of the spiral processing trace 42 and the position of the small region 41 for each small region 41.

In the second case, the direction of the processing mark 42 is obtained based on the calculation result of the image directional strength for each small area 41 sent from step # 4 in accordance with a predetermined instruction. . In this case, the image directionality intensity, that is, the bounded variation amount is graphed as shown in FIG. In this case, since the image is bright and there is no change in brightness in the direction of the processing marks and scratches, the bounded fluctuation amount is small.

FIG. 6 is a comparison of the average value * f (θ) of the bounded fluctuation amount in each direction and the bounded fluctuation amount g (θ) in each direction for each small region. ), * F (θ)
And the difference between g (θ) becomes comparison data, and as shown in FIG. 6, at θ ₃ and θ ₄ which are the directions of the scratch 43, the difference exceeds the judgment value ε, and the existence of the scratch 43 is confirmed. It

[0034]

[Equation 2]

Ε: Judgment value Then, the process proceeds to step # 6.

In the operator determining step of step # 6, the operator selecting circuit 12 shown in FIG. 1 selects a filtering operator for erasing the processing trace 42 from the direction of the processing trace 42 obtained in step # 5, and then The operator storage circuit 13 shown in FIG. 1 selects an operator selection circuit 12 from among the stored filtering operators.
Step # 7 for the filtering operator selected by
Send to.

Filtering circuit 14 in step # 7
However, for each small area, filtering is performed using the filtering operator sent, and the process proceeds to step # 8.

In the determination step of step # 8, it is determined whether or not the image filtering is completed for all the small regions 41 in the inspection target image area 40 in the window frame. If so, return to step # 3.

Thereafter, the binarization circuit 15 and the pass / fail judgment circuit 16 shown in FIG.

[0040]

According to the image filtering method of the conventional example, effective image filtering cannot be performed when the direction of the processing mark changes depending on the area in the image area to be inspected. Then, since the directionality of the processing mark is calculated or measured and the filtering operator is selected, there is an effect that the entire inspection target image area can be effectively filtered even when the direction of the processing mark changes depending on the region.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an exemplary embodiment of the present invention.

FIG. 2 is a flowchart of the operation of FIG.

FIG. 3 is an image of an inspection target image area of the present invention, a processing trace of a small area, and a scratch.

FIG. 4 is an image of processing marks and scratches on a small area according to the present invention.

FIG. 5 is a direction division view of a small area according to the present invention.

FIG. 6 is a comparison diagram of bounded fluctuation amounts according to the present invention.

FIG. 7 is a filtering operator.

FIG. 8 is a block diagram showing a configuration of a conventional example.

9 is a flowchart of the operation of FIG.

[Explanation of symbols]

 2 inspection object 10 small area cutout circuit (small area cutout step) 11 image direction component calculation circuit (image direction component calculation step) 12 operator selection circuit (operator selection step) 13 operator storage circuit (operator storage step) 14 filtering circuit ( Filtering process 15 Binarization circuit (binarization process) 16 Pass / fail judgment circuit (Pass / fail judgment process) 17 Bounded fluctuation amount calculation circuit (bounded fluctuation amount calculation process) 40 Inspection target image area 41 Small region 42 Processing trace 43 Scratch 44 own pixel

Claims (2)

[Claims] 1. A filtering method for removing noise and emphasizing a necessary image in image processing, comprising: a small area cutting step of dividing an inspection object image area obtained by imaging an inspection object into small areas; The image direction component calculation step of calculating the image direction component of the image of each of the small areas that are calculated from the history of the inspection object, and the image component of a specific direction to be emphasized or erased is selected from the image direction component calculation result. An image comprising: an operator selecting step of selecting a filtering operator for emphasizing or deleting the selected image component in a specific direction; and a filtering step of filtering using the selected filtering operator. Filtering method. 2. An image density difference between pixels arranged on the same straight line of the respective straight lines facing the same direction for each small region is calculated and aggregated between the small region cutting process and the image direction component calculating process. There is a bounded fluctuation amount calculating step for calculating the image directionality intensity, and the image direction component calculating step uses the calculated bounded fluctuation amount to enhance or erase the image component in a specific direction. The image filtering method according to claim 1, wherein