JPH11316193A - Method for dividing region of image - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method for dividing a region of image, of which the region can be divided in a short processing time. SOLUTION: All color number obtained from color data of each pixel of a color image on a printed board are replaced by model color candidates with a limited color number by using a limited color displaying algorithm (53). As for the limited color displaying algorithm, a method selecting the model color candidate in order of decreasing a appearance frequency of color, a method of dividing a color space into a predetermined number of subspaces to select the model color candidate of each subspace, or a method converting the three- dimensional color space to one-dimensional arrangement and dividing the one- dimensional pixel array position to set the model color candidate, is used. By applying a rearrangement method to the model color candidate with the limited color number, the region is divided into a desired number of regions (55), and each pixel of the original color image is assigned to one of the divided regions to divide the desired pattern region (57).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to an image area dividing method for dividing an image including a plurality of pattern areas into images of each pattern area.

[0002]

2. Description of the Related Art In a printed circuit board, for example, various pattern areas such as wiring layers for signal lines and power supply lines or pad portions for mounting semiconductor chips are formed on an insulating substrate. In the manufacturing process of such a printed board, a defect inspection of each of these pattern regions is performed. In the defect inspection of the pattern area, the printed circuit board is imaged by the imaging means,
It is preferable to divide each pattern region such as a signal line, a power supply line or a pad portion from the obtained image, and to inspect each pattern region for a defect by applying a different inspection standard to each pattern region.

FIG. 7 is a view showing an image of a main part of a printed board obtained by an image pickup means. The image 10 of the printed circuit board in FIG.
2, pattern regions of a pad region 13 and a silk region 14 are formed, and each pattern region has a different color due to a difference in constituent materials.

FIGS. 8 to 11 show images of a wiring pattern area 11, a resist area 12, a pad area 13 and a silk area 14, respectively, divided from an image 10. FIG.

As a method of dividing the image 10 into images of the respective pattern areas 11 to 14, a conventional rearrangement method (K-me
An method) is generally used. 12 to 15 are explanatory diagrams of the rearrangement method. Here, for convenience of explanation, the pixel values of the image are 3
A method of dividing into two regions (clusters) will be described. 12 to 15, a large number of black points indicate pixel values 15 of pixels distributed in the color space 16.

First, in FIG. 12, the number of areas to be divided, here, "3" is designated. In the color space 16,
Initial value 1 for predicting each center position of the region divided into three
7 is set. Then, each pixel value 15 has three initial values 1
7 is calculated, and all the pixel values 15 are divided into three regions R1 to R3 so that the pixel values 15 close to the same initial value 17 are included. A division line X1 in FIG. 12 indicates a boundary between these three regions R1 to R3.

Next, in FIG. 13, for each of the three divided regions R1 to R3, an average value of the pixel values 15 included in each region is obtained and set as the center value 17a of the next new region. I do. And all pixel values 15
Is determined to be closer to any of the three center values 17a newly set, and each pixel value 15 is again converted into the three regions R1a to R1a.
3a. The dividing line X2 has three regions R1a to R3.
The boundary of a is shown.

Further, in FIG. 14, the same processing as that of FIG. 13 is repeated to divide the pixel value 15 into three new regions R1b to R3b, and calculate the center value 17b of each of the regions R1b to R3b.

Further, in FIG. 15, the above processing is repeated, and a new center value 1 is added to the center value in the previous processing.
The process of dividing all the pixel values 15 into three regions R1i to R3i is completed in a state where the position of 7i has hardly moved. Using the method described above, the image data can be divided into three regions whose pixel values are close to each other.

[0010] By using such a rearrangement method,
Dividing an image of each pattern area from an original image by dividing pixels having similar pixel values into several blocks (areas) based on pixel values of each pixel of the image having a plurality of pattern areas; Can be.

[0011]

However, when the image is divided into regions using the conventional rearrangement method, all the image data, for example, all the pixel values 15 in the examples of FIGS. It is necessary to perform the operation repeatedly. For this reason, when the number of pixels of the image to be processed increases, there is a disadvantage that the time for the repetitive operation increases according to the number of pixels.

An object of the present invention is to provide an image area dividing method capable of performing area division in a short processing time.

[0013]

According to a first aspect of the present invention, there is provided an image area dividing method for dividing an image having a plurality of pattern areas having different colors into images of each pattern area. A method of arranging each pixel constituting an image in a multidimensional space corresponding to a color component, and a predetermined number of colors from a plurality of pixels arranged in the multidimensional space as representative color candidates Selecting, replacing the color of each pixel with one of a predetermined number of representative color candidates, and specifying the number of multi-dimensional spaces designated by using a rearrangement method based on each pixel replaced with the representative color candidate. Divided into regions,
Setting a representative color for each of the divided regions based on the representative color candidates included in the divided region, and determining which of the representative colors set for each of the divided regions is closer to the color of each pixel constituting the image; And generating an image of each pattern area based on the determination result.

In the image area dividing method according to the first invention, the number of colors of each pixel of the image is determined by selecting a representative color candidate from a large number of colors included in the image and replacing the color of each pixel. After the subtraction, the rearrangement method is applied to these pixels. This makes it possible to reduce the number of pixels of a color to be subjected to the repetitive operation by the rearrangement method,
The image can be divided into desired pattern areas at high speed and in a short time.

According to a second aspect of the present invention, there is provided an image area dividing method,
In the configuration of the image area dividing method according to the first invention,
The step of selecting a predetermined number of colors as representative color candidates includes selecting a predetermined number of colors as representative color candidates from the colors of a plurality of pixels forming the image in the order of appearance frequency.

In this case, since the appearance frequency of the pixels in the pattern region constituting the main part of the image is high, it is possible to surely divide these pattern regions from the image.

According to a third aspect of the present invention, there is provided an image area dividing method,
In the configuration of the image area dividing method according to the first invention,
Selecting a predetermined number of colors as representative color candidates, dividing the distribution space into a plurality of subspaces by dividing the distribution space in which pixels are distributed in the multidimensional space along each axis of the multidimensional space; And setting a representative color candidate for each subspace based on the colors of the pixels included in the plurality of subspaces.

In this case, it is possible to reduce the number of colors of a plurality of pixels by dividing a multidimensional space in which a large number of pixels are distributed into a small number of subspaces and setting representative color candidates. . Therefore, the number of pixels of the color to be subjected to the repetition operation by the rearrangement method is reduced, and the image can be divided into a desired pattern area at high speed in a short time.

According to a fourth aspect of the present invention, there is provided an image area dividing method comprising:
In the configuration of the image area dividing method according to the first invention,
The step of selecting a predetermined number of colors as the representative color candidates includes arranging the plurality of pixels in a one-dimensional space by scanning a plurality of pixels arranged in a multidimensional space along a space-filling curve. Creating a frequency histogram indicating the frequency of appearance of pixels corresponding to each of the array positions; and dividing the one-dimensional array position into a predetermined number of regions based on the frequency of appearance of the pixels at each array position in the frequency histogram. And setting a representative color candidate for each of the divided regions based on the colors of the pixels included in the divided regions.

In this case, by scanning the multidimensional space using the space filling curve, it is possible to create a one-dimensional frequency histogram in which pixels having close colors are collected. By dividing the one-dimensional array position and setting representative color candidates, it is possible to reduce the number of colors of a plurality of pixels, and to reduce the number of pixels of a color to be subjected to the repetitive operation by the rearrangement method. Can be done. Therefore, an image can be divided into desired pattern areas at high speed and in a short time.

[0021]

Embodiments of the present invention will be described below. FIG. 1 is a flowchart of an image area dividing method according to an embodiment of the present invention. Hereinafter, a process of dividing the image 10 of the printed circuit board shown in FIG. 7 into images of the respective pattern areas shown in FIGS. 8 to 11 will be described.

An image 10 of the printed circuit board shown in FIG. 7 is composed of a wiring pattern area 11 shown in FIG. 8, a resist area 12 shown in FIG.
0 and the silk region 14 of FIG. Each of these pattern areas 11 to 14
Have different surface materials because they have different constituent materials. Therefore, by determining the color of each of the pattern areas 11 to 14, the pattern area 11 to 1
Divide 4

First, an image 10 of a printed circuit board is input as color data composed of red, green, and blue pixel values for each pixel by image pickup means such as a CCD (charge coupled device) camera (step S1).

Next, let each pixel of the input image 10 be red,
A three-dimensional distribution of pixels is created by arranging in a three-dimensional color space of green and blue (step S2).

Further, based on the color data of each pixel of the input image 10, a limited color display algorithm is applied to represent a large number of colors of the image 10 with a predetermined number of representative color candidates. The following three methods are used to select a representative color candidate using the limited color display algorithm.

The first method is a method in which a frequency histogram representing the frequency of colors appearing in the image 10 is created, and a predetermined number of colors selected in descending order of appearance frequency are selected as representative color candidates. That is, the appearance frequency of the pixel at each arrangement position in the three-dimensional color space is counted. Then, a predetermined number of colors are selected in order from the color of the pixel having the highest appearance frequency, and this is set as a representative color candidate of the image 10. Further, the color data of each pixel is replaced with the color data of the closest representative color candidate from a predetermined number of representative color candidates. Thus, hundreds to thousands of colors included in the image 10 can be replaced with a predetermined number of representative color candidates.

The second method is a method using a color space linear division method. The color space linear division method is a method of linearly dividing only an area where pixels of an original image are distributed in a red, green, and blue space and setting a representative color candidate for each of the partial spaces. FIG. 4 is an explanatory diagram of this color space linear division method.

First, as shown in FIG. 2A, a pixel distribution area along a red axis R in a three-dimensional color space is equally divided into n (integers). Next, as shown in FIG. 2B, the distribution region of the pixels along the green axis G in the three-dimensional color space is equally divided into n (integers).
Further, as shown in FIG. 2C, the distribution area of the pixels along the blue axis B in the three-dimensional color space is equally divided into n (integers). Thereby, the pixel distribution space of the three-dimensional color space is divided into approximately n ³ subspaces 4. Then, a representative value of a color is obtained for each subspace 4, and this color is set as a representative color candidate.

The third method is a method in which a pixel distribution in a three-dimensional color space is made one-dimensional using a space filling curve, and a representative color candidate is selected by integrating adjacent colors. The space filling curve is defined as a curve that passes through a grid point that satisfies a predetermined space only once, and includes, for example, a Peano curve, a Hilbert curve, and the like. FIG. 3 is a pixel distribution diagram in a three-dimensional color space. FIG. 4 is a perspective view of the Peano curve.
FIG. 5 shows a frequency histogram.

In this method, first, a three-dimensional color space is scanned along a Peano curve shown in FIG. 4 and rearranged into a one-dimensional color space as shown in FIG. 5 to create a frequency histogram.
When the frequency histogram is obtained, adjacent colors are integrated along the color array position on the horizontal axis of the frequency histogram, and the color array position is divided so that the appearance frequency of the pixels becomes equal. Then, a representative color candidate is set for each of the divided areas S1 to Si.

By using any of the above three methods, the image 10 consisting of hundreds to thousands of colors can be replaced by a limited number of representative color candidates (step S3).

Note that the number of representative color candidates
Is preferably obtained in advance in accordance with the type of. As a result of the study by the present inventors, when the image 10 is represented by about 700 colors, it was possible to limit the number of representative color candidates to a few percent of the actual number of colors. If the number of the representative color candidates is excessively limited, the area divided from the image 10 becomes inaccurate, or it becomes difficult to divide each pattern area.

Next, the number of divisions (the number of clusters) for dividing the image 10 into a plurality of areas, an initial value used in a rearrangement method described later, and an end condition are input. For example, in the present embodiment, “4” is input as the number of clusters in order to divide the image 10 into the four pattern regions 11 to 14 of FIGS. 6 to 9 (step S4).

Then, based on the input number of clusters, the three-dimensional color space in which the pixels of the image 10 replaced with the representative color candidates are arranged is divided into four regions by using the rearrangement method (K-mean method). Divided into FIG. 6 is an explanatory diagram of area division by the rearrangement method. In FIG. 6, black points schematically show the pixels 6 replaced with the representative color candidates in the three-dimensional color space 5.

First, in FIG. 6A, the number of areas to be divided, here "4", is designated. Also, the input 4
An initial value 7 of the two color data is set at a corresponding position in the color space 5. Then, it is calculated which of the four initial values 7 the representative color candidate pixel 6 is closer to, and all the representative color candidate pixels 6 are set to 4 so as to include the representative color candidate pixel 6 close to the same initial value 7.
Is divided into two regions r1 to r4. A division line Y1 in FIG. 6A indicates a boundary between the four regions r1 to r4.

Next, in FIG. 6B, the divided 4
The average value of the color data of the pixel 6 of the representative color candidate included in each of the regions r1 to r4 is obtained, and this is set as the center value 7a of the next new region. Then, it is determined which of the four center values 7a of the newly set representative color candidate pixels 6 is closer to each other.
Is divided into two regions r1a to r4a. The division line Y2 indicates a boundary between the four regions r1a to r4a.

Further, in FIG. 6 (c), the above processing is repeated, and in a state where the fluctuation amount of the new center value for each processing becomes smaller than the value given in the end condition, the repetitive arithmetic processing is performed. To end. Thus, the pixel of the representative color candidate 6 can be divided into four regions r1i to r4i (step S5).

Further, the final center value 7i in each of the four divided areas is set as a representative color. Thus, four representative colors respectively corresponding to the four divided regions are determined (Step S6).

Then, the color data of each pixel of the image 10 and 4
The color data of the representative color is compared with the color data of the two representative colors, and the color data of the closest representative color is assigned to the color data of each pixel. Thus, the image 10 is divided based on the four color data of each pixel (step S7).

As described above, in the region segmentation of the image according to the present embodiment, the actual colors of the image 10 are replaced with the representative color candidates having a limited number of colors, and the rearrangement method is applied to the representative color candidates to obtain the region. We are doing a split. In the representative color candidates, the number of colors can be limited to about several percent of the total number of colors of the image 10. For this reason, in the process of the rearrangement method, by setting the representative color candidate to be subjected to the repetition operation to a value of about several tens to several hundreds, compared with the conventional case of performing the repetition operation for all the pixels, Area calculation can be performed by repeatedly performing calculations at high speed and in a short time.

In the above-described image region dividing method, when one of the three representative color candidate selection methods is selected and the image region dividing process is performed, if the appropriate region division is not performed, May be configured to perform the image area division process again using another representative color candidate selection method.

Further, in the above embodiment, the image 10
The case where the number of representative color candidates is large with respect to the number of clusters (the number of divisions of the region) has been described. When the area division can be performed, the area division of the image can be performed by the limited color display algorithm used for selecting the representative color candidate. In this case, the process of dividing the representative color candidates by the rearrangement method can be omitted.

By appropriately using the three methods of selecting the representative color candidates according to the characteristics of the image, appropriate color data can be selected as the representative color candidates from the color data of all the pixels. For this reason, it is possible to suppress a decrease in accuracy as compared with the conventional method in which the rearrangement method is applied to all pixels.

In the above-described rearrangement processing, the information of the frequency of appearance of the pixel 6 is added to the pixel 6 of the representative color candidate arranged in the color space 5 so that the center of the four divided areas can be obtained. The accuracy in calculating the value can be improved.

[Brief description of the drawings]

FIG. 1 is a flowchart of an image area dividing method according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram of a color space linear division method.

FIG. 3 is a pixel distribution diagram in a three-dimensional color space.

FIG. 4 is a perspective view of a Peano curve.

FIG. 5 is a diagram showing a one-dimensional frequency histogram.

FIG. 6 is an explanatory diagram of a rearrangement method applied to a representative color candidate.

FIG. 7 is a schematic view of a printed circuit board.

FIG. 8 is a schematic diagram of a wiring pattern area of a printed circuit board.

FIG. 9 is a schematic view of a resist region of a printed circuit board.

FIG. 10 is a schematic view of a pad area of a printed circuit board.

FIG. 11 is a schematic diagram of a silk region of a printed circuit board.

FIG. 12 is an explanatory diagram of a conventional rearrangement method.

FIG. 13 is an explanatory diagram of a conventional rearrangement method.

FIG. 14 is an explanatory diagram of a conventional rearrangement method.

FIG. 15 is an explanatory diagram of a conventional rearrangement method.

[Explanation of symbols]

 5 Color space 6 Representative color candidate pixel 7 Initial value 10 Image 11 Wiring pattern area 12 Resist area 13 Pad area 14 Silk area

Claims (4)

[Claims] 1. An image area dividing method for dividing an image having a plurality of pattern areas having mutually different colors into images of each pattern area, wherein each pixel constituting the image is divided into a multidimensional space corresponding to a color component. And selecting a predetermined number of colors from among the colors of the plurality of pixels arranged in the multidimensional space as representative color candidates; and setting the colors of the pixels to the predetermined number of representative color candidates. Replacing the multi-dimensional space into a specified number of regions using a rearrangement method based on each pixel replaced by the representative color candidate, and the representative color candidate included in the divided region Setting a representative color for each of the divided regions on the basis of: determining which of the representative colors set for each of the divided regions the color of each pixel constituting the image is closer to, and determining Based on the result Area dividing method of the image characterized by comprising the step of generating an image of each pattern region. 2. The step of selecting the predetermined number of colors as representative color candidates includes selecting a predetermined number of colors as representative color candidates from the colors of a plurality of pixels constituting the image in the order of appearance frequency. 2. The method according to claim 1, wherein the image is divided into regions. 3. The step of selecting the predetermined number of colors as representative color candidates includes dividing the distribution space in which pixels are distributed in the multidimensional space along each axis of the multidimensional space, thereby dividing the distribution space. 2. The image according to claim 1, further comprising: dividing the image into a plurality of subspaces; and setting a representative color candidate for each of the subspaces based on colors of pixels included in the plurality of subspaces. Area division method. 4. The step of selecting the predetermined number of colors as representative color candidates includes: scanning a plurality of pixels arranged in the multi-dimensional space along a space filling curve to convert the plurality of pixels into a one-dimensional space. Arranging and generating a frequency histogram indicating the frequency of appearance of pixels corresponding to each array position in the one-dimensional space; and the one-dimensional array position based on the frequency of appearance of pixels at each array position in the frequency histogram. Dividing the image into a predetermined number of regions, and setting a representative color candidate for each of the divided regions based on the colors of the pixels included in the divided regions. How to segment the image.