CN114792310A - A Mura Defect Detection Method with Blurred Edges in LCD Screens - Google Patents

Abstract

The invention provides a Mura defect detection method with blurred edges in an LCD screen. The method realizes accurate detection of defects by adding noise, and is applicable to screens of different shapes and sizes; Filling is performed before detection and masking is performed after detection, so as to avoid the influence of bleeding points on detection results; for the problem of large image resolution, downsampling is used to reduce the image size to the original one

It not only saves the time of image processing, but also retains the valid data in the image; after the entire image processing process is completed, the result image is restored to the original size by means of upsampling, which is convenient for on-site operators to view, analyze and judge.

Description

A Mura Defect Detection Method with Blurred Edges in LCD Screens

technical field

The invention relates to the field of machine vision detection, and more particularly, to a Mura defect detection method with blurred edges in an LCD screen.

Background technique

LCD screens are usually composed of multiple materials and substrate layers that are bonded together. Often during the bonding process, various contaminants, air bubbles, migrations or other defects may enter the LCD screen to cause Mura defects. The existence of Mura defects seriously affects the yield of LCD screens. There are currently three mainstream detection methods for Mura defects with blurred edges. First, traditional detection methods rely on the human eye. This method has shortcomings such as slow detection speed, low efficiency, inability to meet high-speed automated production lines, and low detection accuracy. Secondly, in the edge detection scheme, because the edge of this defect is very blurred, the gray value of the pixel points in the edge area is in a "gradual" state, so the edge cannot be effectively extracted and the defect cannot be effectively detected. The detection effect of the edge detection scheme is shown in Figure 1. Finally, the front-background separation scheme, the existing front-background separation scheme expects to process the input image to obtain a foreground image containing defects and a background image without defects, and then compare and differentiate. However, due to the blurred edge of the defect, both the foreground image and the background image obtained by the front-background separation operation contain the defect or do not have the defect, so the defect cannot be detected by the subsequent contrast difference operation. The detection effect of the front-background separation scheme is shown in Figure 2. In general, the three mainstream methods are insufficient for Mura defects with blurred edges in LCD screens, and it is difficult to achieve the expected results. The expected effect is shown in Figure 3.

The prior art discloses a patent for an online automatic detection method for Mura defects on a mobile phone TFT-LCD screen. The patent first collects the screen image of the mobile phone to be detected through a CCD industrial camera; then the to-be-detected image is extracted, geometrically corrected and filtered. Preprocessing is used to obtain the TFT-LCD screen area in the image; then the screen area is divided into blocks, and the adaptive local enhancement algorithm is used to enhance the Mura defect in the image according to the gray distribution characteristics of each sub-image block; finally, the threshold value is used Mura defects in images were extracted by method and morphological open operation. The invention can automatically identify Mura defects with low contrast and blurred edges, has high accuracy and strong robustness, can effectively solve the problems of high manual detection cost, low efficiency and low accuracy in the production process, and is useful for improving mobile phone TFT-LCD The production efficiency and quality of the screen are of great significance. However, there are few reports in this patent that the problem of defect edge blurring can be effectively solved by adding appropriate noise into the differential.

SUMMARY OF THE INVENTION

The invention provides a method for detecting Mura defects with blurred edges in an LCD screen. The method can effectively solve the problem of blurred edges of defects by adding appropriate noise into the difference, so as to realize the effective detection of Mura defects with blurred edges, and improve the detection efficiency. accuracy.

In order to achieve above-mentioned technical effect, technical scheme of the present invention is as follows:

A method for detecting Mura defects with blurred edges in an LCD screen, comprising the following steps:

S1: Extract the region of interest in the image by image cropping;

S2: Perform Gaussian filtering on the image in step S1 to filter out Gaussian noise;

S3: down-sampling the image after step S2 to reduce the resolution of the image;

S4: perform edge filling on the image after step S3, and fill the non-screen area of the edge area;

S5: Select a certain position in the image after step S4 to add noise;

S6: Differentiate the noise-added image after step S5 and the image obtained in step S4, and enhance the result image after the difference;

S7: Perform binarization and expansion processing on the image obtained in step S6 to screen out defects;

S8: Perform up-sampling and masking on the image obtained in step S7, restore the resolution of the image, and shield the false detection of the non-screen area.

Further, in the step S1, since the LCD screen has various shapes and sizes, in order to ensure that all areas of the product to be tested can be photographed during the detection process of the actual industrial site, the field of view of the camera must be larger than The area to be tested will leave a certain bleeding position in the edge area of the screen. It is necessary to crop the area to be tested on the LCD screen for subsequent processing. Image cropping is performed by manually inputting the screen edge coordinates for cropping and automatically obtaining the edge coordinates for cropping.

Further, in the step S2, the specific operation of Gaussian filtering is: scan each pixel in the image with a template, and replace the grayscale of the center pixel of the template with the weighted average grayscale value of the pixel in the neighborhood determined by the template. value, select a template with a kernel of 45×45 and a standard deviation of 0 for filtering.

利用原图中虚拟点四周的四个真实存在的像素点的灰度值来共同决定目标图中的对应像素点的灰度值，将虚拟点对应的灰度值计算出来，具体地：Further, in the step S3, the downsampling mode of bilinear interpolation is adopted to reduce the height and width of the image to the original ones respectively.

The gray value of the four real pixels around the virtual point in the original image is used to jointly determine the gray value of the corresponding pixel in the target image, and the gray value corresponding to the virtual point is calculated, specifically:

f(dx,0)=f(0,0)*(1-dx)+f(1,0)*dx (1)

f(dx,1)=f(0,1)*(1-dx)+f(1,1)*dx (2)

f(dx,dy)=f(dx,0)*(1-dy)+f(dx,1)*dy (3)

Where f(0,0) represents the gray value of the coordinate (0,0), f(0,1) represents the gray value of the coordinate (0,1), and f(1,0) represents the coordinate of (1) ,0) grayscale value, f(1,1) represents the grayscale value with coordinates (1,1), f(dx,0) represents the grayscale value with coordinates (dx,0), f(dx, 1) Represents the grayscale value with coordinates (dx,1), and f(dx,dy) represents the grayscale value with coordinates (dx,dy).

Further, in the step S4, the filling process is column-by-column filling. First, the average gray level of the 100th pixel point to the 300th pixel point in a certain column is obtained, and then the gray level is found by traversing all the pixel points in the column. All the pixels in the column whose gray value is lower than a certain threshold, make the gray value of these pixels equal to the obtained gray mean value.

Further, in the step S5, when adding noise, it needs to be performed column by column from left to right, and the pixel mean value of each column of pixels is calculated in turn, and the obtained pixel mean value is used as the added noise value, and every two pixels in this column. Click to add the calculated pixel mean.

Further, in the step S6, the obtained noise value is smaller than the gray value of the white group defect, and larger than the gray value of the black group defect, and the noise value is not much different from the normal gray value; Differences with noise-filled plots can detect white mass defects:

Differential detection of black group defects is performed between the noise-filled image and the non-noise-filled image. The larger gray value in the obtained image is where the defect is located, and the smaller gray value is no defect. In order to increase the number of defects. The gray value gap between the place and the non-defective place is enhanced by histogram equalization or normalization. After enhancement through histogram equalization or normalization, the gray value gap between the defective area and the non-defective area is widened. , which is convenient for further processing;

When the pixel grayscale of the image changes randomly and the image histogram is uneven, the histogram equalization makes the image histogram roughly flat;

The principle of normalization is to calculate the gray value of each pixel in the image and map it to the range of 0-255. The minimum gray value before mapping is the lower limit of the range after mapping, and the largest gray value before mapping is the lower limit after mapping. Upper limit of range:

img(n,m) in equation (4) refers to the gray value of the pixel with coordinates (n, m) in the image, min_img is the minimum value of all pixel gray values in the image, and max_img is the gray value of the pixel in the image. The maximum value of the gray value of all pixel points, normalization is a linear transformation, after normalization, the gray value will be mapped to 0-255, which can increase the contrast.

Further, in the step S7, after the difference between the image without noise and the image with noise added or the difference map is enhanced, the place with the largest gray value in the image is the place where the defect is, and the gray value of other places is the place where the defect is located. The difference from the maximum value is at least 10, and the binarization method is used to set the gray value of the pixel whose gray value is greater than a certain threshold to 255, and set it to 0 in other areas. It is composed of multiple white short lines, and the white short lines are condensed into white balls by the expansion operation;

The specific operation method of expansion is to use a rectangle with a width m and a height n as a template, and do the following processing for each pixel x in the image: the pixel x is placed in the center of the template, and according to the size of the template, it traverses all other items covered by the template. Pixel, modify the value of pixel x to be the largest value among all pixels, connect and extend the prominent points on the periphery of the image.

所以需要通过上采样操作使图像恢复原有的尺寸；上采样选择双三次插值法，在这种方法中，(x,y)的灰度值可以通过矩形网格中最近的16个采样点的加权平均得到。Further, in the step S8, due to the down-sampling operation, the width and height of the image are reduced to the original

Therefore, it is necessary to restore the original size of the image through the upsampling operation; the bicubic interpolation method is selected for the upsampling. In this method, the gray value of (x, y) can pass through the nearest 16 sampling points in the rectangular grid. weighted average is obtained.

经过下采样，图像的分辨率降至1759×793。Preferably, a Gaussian filter kernel with a size of 45×45 and a standard deviation of 0 is used to perform Gaussian filtering. The resolution of the Gaussian filtering result graph is 8797×3965. The larger image resolution affects the speed of subsequent image processing. and width are reduced to the original

After downsampling, the resolution of the image is reduced to 1759×793.

Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

既节省了图像处理的时间又保留了图像中的有效数据；在整个图像处理过程完成之后，采用上采样的方式将结果图恢复到原有尺寸，方便现场操作工人进行查看、分析与判断。The present invention can effectively detect Mura defects with blurred edges at any position or size in an LCD screen of any shape and size by adding appropriate noise at an appropriate position, thereby improving the detection accuracy and reducing the missed detection rate and over-detection rate of detection. Accurate detection of defects by adding noise, and it is applicable to screens of different shapes and sizes; for bleeding points in the edge area of the screen, fill in before detection and mask after detection, so as to avoid bleeding The influence of points on the detection results; for the problem of large image resolution, downsampling is used to reduce the image size to the original

It not only saves the time of image processing, but also retains the valid data in the image; after the entire image processing process is completed, the result image is restored to the original size by means of upsampling, which is convenient for on-site operators to view, analyze and judge.

Description of drawings

Fig. 1 is an edge detection scheme detection effect diagram in the prior art;

Fig. 2 is a front-background separation detection effect diagram in the prior art;

Fig. 3 is an expected detection effect diagram in the prior art;

Fig. 4 is the flow chart of the method of the present invention;

Figure 5 is an image taken by a camera;

6 is a schematic diagram of a bilinear interpolation process;

Fig. 7 is a schematic diagram of expansion operation;

Figure 8 shows the effect of image cropping

9 is a schematic diagram of a mask template;

Figure 10 is a schematic diagram of the downsampling effect;

Figure 11 is the overall filling effect diagram;

Figure 12 is an effect diagram of adding noise;

Fig. 13 is a binarization diagram of a white mass defect with blurred edges;

Fig. 14 is a black group binarization map with blurred edges;

Figure 15 is an expansion diagram of a white mass defect with blurred edges;

Figure 16 is an expansion diagram of a black group defect with blurred edges;

FIG. 17 is a graph of detection results.

Detailed ways

The accompanying drawings are for illustrative purposes only, and should not be construed as limitations on this patent;

In order to better illustrate this embodiment, some parts of the drawings are omitted, enlarged or reduced, which do not represent the size of the actual product;

It will be understood by those skilled in the art that some well-known structures and their descriptions may be omitted from the drawings.

The technical solutions of the present invention will be further described below with reference to the accompanying drawings and embodiments.

Example 1

As shown in Figure 4, a method for detecting Mura defects with blurred edges in an LCD screen includes the following steps:

S1: Extract the region of interest in the image by image cropping;

S2: Perform Gaussian filtering on the image in step S1 to filter out Gaussian noise;

S3: down-sampling the image after step S2 to reduce the resolution of the image;

S4: perform edge filling on the image after step S3, and fill the non-screen area of the edge area;

S5: Select a certain position in the image after step S4 to add noise;

S6: Differentiate the noise-added image after step S5 and the image obtained in step S4, and enhance the result image after the difference;

S7: Perform binarization and expansion processing on the image obtained in step S6 to screen out defects;

S8: Perform up-sampling and masking on the image obtained in step S7, restore the resolution of the image, and shield the false detection of the non-screen area.

In step S1, since the LCD screen has various shapes and sizes, in order to ensure that all areas of the product to be tested can be photographed during the detection process of the actual industrial site, the field of view of the camera must be larger than the area to be tested, so A certain bleeding position will be left in the edge area of the screen, and the area to be tested on the LCD screen needs to be cropped out for subsequent processing. Image cropping is performed by manually inputting the screen edge coordinates for cropping or automatically obtaining the edge coordinates for cropping.

For each pixel in step S2, the weighted average gray value of the pixel in the neighborhood determined by the template is used to replace the gray value of the center pixel of the template, and a template with a kernel of 45×45 and a standard deviation of 0 is used for filtering.

利用原图中虚拟点四周的四个真实存在的像素点的灰度值来共同决定目标图中的对应像素点的灰度值，将虚拟点对应的灰度值计算出来，具体地：In step S3, the downsampling mode of bilinear interpolation is adopted to reduce the height and width of the image to the original ones respectively.

The gray value of the four real pixels around the virtual point in the original image is used to jointly determine the gray value of the corresponding pixel in the target image, and the gray value corresponding to the virtual point is calculated, specifically:

f(dx,0)=f(0,0)*(1-dx)+f(1,0)*dx (1)

f(dx,1)=f(0,1)*(1-dx)+f(1,1)*dx (2)

f(dx,dy)=f(dx,0)*(1-dy)+f(dx,1)*dy (3)

Where f(0,0) represents the gray value of the coordinate (0,0), f(0,1) represents the gray value of the coordinate (0,1), and f(1,0) represents the coordinate of (1) ,0) grayscale value, f(1,1) represents the grayscale value with coordinates (1,1), f(dx,0) represents the grayscale value with coordinates (dx,0), f(dx, 1) Represents the grayscale value with coordinates (dx,1), and f(dx,dy) represents the grayscale value with coordinates (dx,dy).

In step S4, the filling process is column-by-column filling. First, the average gray level of the 100th pixel point to the 300th pixel point in a certain column is obtained, and then all the gray levels in the column are found by traversing all the pixel points in the column. For pixels whose value is lower than a certain threshold, let the gray value of these pixels be equal to the obtained gray mean.

In step S5, when adding noise, it needs to be performed column by column from left to right, and the pixel mean value of each column of pixel points is calculated in turn, and the obtained pixel mean value is used as the added noise value, and the calculated value is added every two pixel points in this column. pixel mean.

In step S6, the obtained noise value is smaller than the gray value of the white group defect, larger than the gray value of the black group defect, and the noise value is not much different from the normal gray value; Differences can be performed to detect white lump defects:

Differential detection of black group defects is performed between the noise-filled image and the non-noise-filled image. The larger gray value in the obtained image is where the defect is located, and the smaller gray value is no defect. In order to increase the number of defects. The gray value gap between the place and the non-defective place is enhanced by histogram equalization or normalization. After enhancement through histogram equalization or normalization, the gray value gap between the defective area and the non-defective area is widened. , which is convenient for further processing;

When the pixel grayscale of the image changes randomly and the image histogram is uneven, the histogram equalization makes the image histogram roughly flat;

The principle of normalization is to calculate the gray value of each pixel in the image and map it to a certain range (such as 0-255), and the place with the lowest gray value before mapping is the lower limit of the range after mapping, and the gray value before mapping The highest place is mapped to the upper limit of the range:

img(n,m) in equation (4) refers to the gray value of the pixel with coordinates (n, m) in the image, min_img is the minimum value of all pixel gray values in the image, and max_img is the gray value of the pixel in the image. The maximum value of the gray value of all pixel points, normalization is a linear transformation, after normalization, the gray value will be mapped to 0-255, which can increase the contrast.

In step S7, after the difference between the image without noise and the image with noise added or the difference map is enhanced, the place with the largest gray value in the image is where the defect is, and the gray value in other places is at least different from the maximum value. 10. Use the binarization method to set the gray value of the pixel whose gray value is greater than a certain threshold to 255, and set it to 0 in other areas. In the result of binarization, each defect is composed of multiple white short lines. Composition, using expansion operation to condense short white lines into white balls;

The specific operation method of expansion is to use a rectangle with a width m and a height n as a template, and do the following processing for each pixel x in the image: the pixel x is placed in the center of the template, and according to the size of the template, it traverses all other items covered by the template. Pixel, modify the value of pixel x to be the largest value among all pixels, connect and extend the prominent points on the periphery of the image.

所以需要通过上采样操作使图像恢复原有的尺寸；上采样选择双三次插值法，在这种方法中，(x,y)的灰度值可以通过矩形网格中最近的16个采样点的加权平均得到。In step S8, due to the down-sampling operation, the width and height of the image are reduced to the original

Therefore, it is necessary to restore the original size of the image through the upsampling operation; the bicubic interpolation method is selected for the upsampling. In this method, the gray value of (x, y) can pass through the nearest 16 sampling points in the rectangular grid. weighted average is obtained.

Example 2

As shown in Figure 4, a Mura defect detection method with blurred edges in an LCD screen is divided into eight steps to realize the detection process. The first step is image cropping, which is used to extract the region of interest in the image. The second step of Gaussian filtering is to filter out Gaussian noise. The third step is downsampling, reducing the resolution of the image. The fourth step is edge padding, which fills the non-screen area of the edge area. The fifth step is to add the right amount of noise in the right place. The sixth step is to perform a difference between the image without noise and the image with added noise, and enhance the resulting image after the difference. The seventh step is to perform binarization and dilation processing to screen out defects. The eighth step is to perform upsampling and masking to restore the resolution of the image and mask false detections in non-screen areas. Through the processing of the above eight steps, Mura defects with blurred edges at any position and any size in an LCD screen of any shape and size can be accurately detected. The important key technical points in each step will be introduced in detail below.

1. Image cropping

LCD screens have various shapes and sizes. In order to ensure that all areas of the product to be tested can be photographed during the inspection process in the actual industrial site, the field of view of the camera must be larger than the area to be tested, so it will be at the edge of the screen. Leave a certain bleeding position, and the image captured by the camera is shown in Figure 5. Due to the existence of the bleeding position, it will not only affect the detection speed but also affect the detection effect. Therefore, before the formal detection starts, the area to be tested on the LCD screen needs to be cut out for subsequent processing. Image cropping can be done by manually inputting the screen edge coordinates or automatically finding the edge coordinates for cropping. Both methods can extract regions of interest, and both meet the requirements of subsequent processing.

2. Gaussian filter

Due to the influence of the lighting environment during imaging, the images captured by the camera always contain Gaussian noise. Gaussian noise refers to a class of noise whose probability density function follows a Gaussian distribution. The present invention uses Gaussian filtering to eliminate the influence of Gaussian noise. Gaussian filtering is a linear smoothing filter that is widely used in the noise reduction process of image processing. Gaussian filtering is a process of weighted averaging of the entire image. The gray value of each pixel is obtained by weighted averaging of the gray value of itself and other pixels in the neighborhood. The specific operation of Gaussian filtering is: scan each pixel in the image with a template (or convolution, mask), and replace the gray value of the center pixel of the template with the weighted average gray value of the pixel in the neighborhood determined by the template. value. The present invention selects a template with a kernel of 45×45 and a standard deviation of 0 for filtering processing.

3. Downsampling

将图像的高度和宽度缩小也有助于提升后续的算法运行速度。双线性插值是一种比较好的图像缩放算法，它充分的利用了原图中虚拟点四周的四个真实存在的像素点的灰度值来共同决定目标图中的对应像素点的灰度值，可以将虚拟点对应的灰度值计算出来，计算原理如图6所示。Because the blurred contrast of the edge of the defect is low, the gray value of the edge pixel is in a state of "gradation" and "transition", and the area to be detected is large. Therefore, the downsampling mode of bilinear interpolation is used to reduce the height and width of the image to the original ones.

Reducing the height and width of the image also helps to speed up subsequent algorithms. Bilinear interpolation is a good image scaling algorithm. It makes full use of the grayscale values of four real pixels around the virtual point in the original image to jointly determine the grayscale of the corresponding pixel in the target image. The gray value corresponding to the virtual point can be calculated, and the calculation principle is shown in Figure 6.

According to the calculation principle of bilinear interpolation, the following equation can be obtained from the coordinate relationship in Figure 6:

f(dx,0)=f(0,0)*(1-dx)+f(1,0)*dx (1)

f(dx,1)=f(0,1)*(1-dx)+f(1,1)*dx (2)

f(dx,dy)=f(dx,0)*(1-dy)+f(dx,1)*dy (3)

Where f(0,0) represents the gray value of the coordinate (0,0), f(0,1) represents the gray value of the coordinate (0,1), and f(1,0) represents the coordinate of (1) ,0) grayscale value, f(1,1) represents the grayscale value with coordinates (1,1), f(dx,0) represents the grayscale value with coordinates (dx,0), f(dx, 1) Represents the grayscale value with coordinates (dx,1), and f(dx,dy) represents the grayscale value with coordinates (dx,dy).

4. Edge padding

The LCD screen has various shapes and sizes such as rectangular screen, notch screen, water drop screen, etc. However, since the shape of the down-sampled image can only be a rectangle, there are still some bleeding points in the edge area of the LCD screen. In order to avoid these bleeding points from affecting the subsequent image processing process, a filling operation is required to fill the edge area with appropriate gray values. The filling process is column-by-column filling. First, the grayscale mean of the 100th pixel to the 300th pixel in a column is obtained, and then all the pixels in the column are traversed to find that all the grayscale values in the column are lower than a certain value. Pixel points with a threshold value, let the gray value of these pixels be equal to the obtained gray mean value.

5. Add Noise

Adding noise is the most important part of the present invention and a key step in the entire image processing process. Because there is a certain gap between the gray value of the defect and the gray value of other pixels in the column, appropriate noise can be added to the result image in step 4 for detection. When adding noise, it needs to be performed column by column from left to right, and the pixel mean value of each column of pixels is calculated in turn.

6. Image difference and enhancement

Through experiments, it is found that in general, the calculated noise value is smaller than the gray value of the white group defect, and larger than the gray value of the black group defect, and the noise value is not much different from the normal gray value. White lump defects can be detected by differencing the non-noise-filled plot with the noise-filled plot. Differentiating the noise-filled map with the one that is not noise-filled can detect black mass defects. In the obtained image, the larger gray value is where the defect is located, and the smaller gray value means there is no defect. In order to increase the gray value gap between the defective place and the non-defective place, histogram equalization or normalization can be used for enhancement. After enhancement through histogram equalization or normalization, the gray value gap between the defective area and the non-defective area can be widened, which is convenient for further processing.

Histogram equalization is an important application of grayscale transformation, which is efficient and easy to implement, and is widely used in image enhancement. When the pixel gray level of the image changes randomly, and the image histogram appears uneven, a certain algorithm can be used for histogram equalization to make the image histogram roughly flat. Simply put, histogram equalization is a method of enhancing image contrast by stretching the range of pixel intensity distributions.

The principle of normalization is to calculate the gray value of each pixel in the image and map it to a certain range (such as 0-255). The minimum gray value before mapping is the lower limit of the range after mapping, and the maximum gray value before mapping is mapped. followed by the upper limit of the range.

img(n,m) in equation (4) refers to the gray value of the pixel with coordinates (n, m) in the image, min_img is the minimum value of all pixel gray values in the image, and max_img is the gray value of the pixel in the image. The maximum value of all pixel gray values. Normalization is a linear transformation. After normalization, the gray value is mapped to 0-255, which can increase the contrast.

7. Binarization and Dilation Processing

After the difference between the image without noise and the image with noise added or the enhancement of the difference map, the place with the largest gray value in the image is where the defect is, and the gray value in other places is at least 10 different from the maximum value, so it can be The binarization method is used to set the gray value of the pixel whose gray value is greater than a certain threshold to 255, and the gray value of other areas to 0. Each defect in the binarized result graph is composed of multiple white short lines, and the dilation operation can be used to condense the white short lines into white balls.

The specific operation method of expansion is to use a rectangle with a width m and a height n as a template, and do the following processing for each pixel x in the image: the pixel x is placed in the center of the template, and according to the size of the template, it traverses all other items covered by the template. Pixel, modify the value of pixel x to be the largest value among all pixels. The result of this operation is to connect and extend the salient points on the periphery of the image. Figure 7 shows a schematic diagram of the expansion operation of the 3×3 template.

8. Upsampling and Masking

所以需要通过上采样操作使图像恢复原有的尺寸。上采样可以选择双三次插值法，双三次插值是二维空间中最常用的插值方法。在这种方法中，(x,y)的灰度值可以通过矩形网格中最近的16个采样点的加权平均得到。Due to the downsampling operation, the width and height of the image are reduced to the original

Therefore, it is necessary to restore the original size of the image through the upsampling operation. Upsampling can choose bicubic interpolation, which is the most commonly used interpolation method in two-dimensional space. In this method, the gray value of (x,y) can be obtained by the weighted average of the nearest 16 sampling points in a rectangular grid.

Because the screen is not a very regular rectangle, the bleeding point in the image does not belong to the place that needs to be detected, so in order to avoid the detection of the bleeding point in the final detection result, the mask method can be used for processing. The mask uses a selected image, figure or object to block a part or all of the area of the processed image to control the area or process of image processing. The specific image or object used for overlay is called a mask or stencil.

Example 3

After starting the detection, it is necessary to place the product to be detected in the center of the camera's field of view, then turn on the power to light up the screen, adjust the gray value of the pixels on the screen to a certain threshold (for example, 128), and then use the camera to capture images. The collected images are transferred to the computer, and the collected images are shown in Figure 5. In Figure 5, the black area on the periphery of the screen is the bleeding point, which does not belong to the range that needs to be detected. The screen area can be cut out by manually inputting the screen edge coordinates or automatically calculating the screen edge coordinates. The effect diagram after cutting is shown in Figure 8. According to the cropped screen rendering, edge information of the screen can be obtained, so that the mask template picture shown in FIG. 9 can be created for use in subsequent mask steps.

经过下采样，图像的分辨率降至1759×793.分辨率对比情况如图10所示。Since the Gaussian noise in the image will interfere with the subsequent image processing process, a Gaussian filter kernel with a size of 45×45 and a standard deviation of 0 is used for Gaussian filtering. The resolution of the Gaussian filtering result image is 8797×3965, and the image resolution It greatly affects the speed of subsequent image processing, so the downsampling method is used to reduce the length and width of the image to the original one.

After downsampling, the resolution of the image is reduced to 1759×793. The resolution comparison is shown in Figure 10.

Because the screen is not a standard rectangular screen, there are still some bleeding spots in the edge area of the screen. These bleeding spots are prone to false detections and therefore need to be filled. The present invention uses a column-by-column filling method to firstly calculate the grayscale mean or grayscale mode of the 100th element to the 300th element in the first column, and assigns the area to be filled with the calculated result. And so on, from left to right, fill column by column. The overall filling effect diagram is shown in Figure 11.

Adding noise is a key step of the present invention, and the specific implementation idea is to first obtain the grayscale mean value of each column of pixels from left to right, and the obtained grayscale mean value is the added noise value. Add the calculated noise value every 2 pixels in each column of pixels. The effect diagram after adding is shown in Figure 12.

The difference between the image with added noise and the image without noise can be used to detect white group defects, and the difference between the image without added noise and the image with added noise can be used to detect black group defects. By setting an appropriate binarization threshold, Mura defects with blurred edges can be screened out, as shown in Figures 13 and 14.

The defects in the binarized image are composed of multiple short lines. In order to better meet the actual detection requirements, an expansion kernel with a core of 3 is used for expansion processing. The effect diagrams after processing are shown in Figures 15 and 16. In order to avoid false detections caused by non-screen areas, use the dilated result maps shown in Figures 15 and 16 to perform a difference with the mask template map shown in Figure 9, and filter the contour area greater than 200 pixels in the differential result map. The contour of the point can be obtained as shown in Figure 17.

The same or similar reference numbers correspond to the same or similar parts;

The positional relationship described in the accompanying drawings is only for exemplary illustration, and should not be construed as a limitation on this patent;

Obviously, the above-mentioned embodiments of the present invention are only examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. For those of ordinary skill in the art, changes or modifications in other different forms can also be made on the basis of the above description. There is no need and cannot be exhaustive of all implementations here. Any modifications, equivalent replacements and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the claims of the present invention.

Claims (10)

1. a fuzzy Mura defect detection method of edge in LCD screen, is characterized in that, comprises the following steps: S1: Extract the region of interest in the image by image cropping; S2: Perform Gaussian filtering on the image in step S1 to filter out Gaussian noise; S3: down-sampling the image after step S2 to reduce the resolution of the image; S4: perform edge filling on the image after step S3, and fill the non-screen area of the edge area; S5: Select a certain position in the image after step S4 to add noise; S6: Differentiate the noise-added image after step S5 and the image obtained in step S4, and enhance the result image after the difference; S7: Perform binarization and expansion processing on the image obtained in step S6 to screen out defects; S8: Perform up-sampling and masking on the image obtained in step S7, restore the resolution of the image, and shield the false detection of the non-screen area. 2. The Mura defect detection method with blurred edge in LCD screen according to claim 1, is characterized in that, in described step S1, because LCD screen has various shapes and sizes, the detection process in actual industrial site In order to ensure that all areas of the product to be tested can be photographed, the field of view of the camera must be larger than the area to be tested, so there will be a certain amount of bleeding at the edge of the screen, and the area to be tested on the LCD screen needs to be cropped out for subsequent processing. , the image is cropped by manually inputting the screen edge coordinates for cropping or automatically obtaining the edge coordinates for cropping. 3. the blurred Mura defect detection method of edge in LCD screen according to claim 2, is characterized in that, in described step S2, the concrete operation of Gaussian filter is: scan each pixel in the image with a template, use The weighted average gray value of the pixels in the neighborhood determined by the template replaces the gray value of the center pixel of the template, and the template with a kernel of 45×45 and a standard deviation of 0 is used for filtering. 4. the Mura defect detection method of edge blurring in LCD screen according to claim 3, is characterized in that, in described step S3, adopts the downsampling mode of bilinear interpolation, the height and width of image are respectively reduced to original of

The gray value of the four real pixels around the virtual point in the original image is used to jointly determine the gray value of the corresponding pixel in the target image, and the gray value corresponding to the virtual point is calculated, specifically: f(dx,0)=f(0,0)*(1-dx)+f(1,0)*dx (1) f(dx,1)=f(0,1)*(1-dx)+f(1,1)*dx (2) f(dx,dy)=f(dx,0)*(1-dy)+f(dx,1)*dy (3) Where f(0,0) represents the gray value of the coordinate (0,0), f(0,1) represents the gray value of the coordinate (0,1), and f(1,0) represents the coordinate of (1) ,0) grayscale value, f(1,1) represents the grayscale value with coordinates (1,1), f(dx,0) represents the grayscale value with coordinates (dx,0), f(dx, 1) Represents the grayscale value with coordinates (dx,1), and f(dx,dy) represents the grayscale value with coordinates (dx,dy). 5. The Mura defect detection method with blurred edge in LCD screen according to claim 4, is characterized in that, in described step S4, filling process is to fill column by column, first find out the 100th pixel in a certain column to the 1st pixel. The grayscale mean of 300 pixels, and then by traversing all the pixels in the column to find all the pixels in the column whose grayscale values are lower than a certain threshold, so that the grayscale values of these pixels are equal to the obtained grayscale mean. 6. The Mura defect detection method with blurred edge in LCD screen according to claim 5, it is characterized in that, in described step S5, need to carry out row by row from left to right when adding noise, calculate the pixel of each row of pixel points successively Mean, the obtained pixel mean value is used as the added noise value, and the calculated pixel mean value is added every two pixel points in this column. 7. The method for detecting Mura defects with blurred edges in an LCD screen according to claim 6, wherein in the step S6, the obtained noise value is smaller than the gray value of the white group defect, and is smaller than the gray value of the black group defect. If the value is large, the noise value is not much different from the normal gray value; the white lump defect can be detected by the difference between the image without noise filling and the image filled with noise: Differential detection of black group defects is performed between the noise-filled image and the image without noise-filling. In the obtained image, the larger gray value is where the defect is located, and the smaller gray value is no defect. In order to increase the number of defects The difference between the gray value of the place where there is no defect and the place without defect is enhanced by histogram equalization or normalization. After the enhancement is performed by histogram equalization or normalization, the gray value of the defect area and the non-defect area are opened up. The gap is convenient for further processing; When the pixel grayscale of the image changes randomly and the image histogram is uneven, the histogram equalization makes the image histogram roughly flat; The principle of normalization is to calculate the gray value of each pixel in the image and map it to the range of 0-255. The minimum gray value before mapping is the lower limit of the range after mapping, and the largest gray value before mapping is mapped as The upper limit of the range:

img(n,m) in equation (4) refers to the gray value of the pixel with coordinates (n, m) in the image, min_img is the minimum value of all pixel gray values in the image, and max_img is the gray value of the pixel in the image. The maximum value of the gray value of all pixel points, normalization is a linear transformation, after normalization, the gray value will be mapped to 0-255, which can increase the contrast. 8. The Mura defect detection method with blurred edge in LCD screen according to claim 7, is characterized in that, in described step S7, after the image that does not add noise and the image that adds noise carries out difference or the enhancement to difference map , the place with the largest gray value in the image is where the defect is located, and the gray value in other places is at least 10 different from the maximum value. The gray value of the pixel whose gray value is greater than a certain threshold is set to 255, other areas are set to 0, each defect in the binarized result graph is composed of multiple white short lines, and the expansion operation is used to condense the white short lines into white balls; The specific operation method of expansion is to use a rectangle with a width m and a height n as a template, and do the following processing for each pixel x in the image: the pixel x is placed in the center of the template, and according to the size of the template, it traverses all other items covered by the template. Pixel, modify the value of pixel x to be the largest value among all pixels, connect and extend the prominent points on the periphery of the image. 9. The Mura defect detection method with blurred edge in LCD screen according to claim 8, is characterized in that, in described step S8, due to downsampling operation, the width and height of image is reduced to original

Therefore, it is necessary to restore the original size of the image through the upsampling operation; the bicubic interpolation method is selected for the upsampling. In this method, the gray value of (x, y) can pass through the nearest 16 sampling points in the rectangular grid. weighted average is obtained. 10. The Mura defect detection method with blurred edges in the LCD screen according to any one of claims 1-9, characterized in that, selecting a Gaussian filter kernel with a size of 45×45 and a standard deviation of 0 to carry out Gaussian filtering, and Gaussian filtering The resolution of the resulting image is 8797×3965, and the image resolution greatly affects the speed of subsequent image processing. The length and width of the image are reduced to the original one.

After downsampling, the resolution of the image is reduced to 1759×793.

Images