CN116721082B - Display color mura defect detection method based on channel separation and filtering - Google Patents

Abstract

The present invention discloses a display screen color Mura defect detection method based on channel separation and filtering, which is applied to the field of image processing technology, and aims at the problem that the existing technology has low accuracy in detecting color Mura. The present invention first converts the collected RGB image of the LCD display screen into the Lab color space; secondly, the L, a and b channels are subjected to mean template filtering to obtain color and brightness images, and then the Weber contrast feature map of the Lab three-channel image is calculated; then, the frequency domain contrast sensitive function filtering template is used to perform frequency domain filtering on the Lab three-channel contrast feature map to obtain the Lab three-channel feature map sensitive to the human eye, and the ab channel feature map is fused; finally, the feature map statistics of the L channel and ab color fusion are calculated, and the color Mura defect detection result is obtained by threshold segmentation. The method of the present invention realizes accurate and efficient color Mura defect detection of the display screen.

Description

Display color mura defect detection method based on channel separation and filtering

Technical Field

The present invention belongs to the field of image processing technology, and in particular relates to a display screen color Mura defect detection technology.

Background Art

Machine vision systems are widely used in production and life due to their stability and reliability. Using machine vision systems to detect the quality of display screens brings convenience to industrial production, improves production efficiency and saves costs.

Mura defects are characterized by low contrast and no fixed shape. With the rapid development of display technology in the past decade, the current display Mura defect detection method is mainly a defect detection method based on background reconstruction: one is to use frequency domain-based reconstruction, such as Fan proposed Automatic detection of Mura defect in TFT-LCD based on regression diagnostics [J]; the other is a reconstruction method based on fitting, such as Yan Chengchen proposed a TFT-LCD detection algorithm combining weighted template difference map and bilateral filtering [J], Journal of Electronic Measurement and Instrumentation. However, the reconstruction method based on frequency domain cannot effectively obtain the defect-free background information of color Mura in the frequency domain, which makes this method unable to separate defects from the background; the reconstruction method based on fitting does not take color features into account, so it directly fits the grayscale image and cannot obtain the defect area. Therefore, this type of method has low accuracy in detecting color Mura. The human eye's perception of color information greatly exceeds the perception level of brightness information, so the use of color filtering analysis is an effective technical approach for display color Mura defect detection.

Summary of the invention

In order to solve the above technical problems, the present invention proposes a display screen color Mura defect detection method based on channel separation and filtering.

The technical solution adopted by the present invention is: a display color Mura defect detection method based on channel separation and filtering, comprising:

Step S1, performing Lab space channel separation on the image to be processed; specifically: for the image to be processed, converting the RGB image color space into the Lab color space to obtain separated brightness and color channel images;

Step S2, Weber contrast feature map calculation; specifically: mean filtering the separated brightness and color channel images respectively, and then obtaining the contrast feature maps of the brightness and color channels according to the Weber contrast calculation formula;

Step S3, contrast sensitivity function frequency domain filtering; specifically: construct frequency domain contrast sensitivity function filtering templates for brightness and color channels respectively, use fast Fourier transform to convert the contrast feature maps of brightness and color channels into frequency domain space, and multiply them with contrast sensitivity functions respectively, and obtain brightness and color channel feature maps that are sensitive to the human eye after inverse transformation;

Step S4, color feature map fusion and adaptive threshold segmentation; specifically: fuse the color channel feature maps that are sensitive to the human eye to obtain a color fusion feature map that is sensitive to the human eye, use the mean and standard deviation of the two-channel feature maps that are sensitive to the human eye and the color feature fusion map that is sensitive to the human eye to perform threshold segmentation, perform an OR operation on the segmentation results of the two feature maps, and obtain a Mura defect detection result.

Beneficial effects of the present invention: The present invention separates the brightness and color of the input image, and then uses contrast sensitive functions to filter the color and brightness contrast maps respectively, and uses a comprehensive method of Lab color space, human eye contrast sensitivity function and feature decomposition to detect color Mura defects, thereby achieving accurate detection of brightness Mura and color Mura defects on the display screen.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a flow chart of a method for detecting color Mura defects in a display screen based on channel separation and filtering according to the present invention.

Figure 2 is an example of the result of converting the original image to Lab space mean filtering.

Figure 3 shows an example of the contrast sensitivity function filtering results.

Figure 4 shows a sample of Mura defect detection results.

DETAILED DESCRIPTION

Referring to FIG. 1 , in the embodiment described below, the display color Mura defect detection method based on channel separation and filtering is implemented in the following steps:

Step S1: Input image Lab space channel separation, specifically including the following steps:

S1-1, convert the collected RGB image of the LCD display screen into the Lab color space;

The Lab color space calculation is performed on the input image. The calculation formula is as follows:

Where L is the brightness channel, a is the red and green color channel, b is the blue and yellow color channel, X, Y, Z represent the color coordinates in the XYZ color space, and f(X), f(Y1), and f(Z) represent the color distribution function in the XYZ color space. Get the separated brightness L and color a and b channel images;

S2: Weber contrast feature map calculation, the specific steps include the following steps:

S2-1, mean filtering processing,

The L, a and b channel images are filtered using a mean filter template to suppress the interference of the background texture of the display screen. The image processing results are shown in Figure 2.

S2-2, global mean calculation,

Calculate the global mean of the L, a and b channel images after mean filtering respectively. The calculation formula is as follows:

Where μ _weber represents the global mean, M is the width of the input image, N is the length of the input image, and pi _,j represents the grayscale value at coordinate (i,j) in the image. Calculate the mean of the L channel, a channel, and b channel respectively.

S2-3, Weber contrast calculation,

According to the Weber contrast calculation formula, the contrast feature maps of L, a and b channels are obtained.

The Weber contrast calculation formula is as follows:

Wherein I(x, y) is the gray value of the image of a specific channel, I _b1 (x, y) is the background brightness of the image of the specific channel, and C _w (x, y) is the contrast of the image of the specific channel.

Replace I _b1 (x, y) with the global mean of L, a and b, and calculate _CL , _Ca and _Cb respectively. _CL represents the contrast feature map of L channel, _Ca is the contrast feature map of a channel, and _Cb is the contrast feature map of b channel.

S3: Contrast sensitivity function frequency domain filtering. The result diagram is shown in FIG3 . The specific steps include the following steps.

S3-1, frequency domain contrast sensitivity function construction,

Construct the frequency domain contrast sensitivity functions of L, a and b channels respectively. The contrast sensitivity functions are as follows:

Among them, f represents the frequency variable in the frequency domain, csf _L (f) represents the contrast sensitivity function of the L channel in the frequency domain, csf _a (f) represents the contrast sensitivity function of the a channel in the frequency domain, and csf _b (f) represents the contrast sensitivity function of the b channel in the frequency domain.

S3-2, Fourier transform of channel feature map,

The L, a and b channel feature maps that are sensitive to the human eye are Fourier transformed, and the low-frequency information is moved to the center of the image to obtain the L, a and b channel feature maps in the frequency domain, which are recorded as C _{L_w} , _{Ca_w} and C _{b_w} respectively.

S3-3, feature map frequency domain filtering,

The obtained L, a and b channel feature maps in the frequency domain are multiplied by the contrast sensitivity function respectively to obtain the channel filtering feature results. The calculation formula is as follows.

Where C _{csf_L} is the L channel feature map sensitive to the human eye, C _{csf_a} is the a channel feature map sensitive to the human eye, C _{csf_b} is the b channel feature map sensitive to the human eye, idft means inverse Fourier transform, and the result is shown in Figure 3.

Step S4: color feature map fusion and adaptive threshold segmentation, specifically including the following steps.

S4-1, color feature map fusion,

The a and b channel feature maps that are sensitive to the human eye are fused to obtain a color fusion feature map that is sensitive to the human eye. The calculation formula is as follows:

Where C _{csf_col} represents the color fusion feature map that the human eye is sensitive to.

S4-2, adaptive threshold segmentation of brightness and color,

The mean and standard deviation of the L channel feature map that the human eye is sensitive to and the color feature fusion map that the human eye is sensitive to are used for threshold segmentation. The calculation formula is as follows:

th _col =μc _col +K1δc _col

th _L = μc _L + K2δc _L

Among them, th _col is the segmentation threshold of the color fusion feature map, K1 is a parameter, μc _col is the mean of the color fusion feature map, μc _col is the standard deviation of the color fusion feature map, th _L is the segmentation threshold of the L channel feature map sensitive to the human eye, K2 is a parameter, μc _L is the mean of the L channel feature map sensitive to the human eye, and δc _L is the standard deviation of the L channel feature map sensitive to the human eye. In this embodiment, K1 takes a value of 1 and K2 takes a value of 3.

The segmentation results of the two feature maps are ORed to obtain the Mura defect detection result, as shown in FIG4 .

Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the scope of protection of the present invention is not limited to such specific statements and embodiments. For those skilled in the art, the present invention may have various changes and variations. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of the claims of the present invention.

Claims (5)

1. A display color Mura defect detection method based on channel separation and filtering, characterized by comprising: Step S1, performing Lab space channel separation on the image to be processed; specifically: for the image to be processed, converting the RGB image color space into the Lab color space to obtain separated brightness and color channel images; Step S2, Weber contrast feature map calculation; specifically: mean filtering the separated L, a and b channel images respectively, and then obtaining the contrast feature maps of the L, a and b channels according to the Weber contrast calculation formula; Step S3, contrast sensitivity function frequency domain filtering; specifically: construct frequency domain contrast sensitivity function filter templates for L, a and b channels respectively, use fast Fourier transform to convert the contrast feature maps of L, a and b channels into frequency domain space, and multiply them with contrast sensitivity functions respectively, and obtain L, a and b channel feature maps that are sensitive to the human eye after inverse transformation; the frequency domain contrast sensitivity function filter template expression of L channel constructed in step S3 is: Wherein, csf _L (f) represents the contrast sensitivity function of the L channel in the frequency domain, and f represents the frequency variable in the frequency domain; The frequency domain contrast sensitivity function filter template expression of channel a constructed in step S3 is: Wherein, csf _a (f) represents the contrast sensitivity function of channel a in the frequency domain; The frequency domain contrast sensitivity function filter template expression of the b channel constructed in step S3 is: Where csf _b (f) represents the contrast sensitivity function of the b channel in the frequency domain; Step S4, color feature map fusion and adaptive threshold segmentation; specifically: fuse the a channel feature map and the b channel feature map that are sensitive to the human eye to obtain a color fusion feature map that is sensitive to the human eye, use the mean and standard deviation of the L channel feature map that is sensitive to the human eye and the color fusion feature map that is sensitive to the human eye to perform threshold segmentation, perform an OR operation on the segmentation results of the two feature maps, and obtain a Mura defect detection result. 2. The display color Mura defect detection method based on channel separation and filtering according to claim 1 is characterized in that step S2 specifically comprises the following sub-steps: S21, mean filtering processing, using the mean filtering template to filter the L, a and b channel images respectively; S22, global mean calculation, respectively calculate the global mean of the L, a and b channel images after mean filtering, the calculation formula is as follows: Wherein, μ _weber represents the global mean, M is the width of the input specific channel image, N is the length of the input specific channel image, and pi _,j represents the gray value of the coordinate (i,j) in the specific channel image; S23, Weber contrast calculation, according to the Weber contrast calculation formula, obtain the contrast feature map of L, a and b channels; the Weber contrast calculation formula is as follows: Where I(x, y) is the grayscale value of the image, I _b1 (x, y) is the global mean of the specific channel image, and C _w (x, y) is the contrast feature map of the specific channel image. 3. The display color Mura defect detection method based on channel separation and filtering according to claim 2 is characterized in that in step S4, the color channel feature map sensitive to the human eye is fused to obtain a color fusion feature map sensitive to the human eye; the calculation formula is: Among them, C _{csf_a} represents the a-channel feature map, C _{csf_b} represents the b-channel feature map, and C _{csf_col} represents the color fusion feature map that the human eye is sensitive to. 4. The display color Mura defect detection method based on channel separation and filtering according to claim 1 is characterized in that the segmentation threshold calculation formula of the color fusion feature map that the human eye is sensitive to is: th _col =μ _ccol +K1δC _col Among them, th _col is the segmentation threshold of the color fusion feature map, K1 is the parameter, μc _col is the mean of the color fusion feature map, and δc _col is the standard deviation of the color fusion feature map. 5. According to the display color Mura defect detection method based on channel separation and filtering according to claim 4, it is characterized in that the segmentation threshold calculation formula of the L channel feature map is: th _L = μc _L + K2δc _L Among them, _thL is the segmentation threshold of the L channel feature map sensitive to the human eye, K2 is a parameter, _μcL is the mean of the L channel feature map sensitive to the human eye, and _δcL is the standard deviation of the L channel feature map sensitive to the human eye.