JP3695120B2 - Defect inspection method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a defect inspection method for inspecting defects of display elements such as liquid crystal panels and plasma displays used in computer monitors, TV receivers, liquid crystal projectors, and the like.
[0002]
[Prior art]
When trying to detect minute defects in the display element, the defect is detected depending on the position on the display element due to problems such as uneven brightness of the illumination, problems with the viewing angle of the liquid crystal, and unevenness in the amount of light captured by the optical system Sensitivity changes, and it may be difficult to accurately detect minute defects. Japanese Patent Laid-Open No. 4-336384 discloses a method for solving such a problem. This method is known as a defect detection method by so-called shading correction, and is a method of generating a background image by performing smoothing filter processing of an input image and taking a difference between the background image and the input image.
[0003]
[Problems to be solved by the invention]
The inspection area is within the display area of a display element such as a liquid crystal panel, but the area captured by the TV camera is set to be slightly wider than this display area so as to allow some deviation during workpiece setting. Yes. In order to increase the resolution, a configuration as disclosed in Japanese Patent Laid-Open No. 8-254499 using a plurality of TV cameras is also conceivable. In this case, however, the cameras are arranged so that the entire display area is imaged by the TV camera. The Therefore, when a smoothing filter process is performed on the input image input in this way to create a background image, the luminance difference between the display area and the area outside the display area (usually, the area outside the display area is larger than the display area). Therefore, when the shading correction is applied as it is, the outermost peripheral portion of the display area is always detected as a defect. However, if the false defect signal detected in the outermost peripheral portion of the display area is ignored, if there is a defect in the outermost peripheral portion of the display area, the defect is missed.
[0004]
Therefore, the present invention solves such a problem, and an object of the present invention is to prevent generation of a pseudo-defect signal in the outermost peripheral portion of the display area in the defect detection method by shading correction. The present invention provides a method that enables accurate defect detection over a wide range.
[0005]
[Means for Solving the Problems]
The defect inspection method according to claim 1 is a defect inspection method for detecting a micro defect of a display element of a display element,
An input image generation step of capturing an inspection region of the display element and generating an input image;
A background image generation step of generating a background image by performing density expansion processing of the smoothed input image obtained by performing the smoothing filter processing of the input image;
A comparison step of comparing the input image with the background image is provided in this order.
[0006]
Here, the above-described density expansion process is the following process. First, a pixel P of interest is determined from the smoothed input image obtained by performing the smoothing filter process on the input image. Then, a predetermined region (for example, the pixel P, the pixel on the pixel P, the pixel below the pixel P, the left pixel of the pixel P, and the right pixel of the pixel P) including the pixel P is included. The signal of the pixel having the highest signal intensity in the region including the pixel P) is replaced with the signal of the original pixel P as the signal of the pixel P. This operation is performed for all pixels of the smoothed input image.
[0007]
For this reason, when the signal of the display area of the display element is higher in intensity than the signal outside the display area, sagging occurs in the outermost peripheral portion of the display area. Will be pushed out of the display area.
[0008]
As a result, since no pseudo defect signal is generated in the outermost peripheral portion of the display area, accurate defect detection can be performed over the entire display area.
[0009]
The defect inspection method according to claim 2 is a defect inspection method for detecting a minute defect of a display element,
An input image generation step of capturing an inspection region of the display element and generating an input image;
A background image generation step of generating a background image by performing density reduction processing of the smoothed input image obtained by performing the smoothing filter processing of the input image;
A comparison step of comparing the input image with the background image is provided in this order.
[0010]
Here, the above-described shading reduction processing is the following processing. First, a pixel P of interest is determined from the smoothed input image obtained by performing the smoothing filter process on the input image. Then, a predetermined region (for example, the pixel P, the pixel on the pixel P, the pixel below the pixel P, the left pixel of the pixel P, and the right pixel of the pixel P) including the pixel P is included. The signal of the pixel having the lowest signal intensity in the region including the pixel P is replaced with the signal of the original pixel P as the signal of the pixel P. This operation is performed for all pixels of the smoothed input image.
[0011]
For this reason, when the intensity of the signal in the display area of the display element is lower than that of the signal outside the display area, sagging occurs in the outermost peripheral area of the display area. Will be pushed out of the display area.
[0012]
As a result, since no pseudo defect signal is generated in the outermost peripheral portion of the display area, accurate defect detection can be performed over the entire display area.
[0013]
A defect inspection method according to a third aspect is the defect inspection method according to the first or second aspect, wherein the comparison step is performed by performing an inter-image subtraction process.
[0014]
For this reason, since the type of the image data may be an integer type, the amount of data is relatively small, the calculation processing can be performed at high speed, and as a result, defect inspection can be performed at high speed.
[0015]
A defect inspection method according to a fourth aspect is the defect inspection method according to the first or second aspect, wherein the comparison step is performed by performing an inter-image division process.
[0016]
For this reason, the type of the image data needs to be a floating point type. However, even when the luminance unevenness is large in the display area, the SN ratio between the normal pixel and the defective pixel can be kept constant to a certain extent. There is an effect that it is possible to detect a defect.
[0017]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, the present invention will be described in detail based on examples.
[0018]
(Example 1)
FIG. 1 is a flowchart illustrating the defect inspection method according to the first embodiment. FIG. 2 is a diagram illustrating a captured input image, and FIG. 3 is a configuration diagram of the defect inspection apparatus according to the first embodiment. The inspection target is an image projected by the liquid crystal projector 11. The inspection pattern 2 is projected on the screen 10 by the liquid crystal projector 11, and the projected inspection pattern 2 is imaged by the two-dimensional TV camera 12. The picked-up image is sent to the image processing apparatus 13, where it is inspected whether there is a defect in the inspection pattern 2 projected here, and the quality is judged. Defects to be detected here are only defects per pixel of an active matrix display element called a point defect, and defects that have a relatively large area such as color unevenness are not handled. To do. In the inspection pattern 2 to be projected, the entire display area is switched to white display, black display, and gradation display one after another, and each is imaged and inspected.
[0019]
In the image input process of step S1, a certain inspection pattern 2 displayed on the screen 10 is captured by the TV camera 12, and this image data is stored as an image A in a memory in the image processing apparatus 13. In recent years, the resolution of display elements has increased from VGA to SVGA, XGA and more, and the resolution of TV cameras for inspecting such display elements is inevitably high. Conventionally, image processing of about 512 horizontal pixels × 480 vertical pixels has been common, but TV cameras of 1000 pixels × 1000 pixels or more are also being used. In addition, a method of dividing an image to be inspected and inputting it to increase the resolution is widely used. In any case, the input image A is an image in a larger range than the display area 2 so that all the display elements are captured as shown in FIG. Examining the luminance change on the data extraction line 3 in the input image, the luminance data is as shown in FIG. Here, although omitted in FIG. 2, it is assumed that there are a bright spot defect and a black spot defect on the data extraction line 3. Changes on the graph corresponding to this defect are the bright spot defect 30 and the black spot defect 31 in FIG. The portion corresponding to the display area 2 is the range indicated by 20 arrows in this graph. It is assumed that the display area part 2 has higher luminance than the part outside the display area. When the change in the data in the display area is viewed macroscopically, it has a large mountain shape with relatively bright brightness at the center and dark brightness at the periphery. The cause of this is considered to be uneven illuminance of the illumination light source and characteristics of the liquid crystal panel and the lens optical system. However, it is relatively difficult to improve this situation due to technical and cost problems. Therefore, it is desired to detect the bright spot defect 30 and the black spot defect 31 from the image state input in this manner, but the difference between the maximum value and the minimum value of the brightness unevenness is larger than the brightness change of the defects 30 and 31. It is not possible to simply binarize and detect only defects. That is, when the maximum value of the luminance unevenness is larger than the luminance peak of the bright spot defect 30 or the minimum value of the luminance unevenness is smaller than the minimum value of the black spot defect 31, the luminance unevenness is simultaneously with the defect. This will also detect pseudo defects due to. Therefore, it is considered that only a minute defect such as the bright spot defect 30 or the black spot defect 31 is detected from such luminance unevenness by performing shading correction processing described later.
[0020]
In step S2, a copy of image A is created and smoothing filter processing is performed. The reason for making a copy of the image A is to use the image A when performing the inter-image calculation later. This smoothing filter process is a process of removing a minute area luminance change such as a point defect from image data. Usually, a 3 × 3 or 5 × 5 matrix convolution operation is performed over the entire target image. Realized. An image created as a result of the smoothing filter processing is defined as an image B. The state of this process will be described with reference to FIG. 4B after the smoothing filter process. Compared with FIG. 4A, it can be seen that the luminance change of a fine period is removed, and the luminance change corresponding to the bright spot defect 30 and the black spot defect 31 is also removed on the graph. Furthermore, the luminance change at the boundary between the display area on both sides of FIG. 4B and the outside of the display area is in a sagging state as compared to the graph before smoothing filter processing (FIG. 4A). The brightness change is moderate. The graph of FIG. 4C is a graph showing the result of taking the difference between the input image A and the image B after smoothing filter processing. In FIGS. 4A and 4B and FIG. 4C, the scale of the vertical axis is changed for easy understanding of the defect. Conventionally, in order to remove the influence of luminance unevenness, such a difference image is obtained and binarized to detect a defect. Here, before and after the dotted lines on both sides which are included for the purpose of indicating the position of the display area 20 When attention is paid to the data change in the vicinity, it can be seen that the portion of the pseudo defect 21 is generated. If binarization is performed for the purpose of detecting defects in this state, the pseudo defects 21 are also detected simultaneously with the bright spot defects 32. In order not to detect the pseudo defect 21, the inspection range must be set inside the display area 20 so as not to include the pseudo defect 21. The defect will be overlooked.
[0021]
In order to avoid this, in the defect inspection method of the present embodiment, the density expansion process in step S3 is performed before the difference image is obtained. The density expansion processing is described in column 9 and FIG. 3 of JP-A-4-336384, and the density reduction processing described later is described in column 10 and FIG. 4 of JP-A-4-336384. When gradation expansion processing is performed on the image B subjected to the smoothing filter processing, an effect of stretching the waveform in both directions in the direction of the arrow 22 as shown in FIG. 5C is obtained. This image is called an image C. As a result, the sagging portion of the shoulder in the vicinity of the display area boundary is expelled to the outside of the display area 20, and a difference image between the image A and the image C (this is referred to as an image D) in the image calculation processing in step S4. ).
[0022]
If the brightness of the display area is brighter than the surrounding area, the density expansion process is performed in step S3. Conversely, if the surrounding brightness is brighter, the density reduction process is performed.
[0023]
By doing so, the pseudo defect 23 appears only outside the dotted line of the display area 20 and can be accurately detected over the entire display area, as can be confirmed by the waveform after the difference processing in FIG. It becomes possible. As described above, the calculation between the images in step S4 has been described as taking the difference between the two images. However, as the image data, an integer data type of 8 to 16 bits per pixel is generally used. Therefore, the difference calculation can be processed with the data type as it is, and the processing is simpler, so it is easy to speed up. However, even in the case of defects of the same type and similar, there may be a large difference in defect contrast between the case where the brightness unevenness is bright and the case where the brightness is dark. In such a case, the division processing is used in step S4. But good results are obtained. On the other hand, since it is necessary to make the image data into a floating point type, the amount of data increases and the processing time tends to become longer.
[0024]
In step S5, only a defect is extracted based on a binarized image based on a predetermined threshold value. Here, the size and number of defects, position information, and the like are measured, and in step S6, the quality of the inspection target is determined based on this information.
[0025]
The threshold value for binarization may be set in advance by checking the threshold level of a defect to be detected using a limit sample or the like, or based on the standard deviation of data for a non-defective product. Further, when the direction in which the defect data is output is reversed as in the difference between the bright spot defect 34 and the black spot defect 35 in FIG. 5D, a plurality of threshold values for extracting the defect are prepared. It is also necessary.
[0026]
The above explanation is an example in the case where defect inspection is performed from an inspection pattern projected on a screen by a liquid crystal projector, but the liquid crystal panel itself is placed on the opposite TV camera with a flat backlight placed behind. In the case of inspecting a catch defect, it is possible to inspect based on the same principle.
[Brief description of the drawings]
FIG. 1 is a flowchart showing an example of a defect inspection method according to a first embodiment.
FIG. 2 is a diagram showing an imaged input screen.
FIG. 3 is a configuration diagram of an inspection apparatus using the defect inspection method according to the first embodiment.
FIG. 4 is a graph illustrating a conventional shading correction procedure.
FIG. 5 is a graph for explaining a shading correction procedure according to the first embodiment.
[Explanation of symbols]
1 ... Input area 2 ... Display area

Claims (4)

A defect inspection method having a processing step in which when an image is captured in order to detect a minute defect of a display element and the captured image is processed, luminance sagging occurs in the outermost peripheral portion of the display area ,
An input image generation step of capturing an inspection region of the display element and generating an input image;
And line Cormorant smoothing filter processing step of smoothing filter processing of the input image,
Smoothing filter processing image obtained by the smoothing filter processing step, the shoulder sagging portion of the display region near the boundary, and the line Cormorant gray expansion treatment process the density expansion process for extruding outside the display area,
A defect inspection method comprising: a comparison step of comparing the input image and the background image in this order, using the grayscale expansion image obtained in the grayscale expansion processing step as a background image . A defect inspection method having a processing step in which when an image is captured in order to detect a minute defect of a display element and the captured image is processed, luminance sagging occurs in the outermost peripheral portion of the display area ,
An input image generation step of capturing an inspection region of the display element and generating an input image;
And line Cormorant smoothing filter processing step of smoothing filter processing of the input image,
Smoothing filter processing image obtained by the smoothing filter processing step, the shoulder sagging portion of the display region near the boundary, and the line Cormorant gray reduction process step the density reduction processing for pushing out of the display area,
A defect inspection method comprising a comparison step of comparing the input image and the background image in this order, using the grayscale reduced image obtained in the grayscale reduction processing step as a background image . In the defect inspection method according to claim 1 or 2,
A defect inspection method, wherein the comparison step is performed by performing a subtraction process between images. In the defect inspection method according to claim 1 or 2,
A defect inspection method, wherein the comparison step is performed by performing an image division process.

Images