JPH02277186A - Binarizing device - Google Patents

Abstract

PURPOSE:To turn an unclear picture into a clear one by preparing plural deciding means and binarizing the picture after deciding whether a noted subwindow is equal to a line graphic part or not based on the outputs of those deciding means. CONSTITUTION:A line picture 1 is photographed by an image sensor 2 and converted into a gradation picture by an A/D converter 3. This gradation picture is stored in a frame memory 4 and then read out for each picture element in the same scanning order as the sensor 2. Thus a mobile subwindow is formed by a 2-dimensional local memory 5. Then a local averaging part 6 calculates the average density value Aij and stores this value in an average density value memory 12 for the local area of a noted picture element. Then a threshold value detecting part 7 calculates the density difference between the noted picture element and a nearby picture element and checks whether this difference is kept within a range of the set value or not. If so, the average value is obtained among all picture elements and defined as the threshold value PA. This value PA is compared with the allowable density difference G0 via a comparator 10. When the value PA is smaller than the difference G0, the even density is secured among local areas. Thus a clear binarized picture is obtained.

Description

Detailed Description of the Invention [Field of Industrial Application] The present invention is applicable to optical character reading devices 1ffi (OCR), character and graphic input devices, etc., which are used to capture line graphic images including laser-engraved characters, stamped characters, etc., and objects. The present invention relates to a binarization device that binarizes linear images obtained by imaging surface scratches, cracks, cracks, cracks, dust, etc. for image identification.

FIG. 17 is an explanatory diagram showing an example of a target image to be binarized. The figure shows scratches 2 on the surface of the inspection object 202.
03 is generated, and 201 is a screen. The flaw 203 is characterized by being elongated and shaped like a line segment. Examples of scratches and dirt having such characteristics include a wide variety of things, such as human hair attached to the surface of food or tableware, cracks in castings, and scratches on metal surfaces.

[Conventional technology]

Generally, it is difficult to obtain a clear binary image of low-contrast laser-engraved characters, stamped characters, etc. due to uneven brightness and dirt in the background.

Therefore, instead of setting the binarization level to a fixed value, the applicant took into consideration the fluctuation of the background level of the image, for example, and
A method has been proposed for automatically determining the valuation level so that it is always optimal (see, for example, Japanese Patent Application No. 158032/1982).

[Problem to be solved by the invention]

However, in such a case, if the line figure is binarized at a darkness value level close to the density of the background, the line figure part will appear thicker, but the uneven brightness and dirt in the background will also be binarized, making it difficult to decipher. If you binarize at a darkness value level that does not cause uneven brightness or smudges in the background, the line characters will become blurred.
In other words, since none of the conventional methods distinguishes between line characters and dirt, there is a problem in that if the density values are the same, they are binarized in the same way.

Therefore, the present invention distinguishes smudges that are wider than line figures, density unevenness in the background, etc. from line figures, so that even images with low contrast character lines or linear scratches can be converted into clear binarized images. The object of the present invention is to provide a binarization device that can perform the following operations.

[Means to solve the problem]

In order to achieve the above object, in the present invention, image data obtained by two-dimensionally imaging a background and a line figure drawn thereon is converted into pixels and captured into a memory, and then the pixels are read out from the memory. In a binarization device that binarizes each pixel depending on whether it is a pixel that corresponds to a background part or a pixel that corresponds to a line figure part, At the same time, a sub-window of interest is set for a pixel of interest that is the target of conversion, or a pixel of interest and its surrounding pixels are set as a sub-window. sub-window reading means for setting a plurality of sub-windows each consisting of the individual pixel and its surrounding pixels, reading each of the sub-windows sequentially from the memory; and calculating means for calculating an average density value for each of the sub-windows; an average value calculation means for calculating an average value of difference values between a pixel of interest and its neighboring pixels for each subwindow within a predetermined range; and an average value calculation means based on each average value of the subwindow of interest and its surrounding subwindows by the average value calculation means. a first determining means for determining whether the region within the corresponding window group is a uniform density region; and a second determining means for determining whether each peripheral sub-window corresponds to a background portion based on the average density value of the sub-window of interest and its peripheral sub-windows.
Based on the outputs from the first and second determining means, it is determined whether or not the sub-window of interest corresponds to a line figure portion, and binarization is performed.

Furthermore, in such a binarization apparatus, the coordinates of peripheral sub-windows for a sub-window of interest are selected such that all peripheral sub-windows are located at the same distance from the sub-window of interest.

[Effect]

Hereinafter, the determination principle of binarization according to the present invention will be explained as an operation.

FIG. 18 is an explanatory diagram showing how to select a pixel of interest to be binarized and its surrounding pixels from among the pixels constituting the screen. That is, the neighboring pixels P1 to P1 to the pixel of interest PO
P8 is determined at the following coordinate position.

PO (x, y), PL (x-d, y-d).

P2(x, Y-d), PG(x1d, y-d)
.

P4 (x-d, y), P5 (x+d, y).

P6 (x-d, y 10 d), P7 (x, y + d)
.

P8 (x+d, y+d) FIG. 18 is an example where d=3 (number of pixels). d is a constant determined by the width of the line in the target image.

FIG. 19 is an explanatory diagram showing a specific example of how to select a pixel of interest and peripheral pixels according to the width of a line in a target image.

In the same figure (a), the width of the line of the target image R in the vertical direction is exactly 1.
A line image slightly larger than pixels is shown, and the same figure (
b) shows a line image in which the line width of the target image R in the diagonal direction is a little more than 1 pixel. On the other hand, if one pixel is placed around it, PI, P2. ~P7. P
If peripheral pixels such as 8 are selected, the below-described operation for binarizing the pixel of interest PO will be successful.

Note that if the width of the line in the image R is so wide that it spans a plurality of pixels, a cluster of pixels is considered as a sub-window of interest, and surrounding sub-windows are selected accordingly.

Here, to explain the conditions for binarizing the pixel of interest PO on the image, PO<C1n ((PI-PO>C2n P8-PO
>C2)tJ(P4-PO>C2n P5-PO>C
2) U(P6-PO>C2n PG-PO>C2)U
(P7-PO>C2n p2-po>C2)) When the following conditions hold, the pixel of interest is set to logic 1, and when the condition does not hold, it is set to logic 0.

Here, PO to P8 give the density value of each pixel, and C1 and C2 are constants. Constant QCI.

C2 is comprehensively determined by the surface condition of the inspection image, the degree of illumination, the characteristics of scratches, etc.

Here, the conditions for the binarization described above are explained as follows.

In the above conditional expression for binarization, n means AND, and U means OR. Therefore, when the condition PO<C1 (that is, the condition that the density of the pixel of interest is lower than the constant 01 and black) and any one of the following conditions are satisfied, the pixel of interest is It is set to l, and 0 otherwise.

The following conditions are: PL-PO> C2n P8-PO> C2 (that is, the density difference between the pixel of interest PO and the surrounding pixel P1 is whiter than the constant 02; in other words, if the pixel of interest PO is black, the surrounding pixel Pl If the pixel of interest PO is black, the surrounding pixel P8 is also black. Here, the surrounding pixels P1 and P8 hold together.
Note that they are diagonally opposite to the left with the pixel of interest PO as the center), P4-PO>C2n P5-PO>C2 (if the pixel of interest PO is black, the surrounding pixel P4 is also black) Similarly, if the pixel of interest PO is black, the surrounding pixel P5
The condition that both are also black holds true at the same time.
Note that the peripheral pixels P4 and P5 are opposite to each other in the horizontal direction with the pixel of interest PO as the center), P6-PO>C2n PG-PO>C2 (if the pixel of interest PO is black, the surrounding pixel P6 Similarly, if the pixel of interest PO is black, the surrounding pixel P3
The condition that both are also black holds true at the same time.
Note that the peripheral pixels P6 and PG are opposite to each other diagonally to the right with the pixel of interest PO as the center), P7-PO>C2n P2-PG>C2 (if the pixel of interest PO is black, the surrounding pixels P7 is also black, and similarly, if the pixel of interest PO is black, the surrounding pixels P
2 is also black (note that the peripheral pixels P7 and P2 are vertically opposed to each other with the pixel of interest PO as the center).

As already understood, if the pixel of interest PO is black and the opposing surrounding pixels within 1 m of the pixel of interest PO are both black, then the pixel of interest PO is set to 1, otherwise it is set to 0. This is what the above conditional expression for binarization indicates.

If binarization is performed under these conditions, Figure 19(a),
As can be estimated from (b), it is understood that if the line width of the target line image is slightly larger than one pixel, the binarization of the pixel of interest will be performed correctly. Dew.

Note that when the surrounding pixels are selected as shown in FIG. 19, for example, the distance from the pixel of interest PO to the surrounding pixels is located diagonally to the right or to the left compared to the distance to each pixel on the top, bottom, left, and right. It will be observed that the distance to each pixel is four times longer. Considering that the fact that the distance from the pixel of interest PO to the surrounding pixels is not the same for all surrounding pixels has a negative effect on the binarization of the pixel of interest, the distance from the pixel of interest PO to which surrounding pixel is The principle of another embodiment in this application is to select surrounding pixels such that all pixels are equal.

〔Example〕

Hereinafter, one embodiment of the present invention will be described in detail with reference to the drawings.

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

In the figure, 1 is a line drawing, 2 is an image sensor such as a television camera, 3 is an analog/digital (A/D) converter, 4 is a frame memory, 5 is a known two-dimensional local memory,
6 is a local average section, 7 is a dark value detection section, 8 is an address generation section, 9 is a line segment judgment section, 10A and IOB are comparators, IIA is a permissible density difference variable setting section, 11B is an ambient density difference variable setting section , 12 is an average density value memory, 13 is a uniformity/nonuniformity determination memory, 14 is a target average density value launch circuit, 15 is a subtracter, and 16 is a binary image memory.

The operation will be explained.

An arbitrary line drawing 1 is imaged by an image sensor 2 such as a television camera, converted into a grayscale image by an A/D converter 3, and stored in a frame memory 4. The grayscale image of the line drawing in the frame memory 4 is read out pixel by pixel in the same manner as the scanning order of a television camera, and a two-dimensional station memory 5 forms a moving sub-window of, for example, 3×3. The local averaging unit 6 calculates an average density value Aij for the local area of the pixel of interest, and stores this in the memory 12.

The darkness value detection unit 7 calculates eight density differences between the pixel of interest and neighboring pixels in each sub-window, checks whether the values are within a preset range, and calculates the average value for those within the range. , all of these average values are determined within the processing area, and the average value is used as a darkness value PA as a contrast condition within the processing area, and this is compared with the allowable density difference in the comparator 10A. If the threshold value PA is larger than the allowable density difference Go, it is considered that the local area is uneven in density, so the determination flag Fij is set to "0".
Let it be °°. Also, if the density is smaller than the allowable density Go, the local area can be considered to be uniform in density, so the determination flag Fij is set to "1".
°゛. Here, the allowable value Go is variable by the variable setting section 11A, and the determination flag Fij is set by the comparator 10.
is determined.

The second determination flag Fij is stored in the memory 13.

As described above, when converting characters into binarized characters, the threshold value specified as a contrast condition is changed depending on the situation of the target image.

The reason for this is explained using Fig. 8 A and B. Fig. 8 A
, B schematically shows the density of two images in which the same character type is engraved by varying the illumination. now,
Assuming that each part 101 of the black level corresponds to the density level of a character, the contrast condition is roughly determined by the magnitude of the density difference AM with the background part as shown in 102, but the contrast condition of the target image Dynamic range depends on changes in lighting and background level! The optimal dark value for binarization (the dark value is selected in the middle of the width AM) differs depending on the Jl (for example, from whitish to grayish).

Therefore, if line parts are not emphasized sufficiently, characters will become blurred, but according to the present invention, it is possible to automatically follow the darkness value that should be given as a contrast condition, so a good image can be created. It is possible to maintain the obtained state and contribute to improving the character reading rate.

Returning to FIG. 1, the contents of the average density value memory 12 and the determination flag memory 13 are read out in the order of the target subwindow and the peripheral subwindow by the discrete subwindow address generation unit 8. The average density value Co in the sub-window of interest is latched in the latch circuit 14, and the subtracter 15 calculates the difference Dk between this and the surrounding average density value Ck (k=1...8). Here, for black letters, Dk = max (0,
Ck-Co ), and for white letters Dk = max (
O, Co-Ck). Note that the allowable value Do of the ambient concentration difference can be changed by the variable setting section 11B.

The ambient density difference Dk and the allowable value Do are compared with the comparator 'l OB, and it is determined whether the density difference is the difference between the background density value and the line segment density value, and the value of the background determination flag Hk is determined. Determine.

The character line determination unit 9 determines whether the pixel of interest is a line segment pixel or not based on the uniformity determination flag Fk and the background determination flag Hk, and stores the result in the memory 16. Binary image T is stored in memory 16
is read from. The above is the general operation of the embodiment.

Hereinafter, a detailed explanation will be given with reference to FIGS. 2 to 9.

FIG. 2 shows an example of a grayscale image of the line drawing (including characters) 1 captured by the image sensor 2. Here, the black character 20
is in the white background 21, and 3× in the same direction as the scanning direction.
The subwindow 22 of No. 3 moves. Sub window 2
2 is created in a two-dimensional local memory 5. Note that the reason for forming a 3×3 pixel sub-window is to remove the influence of isolated noise and the like.

A specific example of the subwindow group is shown in FIG. In this embodiment, the average density value and uniformity determination flag are calculated for each sub-window and stored in the memory 12.13, so one pixel in the memory 12.13 corresponds to each sub-window (3 x 3 pixels of the original image). This will correspond to In other words, the average density value and the uniformity determination flag are stored in the memories 12 and 13.
Each pixel is mapped to one pixel. FIG. 3 shows an example in which 3×3 sub-windows are arranged in the X and Y directions with spans of Sx and SY, respectively.

In other words, 3×3 configuring the sub-window of interest Wso
Let the density value of the pixel of interest in Pij be Pij, and surrounding subwindows W s (W s
1 = W s8) are defined.

FIG. 4 shows an example of a local averaging unit that calculates the average density value of each subwindow Ws consisting of 3×3 pixels. This local averaging section 6 includes a switching circuit 61 for switching the nine human power outputs, an adding circuit 62, a latch circuit 63, and a dividing circuit 6.
Consists of 4.

That is, the nine pixel density values P within the subwindow Ws
The switching circuit 61 adds sO=Ps8 by 1 to the adding circuit 6.
2, the output result is latched by a latch circuit 63 to perform cumulative addition, and the result is divided by a constant 9 in a division circuit 64 to calculate an average value Aij. Note that at this time, the initial value of the latch circuit 63 is set to zero.

Moreover, when the average value Aij is expressed in a mathematical formula, it becomes as shown in the following formula.

Aij = (PsO+Psl+Ps2+Ps3+P
s4+Ps5+Ps6+Ps7+Ps8)/9 The average density value for each sub-window thus obtained is stored in the memory 12 in FIG.

A specific example of the dark value detection section is shown in Part 5. This includes a differential density detection circuit 71, an average value calculation circuit 72, and a dark value calculation circuit 7.
3, and the darkness value PA is calculated as follows.

FIG. 9 shows a two-dimensional local memory area used when binarizing a character part, and is composed of eight neighboring pixels shown at 103 and one pixel of interest shown at 104,
It is given by the following formula.

Neighboring pixels: N=(P(i-α, j-α), P(i-α, j), P(
i−α, j+α).

P(IIJ-α), P(i, j+α), P(i+α+j
−α) IP(i+α, j), P(++α, j+α)) Pixel of interest: P (11j) where P(i, j) is a function α:il! that indicates the density of point (ij). ii. Distance between elements A method for determining the optimum contrast threshold value P^ within an image with this distance between elements will be explained.

First, Pm1n is the absolute change range of P^.

Set P wax.

Pm1n≦PA≦Pmax Next, the difference density is obtained by subtracting the density value of the pixel of interest from each point of the neighboring pixels N (the difference density detection circuit 7 in FIG.
1).

N' = (N-P(i, j)) Next, among the elements of N'', those between Pm1n and P11lax are extracted.

Let P'(i, j) be the average of the elements of Plll1n≦N"≦Pmax N" (the fifth
Average value calculation circuit 72) in the figure.

This P(i, j) is determined over the entire processing screen, and the average value of the difference density values is used as the threshold value PA to be determined (darkness value calculation circuit 73 in FIG. 5).

PA = Σ P’(i, j＞/nrJ Its output PA is the set value G in the comparator 10A.

The uniformity determination flag Fij is determined as described above. Note that the determination flag Fij is stored in the memory 13.

FIG. 6 shows a specific example of an address generator that reads out the contents of the memories 12 and 13 for the subwindow of interest and peripheral subwindows. In the figure, 81 is a basic clock oscillator, 82 is a timing control circuit, 83A, 8
3B is the X direction, X direction address counter, 84A, 84
B is the X direction.

X-direction address latch circuits, 85A and 85B are addition control circuits, 86A and 86B are X-direction span and X-direction span variable setting units, and 87A and 87B are addition and subtraction circuits.

Timing control circuit 82 generates a control signal and a timing signal based on the clock from oscillator 81, and provides them to address counter 83A. The address counter 83A counts the timing clock and outputs the X-direction address To (see FIG. 3) of the subwindow of interest, while counting up the address counter 83B every horizontal scanning line and outputs the Y-direction address Jo. Output (see Figure 3). The address of this window of interest (Io.

Jo) are latched by latch circuits 84A and 84B, respectively. Therefore, X (Y) of the peripheral subwindow
The direction address (1, J) is determined by this latch circuit 84A (
84B) and the X (Y) direction span setting section 8
It can be determined by adding or subtracting the span 5x (SY) given from 6A (86B) using an addition/subtraction circuit 87A (87B). At this time, the addition/subtraction control circuit 85A (85B) gives an instruction for addition or subtraction to the addition/subtraction circuit 87A (87B). The following table shows the contents of control by the addition/subtraction control circuit 85A (85B).

In this way, by controlling the addition/subtraction control circuits 85A and 85B in the order shown in the table in the X direction and the X direction, the address (l J
) occurs. Then, the average density value Aij and uniformity determination flag Fij of each peripheral sub-window are set as Ck, respectively.
Read out as Fk (k = 0.1...9).

A specific example of the line segment determination unit that performs binarization from the background determination flag Hk and uniformity determination flag Fk regarding the surrounding subwindows is shown in FIG. 7(a), and the arrangement of the subwindows is shown in FIG. 7(a).
Shown in b). The line segment determination part is an AND (and) game)
91A, 91B, 91C.

Background counter 93, background number variable setting section 95A, comparator 9
Consists of 6A.

In the circuit of FIG. 7(a), the uniformity determination flag Fk and the background determination flag Hk are set to k=1, k=5, k=2, and
6. ni=3 and ni=7. Four sets of data, 4 and 8, are supplied, and Fk of both data.

If each Hk is valid, it is counted in the background counter 93. Therefore, the background counter will receive any value from O to 4, the variable setting section 95A will be set with the number of opposing sub-windows required to binarize the sub-window of interest, and the comparator 96A will set When the number of pairs of opposing sub-windows is larger than the number set in the variable setting section 95A, it is determined that the sub-window of interest, and therefore the pixel of interest, is a line segment, that is, l.

Note that the line segment determination section 9 can also be configured as shown in FIG. 7A. This is provided with shift registers 98A, 98B and a binarization determination memory (ROM) 99.
Introduce flags Fk and Hk to A and 98B, respectively, and set Fo
=FB, Ho to H8, and by reading out the contents of the memory 99 using this as an address, the output Q is obtained.

The embodiments have been described above, but the various variable setting values are as follows:
Needless to say, the value is adjusted to the optimum value depending on the properties of the target line figure. Also, as a sub window, 3×
Although an area of three pixels was considered, this number can be determined as appropriate, and in the extreme, it can be one pixel.

As mentioned earlier with reference to FIG. 18 or FIG. 19, in the previous embodiment, the distance 11 between the pixel of interest PO and the surrounding pixel P1, and the distance 11 between the pixel of interest PO and the surrounding pixel P2 The relationship between the distance 12 becomes the relationship 11=121.2, which causes anisotropy in the relative positions of pixels PO to P8, and the binarized image changes slightly depending on the angle of the line segment scratches that occur. However, a problem arises in that the binarization accuracy is ultimately affected. As long as we use the x,y orthogonal coordinate system, l=/! 2
It is impossible in principle to do so.

On the other hand, pixels Pi, P3. P6. Although it is possible to select a pseudo-optimal position for P8 so that ff1l=f2, this is inconvenient for high-speed processing.

In FIG. 18 referred to earlier, when PO is the pixel of interest, the surrounding pixels required for binarization are 8 pixels of PI-P8, and then 8 pixels of P5 next to PO. When PO is the pixel of interest, the surrounding pixels required for its binarization are the surrounding pixels P2 . P3. P.O., P.O.
5. Since only three pixels, P9, PIO, and pH, in addition to P7 and P8 need to be newly captured, high-speed processing is possible. There is a distance of 3 pixels between the first pixel of interest PO and the next pixel of interest P5, so PO, P
5. After completing one scan with PIO,... as the pixel of interest, repeat the same scan by shifting the selection of the pixel of interest by one pixel.If you perform a total of three scans in the same way, all of the Binarization of pixels is now complete.

FIG. 10 is an explanatory diagram illustrating an example in which surrounding pixels for a pixel of interest are selected at optimal positions such that all surrounding pixels are pseudo equidistant from the pixel of interest.

In the figure, when P1 to P8 are selected as peripheral pixels necessary for binarization of the pixel of interest PO, the peripheral pixels P1 to P
8, all the surrounding pixels are approximately equidistant from the pixel of interest PO, which is perfect in that sense, but when the pixel of interest moves to the neighboring pixel PO, the surrounding pixels necessary for its binarization are P9. It is necessary to newly capture many (peripheral pixels) such as ~P15, and high-speed processing cannot be expected.

-Although it is known to use a two-dimensional local memory of nXn pixels for boat fishing, in the range of n>5, the scale of the hardware that can be realized increases exponentially with n, making it extremely difficult to realize. It is difficult to

It enables high-speed processing and theoretically realizes that all surrounding pixels are approximately equidistant from the pixel of interest, allowing binarization to be performed without depending on the directionality of the target image, such as scratches, dust, cracks, etc. An example of a binarization device that can perform this is shown below.
explain.

To put it simply, the second embodiment manipulates the sampling period and its phase when capturing and storing a video signal from a video camera in a frame memory, and uses equidistant coordinates instead of the conventional orthogonal coordinate system. It can be said that the system converts the image data into a system, and then performs binarization processing on the image data.

FIG. 11 is an explanatory diagram showing an equidistant coordinate system that can realize peripheral pixels P1 to P6 that are equidistant from each other with respect to the pixel of interest PO, and image data is sampled as follows.

That is, when the vertical resolution is 1, the sampling time of one pixel is set to an interval of -L4, and the sampling phase of the odd field E1 of the incremental signal and the sampling phase of the even field E2 are set to 1/2. This is achieved by shifting the pixels by 4 pixels (with respect to the vertical resolution).

Binarization of target images such as scratches is performed using PO as the point of interest (pixel of interest) in the pixel arrangement shown in Figure 12, and using P1 to P6.
When is a neighboring point (peripheral pixel), PO<C1n (
(PI-PO>C2n P4-PO>C2)U(P2
-PO>C2n P5-PO>C2)U(P3-PO
>C2n P6-PO>C2)), the pixel of interest PO is set to 1 (scratch).

FIG. 12 is an example in which the inter-pixel distance d=2. In this coordinate system, the distance between the point of interest and neighboring points is always equal for any d. When moving the pixel of interest to an adjacent pixel and performing the next process, pixels P7 to P9 are newly read, and P5 is set as the new pixel of interest. Therefore, the process is completed after d scans.

Furthermore, in this embodiment, since the number of neighboring pixels can be reduced to six pixels, high-speed processing can be expected.

FIG. 13 is an explanatory diagram showing an embodiment in which a binarization device based on the above-described principle is applied as a surface flaw inspection device.

In the figure, 301 is an image processing device, 302 is a camera, 303 is a light, and 304 is a position sensor. The inspection object 305 is transported in the direction of arrow Y by a pallet 306. When the inspection object 305 enters the field of view of the camera 302,
The image is sensed by the position sensor 304 and imaged, and the image is processed in the image processing device 301.Finally, the presence or absence of scratches in the image (1 or 0 due to binarization) is determined by the upper control device. can be transmitted to.

FIG. 14 shows the main blocks constituting the image processing device 301, including an image processing unit 301A.

It consists of an image binarization section 301B and a frame memory 301C.

FIG. 15 is a more detailed block diagram of the image processing device 301, in which 3011 is a synchronization signal generation circuit for generating an external synchronization signal for the camera, and a vertical synchronization signal (VD) for the reference signal generation circuit 3012. , horizontal synchronization signal C
HD). In the reference signal generation circuit 3012, VD
, ) (Multiply D so that it becomes an equidistant coordinate system,
When inputting an image, the address generator 3013 is controlled and the image data is stored in the frame memory 3014. For binarization, the address generator 3015 generates an address for the frame memory 3014, and the latch circuit 3016 generates an address for the frame memory 3014.
The image data is latched sequentially.

Note that once all the data has been latched by the latch circuits PO-P6, unless the row line is updated, data for three pixels need only be updated for processing at the next point.

The binarization arithmetic circuit 3017 is a part that performs arithmetic operations between pixels and the like to binarize input seven-pixel data. The area counter 3018 is a counter that counts pixels that are determined to be flaws as a result of binarization, and is used to determine whether the inspection object is a good product, a defective product, etc. based on the count result.

Address generator 3015 during image binarization in FIG.
Specifically, the following address will be generated.

When PO is on odd field El (Fig. 16 (A)
), PO(x, y) and PI(x −2X1nt(d/2) 1, y
d) P2(x-d, y) P3(x-2Xint(d/2) 1.)'+
d) P4(x+2Xint(d/2), )'+d)
P5(x+d, y) P6(x+2Xint(d/2), V d)P
When O is on even field E2 (Fig. 16(B)
), for PO(x, y), PI(x 2Xint(d/2), V d)P
2(x - d, y) P3(x - 2X1nt(d/2), y +
d) P4(x+2Xint(d/2)+1.y+d
)P, 5(x+d, y) P6(x+2Xint(d/2)+1.)' d)(
However, int is a function that cuts off fractions below the decimal point and takes only the integer part.

[Effects of the Invention] According to the present invention, an optimal binarized darkness value can be determined depending on the image situation. Furthermore, the area near the pixel of interest (
Since the average density value and the uniformity determination flag are determined for each pixel (3×3 pixels), it is possible to remove the influence of small noise and perform accurate binarization.

Furthermore, according to the second embodiment, since it is possible to remove anisotropy when viewing surrounding pixels from the pixel of interest, there is an advantage that even higher precision binarization can be expected.

[Brief explanation of drawings]

Fig. 1 is a block diagram showing the configuration of an embodiment of the present invention, Fig. 2 is an explanatory drawing showing an example of a line drawing, Fig. 3 is an explanatory drawing showing a sub-window of interest and its surrounding sub-windows, and Fig. 4 is a local average. 5 is a block diagram showing a specific example of the local maximum concentration difference detection section, FIG. 6 is a block diagram showing a specific example of the address generation section, and FIG. 7 is a line segment determination section. FIG. 9 is an explanatory diagram showing an example of the configuration of the local memory.
Figure 0 is an explanatory diagram showing an example of a pixel arrangement in which the distance between the pixel of interest and surrounding pixels is the same for all surrounding pixels, and Figures 11 and 12 are each adopted in another embodiment of the present invention. FIG. 13 is an explanatory diagram showing an application example of the binarization device according to the present invention, FIG. 14 is a block diagram showing details of the main parts in FIG. 13, and FIG. FIG. 16 is an explanatory diagram of generated addresses of the address generation circuit during image binarization in FIG. 15, FIG. 17 is an explanatory diagram showing another example of line drawing, and FIG. The figure is an explanatory diagram showing an example of the arrangement of a pixel of interest and surrounding pixels, and FIG. 19 is an explanatory diagram showing the relationship between the line width of an image and the arrangement of the pixel of interest and surrounding pixels. Explanation of symbols 1...Line drawing, 2...Image sensor, 3...A/
D converter, 4... Frame memory, 5... Two-dimensional local memory, 6... Local average unit, 7... Local maximum density difference detection unit, 8... Address generation unit, 9...・Line Segment Judgment Department, S 13 Agent: Patent Attorney Akio Namiki Agent: Patent Attorney Kiyoshi Matsuzaki First Father-in-law Diagram Y (J) Figure Figure 6! JITA Figure No. 92 Uj Tomi 7 Figure (α) k13] ~ 412; Frame [I+11) Waffle window arrangement! Figure 8A Figure 9! Figure 8B J Horse Y π1O Figure Manma■ Figure 13 Shin 14 Cause 01C M12 Prisoner Tomi 15g Atsushi 16 Prisoner (A) Meshiriniku ■ Arrangement Figure 17 Figure 118 Remains Figure 16 (B) Meushi Placement within 1! II9WJ

Claims (1)

[Claims] 1) Image data obtained by two-dimensionally imaging a background and a line figure drawn thereon is converted into pixels and imported into a memory, and then the pixels are read out from the memory and each pixel is In a binarization device that binarizes each pixel depending on whether it is a pixel that corresponds to a background part or a pixel that corresponds to a line figure part, the binarization is performed according to the width of the line that makes up the line figure. A sub-window of interest is set for the target pixel of interest alone or for the pixel of interest and its surrounding pixels together, and similarly, depending on the width of the line constituting the line figure, the pixel of interest or the pixel of interest and the surrounding pixels are set as a single pixel or the pixel of interest and the surrounding pixels. A subwindow reading means that sets a plurality of subwindows each consisting of a pixel and its surrounding pixels, and sequentially reads out each of the subwindows from the memory; A calculating means for calculating, an average value calculating means for calculating an average value of difference values between a pixel of interest and its neighboring pixels for each sub-window within a predetermined range, and a sub-window of interest and its surrounding sub-windows by the average value calculating means. a first determination means for determining whether or not a region within a corresponding window group is a uniform density region based on each average value of the subwindows; a second determining means for determining whether the first and second
A binarization device characterized in that it determines whether or not a sub-window of interest corresponds to a line graphic part based on an output from a determination means, and then performs binarization. 2) In the binarization device according to claim 1, how to select surrounding sub-windows with respect to the sub-window of interest;
A binarization device characterized in that coordinates are selected such that all peripheral sub-windows are located at the same distance from a sub-window of interest.