JPH01313743A - Method for inspecting colored periodic pattern - Google Patents

Abstract

PURPOSE:To make a pitch at the level of a signal by a colored periodic pattern short and to prevent the occurrence of moire by performing inspection based on image data obtained by picking up an image after the levels of the video signals of respective colors constituting the colored periodic pattern become equal. CONSTITUTION:Light sources 1a-1c for R, G and B are constituted so that they can respectively and independently adjust the intensity of light and an optical fiber bundle 3 obtained by bundling optical fibers 2a-2c which receive the light from the respective light sources is bundled so that the fibers of the respective colors are uniformly mixed at an outgoing port 4. By illuminating a sample 46 placed on an XY stage 50 with said fiber as a light source, the image is picked up with a television camera 41. While observing the video signals, the light quantities of R, G and B are adjusted so that the levels of the signals become equal in the respective colors and the spectrum of illuminating light is set. Thus, the colored periodic pattern can be inspected based on the image data obtained by irradiating the sample 46 having the colored periodic pattern with the light and picking up the image.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention is applicable to color separation filters used in picture tubes and solid-state image pickup tube elements used in color television cameras, and color filters used in liquid crystal display devices. The present invention relates to a method of inspecting a periodic pattern in which unit patterns are repeatedly arranged for defects, unevenness, etc.

[Conventional” technology]

Conventionally, CCD color filters and liquid crystal color filters are
Defects or unevenness in the pattern will cause defects in captured images or displayed images, and minute irregularities on the surface will cause products using the color filter to be defective. Afterwards, an inspection is carried out,
Defective color filters must be excluded.

Inspection of such color filters has conventionally been carried out by microscopic inspection or visual inspection using various types of illumination, but there have been problems such as individual differences and overlooked defects.

Therefore, various methods have been proposed to solve these problems.

An inspection method of photographing a periodic pattern with a television camera and processing the image will be described with reference to FIGS. 9 to 11.

In the inspection apparatus shown in FIG. 9, in order to detect an abnormality in the opening area of the object to be inspected 46 having periodic openings as a unit pattern 51 as shown in FIG. The object to be inspected 46 is illuminated by a transmitted illumination unit composed of a light beam 48 and a diffuser plate 47,
The image processing device 1f42, which photographs the inspection area, A/D converts the output signal of the TV camera into digital image data, and performs various image processing including addition and subtraction on the screen at high speed using the frame memory and arithmetic unit. conduct. The control device 43 includes the image processing device 42, the XY stage 50, and the drive mechanism 4.
The pattern is moved by controlling the pattern moving mechanism comprised of 5. Note that in FIG. 10, 52 and 53 are unit patterns with defects.

The photographing conditions are such that changes due to unit apertures of the video signal by the TV camera 41 can be ignored, for example, about 10 unit apertures 11 are included in the pattern area corresponding to one pixel, and the direction in which the pattern is moved and displaced is the scanning line of the TV camera 41. A case where the displacement distance of the pattern is an integral multiple of the pixel pitch will be explained with reference to FIG. 11.

The light transmission weight distribution on a straight line passing through a pattern defect is, for example, as shown in FIG. 11(a). , that is, a change in light transmittance 54 due to white defects,
As shown by 52 in FIG. 10, a change 55 in light transmittance due to a defect whose opening area is smaller than that of the normal pattern 51, that is, a black defect, is detected. In addition, when the video signal is scanned on the same line as in the case of Fig. 11 (a), it becomes as shown in Fig. 11 (b), and the gradual signal change (shading) due to uneven illumination of the pattern, uneven sensitivity of the imaging surface, etc. ), random noise generated in the video signal processing device, and local changes 56 in the signal due to dust adhering to the optical system.

By adding multiple frames of such video signals, the ratio of random noise components can be reduced to 1/J when the number of additions is N (see Figure 11 (C).
)). Next, when the pattern is displaced and similar screen addition processing is performed, as shown in Figure 11 (d), the signal due to the defect on the pattern also moves as the pattern moves, but the shading of the imaging system The position of the signal 56 has not changed due to dust in the optical system or the like. Therefore, Figure 11 (
When the data shown in FIG. 11(d) is subtracted from the data shown in c), the signal 56 due to shading, dust, etc. contained in both data is erased, and the signal 56 due to the light transmittance change of the pattern and the reduced random noise component are removed. As a result, the signal due to the defect appears with almost the same difference in value from the average value in the vicinity at a position where the number of pixels corresponds to the amount of pattern movement, and the sign appears inverted, and the order of the inversion is 2
The order is reversed depending on the defect type W4 (white defect, black defect).

In the image data processed as above, the brightness changes locally only in the defective area, so it can be easily recognized as a defect by observing it on a monitor, and the brightness at the defective area is reversed relative to the surrounding area. It is also possible to identify the type of defect in the order of .

Next, a method of moving the pattern will be explained with reference to FIG.

As shown in FIG. 12, there are inspected object positions P0 to P, and the above-mentioned movement corresponds to Po and P in this figure. In this case, if processing is performed using only image data obtained from two locations, spots caused by changes in the aperture ratio in the H direction in the figure will be detected, but changes in other directions, such as the ■ direction, will be detected. There is a drawback that large spots with small changes in the H direction are not detected, which causes anisotropy in detection sensitivity. ni, Po and P,,P, and P! , Po and Pi
, Po and P4 can be subjected to predetermined processing to eliminate anisotropy. Also, Pl in the figure
Similarly, anisotropy can be avoided even if detection is performed based on image data obtained by subtracting the sum of image data at each position arranged on the circumference from the image data at the center position P0, as shown in P8. In addition, a value that approximates the curvature of the brightness distribution that is easily visually responsive can be obtained, resulting in test results that are close to visual inspection. The image data obtained by imaging at each position includes the aforementioned random noise component, shading component, and fixed noise component, and the random noise component is obtained by adding multiple frames at each moving position. In addition, shading components and fixed noise components can be eliminated by performing calculations on image data such that the sum of the number of frames of image data becomes "0". For example, add 32 frames at position P and 8 frames at PI in Figure 12, and subtract the image data at Po from the image data obtained by multiplying the image data at P by 4. In this case, the sum of both image data is “0”.
As a result, shading components and fixed noise components are eliminated. Also, if screen addition for 4 frames is performed at each position from P to P, the addition result of image data from P to pH is equivalent to the addition result of image data for 32 frames. Therefore, by subtracting from the result of adding 32 frames worth of image data at the position of Po, the shading component and fixed noise component will be deleted in the same way, and
A two-dimensional differential value of the brightness distribution is obtained. Furthermore, if smoothing processing is applied to the image data whose shading components and fixed noise components have been reduced through the above processing,
Random noise components are further reduced, making it possible to detect extremely slight unevenness components, and also suppressing changes in image data due to minute defects or short-cycle unevenness.

A method has also been proposed for automatically determining whether a product is good or bad based on image data that has been subjected to the above-described image processing.

Figure 13 shows an example measured under the same conditions as Figure 11,
Figures 13 (a) to (e) are the same as Figure 11, where A is a portion where the aperture ratio changes gradually, B is a portion where the aperture ratio changes periodically, and C is a change in the aperture ratio. The large and isolated portion D is a local variation component (fixed noise component) of the video signal due to dirt in the optical system, etc. 1st
It can be seen from FIG. 3(e) that components due to changes in the aperture ratio of the object to be inspected are extracted. In this case, the amount of movement of the object to be inspected varies depending on the state of the unevenness to be detected, but as the period of variation of the unevenness increases, the amount of movement of the object to be inspected also increases. The number of pixels is set to about 2 to 20 pixels. Next, by subtracting the average value of the image data in a sufficiently wide area from the image data in FIG. 13(e), the result shown in FIG. 13(f) is obtained.
When the image data in Fig. 3(e) is differentiated, it becomes as shown in Fig. 13(g), and by setting a predetermined darkness value S+, 32 according to the nature of the unevenness to be detected, the unevenness can be automatically detected. can be detected. Figure 13 (h) is the 13th
Threshold values S+ and S2 are set for the image data in figure (g), and when the darkness value is exceeded, it is set as "1", and when it is not exceeded, it is set as "0".
This is the binarized data shown as . And Figure 13 (
If an isolated unevenness as shown in C of a) is detected and the product is determined to be defective, a darkness value such as Sl may be set for the image data of FIG. 13(f) or (g).

In addition, when detecting periodically changing unevenness as shown in B in FIG. 13(a) and determining the product as defective,
) or (g), set a darkness value such as S2, convert it to binarized data as shown in FIG. The number of irregularities in a predetermined area as shown in i) is converted into density data, and by setting a predetermined darkness value S3 for this density data and comparing it, only irregularities that are changing periodically can be determined. It is possible to detect whether the product is good or bad based on the results.

[Problem to be solved by the invention]

However, when the above-mentioned inspection method is applied to the inspection of multicolor periodic patterns, in order to treat a set of patterns colored in R, G, B, etc. as a unit pattern, the periodic patterns are As the array pitch becomes larger and the period of the image pixels becomes closer, moiré in the captured image becomes stronger, and there is a problem in that the detection sensitivity cannot be increased. Further, when inspecting only using transmitted light, there is a problem in that irregularities in the transparent layer on the surface or defects on the light-shielding pattern cannot be detected reliably.

The present invention has been made to solve the above-mentioned problems, and an object of the present invention is to provide a method for inspecting a colored periodic pattern, which enables highly sensitive and reliable inspection of color filters and the like.

[Means to solve the problem]

To this end, the present invention provides a method for inspecting a colored periodic pattern from image data obtained by irradiating light onto a sample having a colored periodic pattern and reimaging the sample. It is characterized in that the inspection is performed based on image data captured so that the signal levels are equal.

[Effect]

The present invention inspects a colored periodic pattern based on image data captured such that the video signal level of each color constituting the colored periodic pattern is equal, and the pitch in the signal level due to the colored periodic pattern is It is possible to shorten the time to prevent moiré, and set a high resolution to perform highly sensitive defect detection.

〔Example〕

Examples will be described below with reference to the drawings.

FIG. 1 is a diagram showing an embodiment of the present invention, and FIG. 2 is a diagram showing an embodiment of the configuration of a light source, and the same numbers as in FIG. 9 indicate the same contents. In addition, la to IC are R, G, and B, respectively.
2a to 2C are optical fibers, 3 is an optical fiber bundle,
4 is an exit aperture, and 5 is a lens.

For example, as shown in FIG. 2, the R, G, and B light sources 1a to lc condense light from a halogen lamp 11 with a reflecting mirror 12, and pass the light of a specific color through a color filter 13 to optical fibers 2a to lc. They are configured to receive light at 2C, and each light intensity can be adjusted independently. The optical fibers 3 are made up of bundled optical fibers 2a to 2c that receive light from each light source, and are bundled so that the fibers of each color are uniformly mixed at the output port 4. Use this fiber as a light source on the XY system and Daige 50! ! The placed sample 46 is illuminated and imaged by a television camera 41. While observing this video signal, set the spectrum of the illumination light by adjusting the R, G, and B light intensity so that the signal level is between each color, and the arrangement pitch of the color filters will be changed as shown in Figure 14. What used to be R, G, and B as units becomes 1/3 as shown in Figure 3,
As a result, the occurrence of moiré is suppressed, inspection can be performed with a high resolution, and defects can be detected with high sensitivity.

By the way, in detection using transmitted light as shown in Figure 1,
As shown in FIG. 4, the unevenness 25 and foreign matter 26 of the transparent overcoat layer 24 on the colored pattern 22 and the light-shielding pattern 23 formed in the glass 21 cannot be detected or cannot be detected reliably. In this case, the detection described in FIG. 1 may be performed using dark-field epi-illumination as shown in FIG. 5 or other bright-field epi-illumination (not shown). When the sample 35 in FIG. 5 is a reflective colored image, a light source that can independently adjust R, G, and B as described in FIG. 1 may be used.

In FIG. 5, light from a light source 31 is reflected by a reflecting mirror 33, and a sample 35 is illuminated through an annular condenser lens 34 disposed around an objective lens 36.

Although FIG. 5 shows that a minute portion of the sample is illuminated, the illumination area may be set by changing the distance between the lens 34 and the sample so that a desired range can be imaged by the television camera.

FIG. 6 is a diagram showing another embodiment of the present invention using a cooled CCD camera as an imaging device, and in the figure, 61 is a cooled CCD camera.
CD camera, 62 is an image processing device, 64 is a shutter, 6
5 is a shutter driving device, and 66 is a sample.

The cooled CCD camera 61 is a CCD camera that is cooled using an electronic cooling method or the like to significantly reduce dark current and noise to a negligible level, and is capable of long-time exposure in dark areas.
It is characterized by good linearity of the video signal with respect to the integrated light amount, and can clearly capture dark areas with high image quality that could not be captured even with conventional high-sensitivity TV cameras.
It is possible to detect up to several photons per pixel per second.

Using such a cooled CCD camera 61, the light intensity of each color is adjusted so that the signal level of each color is the same when the image is captured, and the sample 66 is irradiated with the optical fiber 3 to obtain CCD.
Take an image with a CD camera. Then, the image data taken without the sample is Io, the image data taken with the sample inserted is I, and the image data taken with the shutter 4 closed is ■. Then, the transmittance T of a point on the sample is 1, i.

It can be calculated as Here I, 1. , I is the image data at the corresponding position, and if the image data when the shutter is closed (light amount = 0) can be ignored, then T=1/I.

The transmittance is obtained as . This calculation can be performed by inter-screen calculation after each image data is stored in the frame memory by the image processing device 2. If the number of pixels of the cooled CCD camera is 512x5L2, this calculation will result in approximately 2
Transmittance data for 50,000 points will be obtained. The image data obtained in this way does not contain components such as shading or dust in the imaging system, so by performing image processing on this image according to each inspection item as described above, it is possible to perform a single measurement. It is possible to conduct an inspection.

Further, if the light source is stable, the image data I is stored in the memory and used, thereby eliminating the need to perform measurement for each sample and reducing the measurement time.

With ordinary solid-state image sensors and image pickup tubes, the linearity of the light intensity and video signal is insufficient, and the thermionic form is large, making it difficult to measure with high precision. are spatial resolution, linearity, S/N
was good, but there was a problem that the imaging time was long (
(about 1 ms per pixel, about 4 minutes for 250,000 pixels), but in the present invention, by using a cooled CCD camera, an image can be captured within several seconds and the linearity is also good.

In addition, defect detection is performed on image data by performing differential processing on image data, which is equivalent to the result of conventional imaging (frame integration) → sample movement → imaging (frame integration), without moving the sample. After this is obtained, defects can be detected by performing similar image processing. In this case, feature extraction may be performed by using spatial filters for differential processing, such as those shown in FIGS. 7(a), (b), and (C). One-dimensional edge extraction can be performed by using the spatial filters shown in Figures 7(a) and (b), and Figure 7(c)
) can perform two-dimensional edge enhancement. Then, the spatial filter can be selected according to the variation in image data, resolution characteristics, and the nature of the defect to be detected.
) is used to determine the size of the pixel of interest, and identification can be made based on the size of the pixel of interest relative to surrounding pixels.

Furthermore, by using the spatial filter shown in Fig. 7(C) to detect unevenness, the same results as when performing a series of imaging operations as explained in Fig. 13 can be obtained, and in the case of Fig. 13, Similarly, detection and judgment can be performed by performing screen addition processing and setting the darkness value.

By the way, in the transmittance measurement method shown in Fig. 6, - Variations in data for each pixel directly affect the data, so in order to perform highly accurate measurements, image data of a small area centered around the measurement point, for example 5 It is necessary to calculate an average value such as ×5 pix to 10×10 pix. If the number of measurement points is limited, the time required for calculation by the CPU can be reduced, but if the number of measurement points is large, the processing by the CPU will require a lot of time. If predetermined image data is read out after performing smoothing filter processing, which is a type of image processing, on the frame memory that has been stored, speeding up can be achieved.゛As mentioned above, a cooled CCD is characterized by good linearity of the video signal with respect to the integrated light amount, but when measuring a sample with low transmittance, image data captured without a sample ■1 Even if is set to a value close to max, the value of the image data (2) captured by inserting the sample becomes small, and slight errors in linearity, dark current, etc. affect the transmittance value.

Therefore, when capturing images of ■ and I, it is better to capture images so that the data levels of these two images are approximately equal and have sufficient size, and in that case, image processing requires , correction calculations are required.

One way to do this is to set the exposure time during image capture in 11.
Operate the shutter so that it becomes 1/T at ■, and
, can be set to almost the same value.

However, mechanical shutters have the disadvantage that the operating time varies, so the error becomes large especially when the shutter open time is short. If this error cannot be ignored, measure the shutter open time and use that value. Just correct the data.

FIG. 8 is a diagram showing another embodiment of the present invention for realizing this, and the same numbers as in FIG. 6 indicate the same contents. In the figure, 70' is a half mirror 171 which is an optical sensor, and 72 is a measuring device.

In this embodiment, a sensor 71 detects a part of the light when an image is captured by a half mirror 70, and a measuring device 72 measures the time during which a detection signal is obtained, thereby determining the exposure time. Correct the data using the exposure time determined in this way. It is possible to make the two image data levels approximately equal.

In addition, by changing the storage time of the cooled CCD, for example, the image data (I) and (I) captured by inserting the sample are made longer.
may have the same level. In this case, the time can be set with sufficient accuracy, but if the light transmittance is low, 1. Since the time to image is 1/T times that of I, for example, when the transmittance is about 0.1%, it takes 1000 times as much time.

Also, similar results can be obtained by changing the brightness of the light source when taking images of I and 1.

In this way, in this example, the transmittance data is obtained as high-resolution image data, so this image data can be processed to automatically recognize the position and rotation of the sample to obtain data at a predetermined position. This allows for easy automatic measurement. In addition, changes in the product to be measured can be handled simply by changing the program without mechanical adjustment.

〔Effect of the invention〕

As described above, according to the present invention, it is possible to prevent the occurrence of moiré due to colored patterns, so it is possible to set a high resolution and perform highly sensitive defect detection, and by using epi-illumination etc. It is possible to reliably detect defects that cannot be detected with conventional lighting.

[Brief explanation of the drawing]

Fig. 1 is a diagram showing an embodiment of the present invention, Fig. 2 is a diagram showing an embodiment of the configuration of the light source, Fig. 3 is a diagram showing the relationship between the coloring periodic pattern and the imaging signal level, and Fig. 4 is a diagram showing the relationship between the coloring periodic pattern and the imaging signal level. Figure 5 is a diagram to explain a specific defect, Figure 5 is a diagram to explain an epi-illumination method, Figure 6 is a diagram showing an example of I-imaging with a cooled CCD camera, and Figure 7 is a diagram showing a space. Diagram showing the filter,
FIG. 8 is a diagram showing another embodiment of the present invention in which exposure time is measured, FIG. 9 is a diagram for explaining a conventional periodic pattern inspection method, and FIG. 10 is a diagram illustrating a periodic pattern inspection method. Figure 11 is a diagram to explain the defect, Figure 11 is a diagram to explain the conventional detection method, Figure 12 is a diagram to explain the pattern movement method, and Figure 13 is a diagram to explain the unevenness detection method. and FIG. 14 are diagrams showing the relationship between the coloring periodic pattern and the imaging signal level. 1 a ~ 1 c-R, light source for GXB, 2 a ~ 2 c
-... Optical fiber, 3... Optical fiber bundle, 4...
Output port, 5...lens. Applicant Dai Nippon Printing Co., Ltd. Agent Patent attorney Masanobu Hirukawa (4 others) Figure 4 Figure 5 (a) Figure 6 Figure 7 (a) (b) (c) Figure 8 S Figure 10 'l', : ', c・ -o○○○go- 1---○σ○○○- -○○○○○-・\ : : 152: : Figure 11 Figure 12 The 14th school

Claims (4)

[Claims] (1) Irradiating a sample with a colored periodic pattern with light,
A method of inspecting a colored periodic pattern from image data obtained by imaging, characterized in that the inspection is performed based on image data captured such that the video signal level of each color constituting the colored periodic pattern is equal. A method for inspecting colored periodic patterns. (2) The method for inspecting a colored periodic pattern according to claim 1, characterized in that the sample is illuminated by a light source whose light intensity for each color of R, G, and B can be adjusted independently. (3) The method for inspecting a colored periodic pattern according to claim 1, characterized in that the sample is illuminated by epi-illumination. (4) The method for inspecting a colored periodic pattern according to claim 1, characterized in that the sample is imaged by a cooled CCD camera.