Display screen detection method and system based on image processing

Abstract

The invention relates to the technical field of image processing, in particular to a display screen detection method and system based on image processing, and the method comprises the following steps: obtaining original RGB pixel data of a display screen optical image, converting the original RGB pixel data into a brightness channel value, a chromaticity a channel value and a chromaticity b channel value of a CIELAB space, and integrating the chroma a channel value and the chroma b channel value. According to the method, the RGB pixel data of the optical image of the display screen is extracted and converted into the CIELAB space numerical value, so that the influence of the shooting color difference of different devices on the detection result is reduced, and the color evaluation is more objective; through integration of chroma a and b channel data and Gaussian smoothing processing, random noise interference is suppressed, and stability of chroma characteristics is improved. A local chromaticity standard deviation is calculated based on a sliding window to position a chromaticity high variation area in a display area, and the sensitivity to a local defect area is enhanced.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a display screen detection method and system based on image processing. Background Art

[0002] The field of image processing technology involves the comprehensive process of using computers to collect, preprocess, analyze and identify image data, mainly including various technical links such as image enhancement, feature extraction, image segmentation, target detection, image compression and pattern recognition.

[0003] Existing technologies fail to fully consider the characteristics of localized chromaticity variations and directly employ global analysis methods, making it difficult to effectively distinguish random noise from true defect areas, increasing the error in defect location. Furthermore, they ignore the information correlation between frequency and chromaticity spaces and lack a cross-validation mechanism for spectral and spatial features. This results in insufficient accuracy in identifying periodic defects, easily leading to false detections or overlooking critical defects, impacting the reliability of display production quality control. Therefore, improvements are needed. Summary of the Invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a display screen detection method and system based on image processing.

[0005] In order to achieve the above object, the present invention adopts the following technical solution, a display screen detection method based on image processing, comprising the following steps: Obtaining original RGB pixel data of the display screen optical image, converting the original RGB pixel data into luminance channel values ​​and chroma a channel values ​​and chroma b channel values ​​in the CIELAB space, integrating the chroma a channel values ​​and chroma b channel values, applying a Gaussian filter for smoothing, and obtaining a smoothed chroma channel image; Based on the chromaticity a channel values ​​of the smoothed chromaticity channel image, a sliding window is set on the display panel pixel information to calculate the standard deviation of the chromaticity a channel values ​​within the window to obtain local chromaticity standard deviation data; based on the local chromaticity standard deviation data, a standard deviation threshold is set to identify pixel areas that exceed the standard deviation threshold, and obtain the location of the high chromaticity variation area; Based on the smoothed chromaticity channel map, selecting an optical uniformity reference block of the display area that is not marked by the chromaticity high variation area location and calculating an average chromaticity value to obtain an average chromaticity parameter of the reference area; and based on the average chromaticity parameter of the reference area, calculating and obtaining a relative chromaticity deviation map; Based on the chromaticity b-channel values ​​of the smoothed chromaticity channel image, a two-dimensional Fourier transform operation is performed on the chromaticity b-channel values ​​of the display light-emitting unit array to obtain a chromaticity channel Fourier spectrum. Based on the chromaticity channel Fourier spectrum, frequency peaks that are different from the reference spectrum are detected to obtain candidate frequency points of periodic artifacts.

[0006] Preferably, the step of obtaining the smoothed chromaticity channel image is: Obtain the original RGB pixel data of the display screen's optical image, collect the entire screen's pixel matrix, and store the image data into the RGB three-channel matrix according to the red channel value, green channel value, and blue channel value of each pixel point to form the original RGB pixel data; According to the original RGB pixel data, the RGB values ​​are mapped to the CIEXYZ color space, and then the luminance channel value, chroma a channel value, and chroma b channel value of each pixel in the CIELAB space are calculated based on the mapping relationship of the CIEXYZ color space to obtain the CIELAB space channel value; According to the CIELAB spatial channel values, the chromaticity a channel values ​​and the chromaticity b channel values ​​of all pixels are extracted, the chromaticity a channel values ​​and the chromaticity b channel values ​​are aligned and combined into a two-dimensional vector matrix according to the pixel positions, and the two-dimensional vector matrix is ​​convolved and smoothed using a two-dimensional Gaussian kernel to form a smoothed chromaticity channel map.

[0007] Preferably, the steps for obtaining the local chromaticity standard deviation data are: Based on the chromaticity a channel value of the smoothed chromaticity channel image, with each pixel in the display panel pixel information as the center point, a square sliding window of a fixed size is set, so that the window slides row by row and column by column to cover each pixel position on the smoothed chromaticity channel image, forming a sliding window positioning matrix; According to the sliding window positioning matrix, traverse and extract the chroma a channel values ​​corresponding to all pixels covered by the window in each sliding window, calculate the average value of all the extracted chroma a channel values ​​within each sliding window, and record the average value to form a chroma a channel average value matrix within the sliding window; Based on the chroma a channel average value matrix in the sliding window, taking the average value in each sliding window as a benchmark, traverse all the extracted chroma a channel values ​​in each sliding window, calculate the difference between each chroma a channel value and the average value in the corresponding sliding window, calculate the average of the squares of the differences and take the square root to obtain the local chroma standard deviation data.

[0008] Preferably, the steps for obtaining the location of the high chromaticity variation region are: Based on the local chromaticity standard deviation data, the numerical distribution of the standard deviation of the chromaticity a channel values ​​within all sliding windows is counted, the numerical distribution median in the local chromaticity standard deviation data is extracted, and the numerical distribution median is used as a basic reference to set a standard deviation threshold for identifying high variation areas to form a standard deviation threshold; According to the standard deviation threshold, each value in the local chromaticity standard deviation data is judged one by one. If the standard deviation of the chromaticity a channel value is greater than or equal to the standard deviation threshold, the corresponding sliding window coverage area is marked as a candidate high-variance pixel area, otherwise it is marked as a normal area, forming a candidate high-variance area marking map; Based on the candidate high-variability region label map, adjacent and continuous candidate high-variability pixel regions are extracted and merged, region boundaries are determined, and region position information is established to obtain the chromaticity high-variability region positioning.

[0009] Preferably, the steps for obtaining the average chromaticity parameter of the reference area are: Based on the chromaticity high variation region location, identifying continuous display regions not marked by the chromaticity high variation region location in the smoothed chromaticity channel image, excluding regions that overlap or are adjacent to the chromaticity high variation region location, to form candidate display regions; According to the candidate display areas, statistics are collected and compared on the changes in chromaticity values ​​within each candidate display area, the standard deviation of the chromaticity values ​​of the pixels within each area is calculated, and the candidate display area with the lowest standard deviation and the smallest change in internal chromaticity values ​​is determined to form an optical uniformity reference block; Based on the optical uniformity reference block, the chromaticity a channel value and the chromaticity b channel value of each pixel point in the reference block are traversed and extracted, all the extracted chromaticity a channel values ​​and chromaticity b channel values ​​are averaged respectively, and the average chromaticity a channel value and the average chromaticity b channel value are calculated and recorded to obtain the average chromaticity parameter of the reference area.

[0010] Preferably, the steps of obtaining the relative chromaticity deviation map are: Based on the average chromaticity parameter of the reference area, the chromaticity a channel value and the chromaticity b channel value of each pixel point are read pixel by pixel from the smoothed chromaticity channel image, and the column coordinate value and the row coordinate value of the corresponding pixel point in the image are extracted to obtain a pixel point attribute set including the chromaticity channel and the pixel position; Calculating the relative chromaticity deviation between each pixel and the average chromaticity parameter of the reference area according to the pixel attribute set; Based on the relative chromaticity deviation value of each pixel, the relative chromaticity deviation value is filled into the corresponding position of the two-dimensional matrix of the same dimension as the smoothed chromaticity channel map according to the row and column coordinate order of each pixel in the image to generate a relative chromaticity deviation map.

[0011] Preferably, the steps of acquiring the Fourier spectrum of the chrominance channel are: Based on the smoothed chromaticity channel map, extracting chromaticity b channel values ​​corresponding to all pixel points of the display light-emitting unit array row by row and column by column, and storing the chromaticity b channel values ​​in a matrix according to their original row and column positions to form a chromaticity b channel value matrix; According to the chroma b channel value matrix, performing a two-dimensional Fourier transform on the matrix, performing frequency domain mapping on the chroma b channel values ​​of all pixels along the row and column directions of the matrix, obtaining an intermediate frequency domain representation matrix in complex form, and forming a two-dimensional frequency domain mapping matrix; Based on the two-dimensional frequency domain mapping matrix, the amplitude of each frequency position is calculated respectively and stored in the spectrum matrix according to the corresponding frequency position. The distribution state of the amplitude is reconstructed through the frequency coordinates in the two-dimensional space to obtain the chrominance channel Fourier spectrum.

[0012] Preferably, the step of obtaining the candidate periodic artifact frequency points is: Based on the Fourier spectrum of the chrominance channel, traverse each frequency point in the spectrum graph, extract its horizontal frequency index, vertical frequency index and corresponding amplitude information, count the amplitude sets of all frequency points, and calculate the mean and standard deviation of the amplitude sets, screen out frequency points with amplitudes greater than the mean plus one standard deviation as preliminary suspected frequency peaks, and generate a candidate set of frequency peaks; Calculating the frequency space deviation verification strength based on the frequency peak candidate set and the pixel chromaticity deviation value of the relative chromaticity deviation map; Based on the frequency-space deviation verification strength, the frequency-space deviation verification strength is traversed and compared with the set verification strength reference threshold, and all frequency peaks whose frequency-space deviation verification strength is higher than the verification strength reference threshold are identified to obtain candidate periodic artifact frequency points.

[0013] The present invention also provides a display screen detection system, comprising: An image preprocessing module obtains the original RGB pixel data of the display screen optical image, converts the original RGB pixel data into the brightness channel value and the chroma a channel value and chroma b channel value in the CIELAB space, integrates the chroma a channel value and the chroma b channel value, applies a Gaussian filter to smooth, and obtains a smoothed chroma channel image; A chromaticity variation detection module sets a sliding window on the display panel pixel information, based on the chromaticity a channel values ​​of the smoothed chromaticity channel image, to calculate the standard deviation of the chromaticity a channel values ​​within the window, thereby obtaining local chromaticity standard deviation data. Based on the local chromaticity standard deviation data, a standard deviation threshold is set to identify pixel areas exceeding the standard deviation threshold, thereby obtaining the location of areas with high chromaticity variation; a chromaticity deviation calculation module, based on the smoothed chromaticity channel map, selecting an optical uniformity reference block of the display area that is not marked by the chromaticity high variation area location, calculating an average chromaticity value, obtaining an average chromaticity parameter of the reference area, and calculating a relative chromaticity deviation map based on the average chromaticity parameter of the reference area; The periodic artifact detection module performs a two-dimensional Fourier transform operation on the chromaticity b channel values ​​of the display light-emitting unit array based on the chromaticity b channel values ​​of the smoothed chromaticity channel image to obtain a chromaticity channel Fourier spectrum. Based on the chromaticity channel Fourier spectrum, the module detects frequency peaks that are different from the reference spectrum to obtain candidate frequency points of periodic artifacts.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are: The present invention extracts RGB pixel data of the display screen's optical image and converts it into CIELAB space values, thereby reducing the impact of color differences between different devices on the test results and making color evaluation more objective. By integrating the chromaticity a and b channel data and performing Gaussian smoothing, random noise interference is suppressed and the stability of the chromaticity features is improved. Local chromaticity standard deviations are calculated based on a sliding window to locate areas with high chromaticity variation within the display area, thereby enhancing sensitivity to local defective areas. Optically uniform areas are then selected as a benchmark for calculating relative chromaticity deviations, thereby improving the accuracy of abnormal area detection. A two-dimensional Fourier transform is then performed on the chromaticity b channel values ​​to capture the spectral characteristics of periodic artifacts and identify periodic defects caused by the display structure or manufacturing process. This achieves more accurate identification of optical quality defects in the display screen and ensures product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 Schematic diagram of the steps of the present invention. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0017] See also Figure 1 The present invention provides a technical solution, a display screen detection method based on image processing, comprising the following steps: Obtaining original RGB pixel data of the display screen optical image, converting the original RGB pixel data into luminance channel values ​​and chroma a channel values ​​and chroma b channel values ​​in the CIELAB space, integrating the chroma a channel values ​​and chroma b channel values, applying a Gaussian filter for smoothing, and obtaining a smoothed chroma channel image; Based on the chromaticity a channel value of the smoothed chromaticity channel image, a sliding window is set on the display panel pixel information to calculate the standard deviation of the chromaticity a channel value within the window to obtain local chromaticity standard deviation data. Based on the local chromaticity standard deviation data, a standard deviation threshold is set to identify pixel areas that exceed the standard deviation threshold to obtain the location of high chromaticity variation areas; Based on the smoothed chromaticity channel map, select the display area optical uniformity reference block that is not marked by the chromaticity high variation area positioning to calculate the average chromaticity value, obtain the average chromaticity parameter of the reference area, and calculate the relative chromaticity deviation map based on the average chromaticity parameter of the reference area; Based on the chromaticity b-channel values ​​of the smoothed chromaticity channel image, a two-dimensional Fourier transform operation is performed on the chromaticity b-channel values ​​of the display light-emitting unit array to obtain the chromaticity channel Fourier spectrum. Based on the chromaticity channel Fourier spectrum, the frequency peaks that are different from the reference spectrum are detected to obtain candidate frequency points of periodic artifacts.

[0018] The steps for obtaining the smoothed chromaticity channel map are: Obtain the original RGB pixel data of the display screen's optical image, collect the entire screen's pixel matrix, and store the image data into the RGB three-channel matrix according to the red channel value, green channel value, and blue channel value of each pixel point to form the original RGB pixel data; According to the original RGB pixel data, the RGB values ​​are mapped to the CIEXYZ color space, and then the brightness channel value, chroma a channel value, and chroma b channel value of each pixel in the CIELAB space are calculated based on the mapping relationship of the CIEXYZ color space to obtain the CIELAB space channel value; According to the CIELAB space channel values, the chromaticity a channel values ​​and chromaticity b channel values ​​of all pixels are extracted, and the chromaticity a channel values ​​and chromaticity b channel values ​​are aligned according to the pixel positions and combined into a two-dimensional vector matrix. The two-dimensional vector matrix is ​​convolved and smoothed with a two-dimensional Gaussian kernel to form a smoothed chromaticity channel map.

[0019] Specifically, the original RGB pixel data of the display screen optical image is obtained by using a high-resolution industrial camera, such as a CCD or CMOS sensor with at least the number of pixels covering the display screen, to shoot the lit display screen in a preset darkroom environment, ensuring that there is no interference from external stray light. When shooting, the display screen usually displays a pure color image such as full white or full gray to expose the subtle differences between pixels. The camera's parameters such as aperture, focal length, white balance and exposure time are pre-calibrated and fixed. For example, for a 1920x1080 resolution display screen, the camera resolution can be set to 2048x1536 or higher for oversampling. The exposure time is set according to the screen brightness. For example, at a brightness of 100cd / m², the exposure time is set according to the screen brightness. The light time may be set to 1 / 60 second, and the captured image is organized into a matrix representing the pixels of the entire screen. The dimension of the matrix is ​​the height (H) multiplied by the width (W) of the display screen, where each pixel contains three independent color channel intensity values, namely the red (R) channel value, the green (G) channel value, and the blue (B) channel value. These values ​​are usually 8-bit or 10-bit depth, such as integers in the range of 0-255. Subsequently, these image data are separated by color channel and stored in three independent two-dimensional matrices, namely the red channel value matrix (dimension HxW), the green channel value matrix (dimension HxW), and the blue channel value matrix (dimension HxW). These three matrices together constitute the original RGB pixel data.

[0020] According to the original RGB pixel data, the RGB channel values ​​of each pixel are first normalized. For example, if the original value range is 0-255, they are divided by 255 to map them to the floating point range of 0-1. Then, the normalized RGB values ​​are mapped to the CIEXYZ color space through the standard color space conversion formula. This conversion is usually based on a 3x3 conversion matrix, which is composed of the chromaticity coordinates of the RGB primary colors used by the display and the chromaticity coordinates of the reference white point (for example, the commonly used D65 light source, whose CIEXYZ tristimulus values ​​are ) is determined, and the specific conversion process can be expressed as multiplying the normalized RGB column vector of each pixel by the conversion matrix to obtain the corresponding CIEXYZ tristimulus value (X, Y, Z). Subsequently, based on the obtained values ​​of the CIEXYZ color space and the preset reference white point (such as the aforementioned D65 Value), further calculate the brightness channel value of each pixel in the CIELAB color space ( ), chroma a channel value ( ) and the chroma b channel value ( ), the calculation process follows the standard definition of CIELAB color space to obtain the CIELAB space channel value.

[0021] According to the CIELAB spatial channel value, from each pixel In the data, extract the chroma a channel value of all pixels ( ) and the chroma b channel value ( ), these extracted The channel values ​​are arranged according to their pixel positions in the original image to form a two-dimensional matrix, which is recorded as , similarly, The channel values ​​are arranged to form another two-dimensional matrix, denoted as , the dimensions of these two matrices are consistent with the pixel height and width of the original image. Then, the two matrices are aligned at the pixel position and combined into a two-dimensional vector matrix, each element of which is a vector containing Then, in order to suppress noise and smooth the drastic changes in chrominance in space, a two-dimensional Gaussian kernel is used to perform convolution smoothing on the two-dimensional vector matrix. Specifically, Matrix and The matrix applies a two-dimensional Gaussian filter with the same parameters, the parameters of the two-dimensional Gaussian kernel, including the standard deviation of the kernel ( ) and the kernel size (such as width and height), need to be set in advance, the standard deviation The choice of depends on the desired smoothing scale. For example, if you want to smooth out small noises smaller than 3 pixels, you can set is 1.0 pixel, the size of the Gaussian kernel is usually set to cover range, for example, , the kernel size can be chosen to be Pixels, the convolution operation will convert each pixel The value is replaced by its neighborhood The Gaussian weighted average of the values The same processing is performed on the values. After convolution smoothing, the smoothed chroma a channel numerical matrix and the smoothed chroma b channel numerical matrix are obtained respectively. These two smoothed matrices together constitute the smoothed chroma channel map.

[0022] The steps for obtaining local chromaticity standard deviation data are: Based on the chromaticity a channel value of the smoothed chromaticity channel image, with each pixel in the display panel pixel information as the center point, a square sliding window of fixed size is set, so that the window slides row by row and column by column to cover each pixel position on the smoothed chromaticity channel image, forming a sliding window positioning matrix; According to the sliding window positioning matrix, traverse and extract the chroma a channel values ​​corresponding to all pixels covered by the window in each sliding window, calculate the average value of all extracted chroma a channel values ​​within each sliding window, and record the average value to form the chroma a channel average value matrix within the sliding window; Based on the average value matrix of the chromaticity a channel in the sliding window, taking the average value in each sliding window as the benchmark, traverse all the extracted chromaticity a channel values ​​in each sliding window, calculate the difference between each chromaticity a channel value and the average value in the corresponding sliding window, calculate the average of the squares of the differences and take the square root to obtain the local chromaticity standard deviation data.

[0023] Specifically, based on the chromaticity a channel value of the smoothed chromaticity channel image, the specific operation is to first determine a square sliding window of a fixed size. The size of the window is, for example, set to The pixel is selected based on the typical physical size of the display defect to be detected and the resolution of the display. For example, if the minimum chromaticity unevenness area to be detected is about 5x5 pixels, the window size can be set to 5x5 pixels. This size is selected by testing and analyzing a large number of sample images containing known defect types to most effectively capture the local chromaticity changes of the target size while suppressing the influence of single pixel noise. After setting, each pixel in the smoothed chromaticity channel image is used as a potential center point, starting from the first pixel in the upper left corner of the image. The center of the sliding window is aligned with the current pixel point, and then the window slides in sequence from left to right and then from top to bottom, moving one pixel each time until the center of the window traverses all pixel positions on the smoothed chromaticity channel map. For pixels at the edge of the image, when the sliding window cannot be completely accommodated in the image, an edge padding strategy can be adopted, such as using the image edge pixel values ​​for constant filling or mirror filling, or only performing effective calculations in the area where the window is completely located inside the image, thereby defining all valid window center positions. The specific coverage range of each window on the image or its center coordinate set that are systematically scanned and determined together constitute the sliding window positioning matrix based on which subsequent processing is based.

[0024] According to the sliding window positioning matrix, for each specific sliding window position defined by the matrix, the following operations are performed: first, according to the precise pixel area covered by the current sliding window on the smoothed chromaticity channel map, the chromaticity a channel values ​​corresponding to all pixels in the area are traversed and extracted. For example, for a 5x5 pixel sliding window, the chromaticity a channel values ​​of 25 pixels in the window will be extracted, and these extracted values ​​are collected to form a temporary value set. Then, all the chromaticity a channel values ​​in this temporary value set are averaged, that is, all the extracted chromaticity a channel values ​​in the window are averaged. The chroma a channel values ​​are summed up and then divided by the total number of pixels in the window (for example, for a 5x5 window, the total number is 25) to obtain the average value of the chroma a channel value within the sliding window. The calculated average value is associated with the center pixel position of the sliding window and recorded. This extraction and average value calculation process is repeated for all window positions indicated by the sliding window positioning matrix. Finally, the calculated average values ​​corresponding to all window positions are organized according to their corresponding center positions in the original image to form a two-dimensional matrix of the same size as the smoothed chroma channel map, that is, the chroma a channel average value matrix within the sliding window.

[0025] Based on the average value matrix of the chromaticity a channel in the sliding window and the set of original chromaticity a channel values ​​previously extracted in each sliding window, the following calculation is performed for each sliding window position. First, the average chromaticity a channel value corresponding to the center position of the current window is obtained, which has been recorded in the average value matrix of the chromaticity a channel in the sliding window. The average value will be used as the basis for calculating the chromaticity change in the current window. Then, each pixel point in the area covered by the current sliding window is traversed again, and its original chromaticity a channel value in the smoothed chromaticity channel map is extracted. For each original chromaticity a channel value in the window, its difference from the window is calculated. The difference between the average chroma a channel values ​​is used to obtain a series of difference data. Then, each value in these difference data is squared, all the squared differences are added up, and then the average of the sum of these square differences is calculated, that is, divided by the total number of pixels in the window. Finally, the square root of the average value of the calculated square difference is taken. The result is the standard deviation of the chroma a channel value in the current sliding window. This standard deviation calculation process is performed for all sliding window positions, and the standard deviation values ​​calculated for each window are recorded and arranged according to the position of the window center pixel in the original image. Finally, the local chroma standard deviation data is summarized.

[0026] The steps for obtaining the location of high chromaticity variation areas are: Based on the local chromaticity standard deviation data, the numerical distribution of the standard deviation of the chromaticity a channel within all sliding windows is counted, the numerical distribution median in the local chromaticity standard deviation data is extracted, and the numerical distribution median is used as a basic reference to set the standard deviation threshold for identifying high variation areas to form a standard deviation threshold; According to the standard deviation threshold, each value in the local chromaticity standard deviation data is judged one by one. If the standard deviation of the chromaticity a channel value is greater than or equal to the standard deviation threshold, the corresponding sliding window coverage area is marked as a candidate high-variance pixel area, otherwise it is marked as a normal area, forming a candidate high-variance area marking map; Based on the candidate high-variability region labeling map, adjacent and continuous candidate high-variability pixel regions are extracted and merged, the region boundaries are determined, and the region position information is established to obtain the chromaticity high-variability region positioning.

[0027] Specifically, based on the local chromaticity standard deviation data, firstly, the standard deviations of the internal chromaticity a channel values ​​calculated by all sliding windows are collected to form a data set containing multiple standard deviation values. The data set is statistically analyzed to grasp its overall distribution characteristics. Specifically, these standard deviation values ​​are sorted from low to high, and then the median of the sorted data set is determined. For example, if there is standard deviation values, after sorting, if If is an odd number, the median is A value, if If it is an even number, the median is The value and The median of the calculated value distribution is used as the basis for setting the standard deviation threshold. The specific setting method of the standard deviation threshold is: Preset multiple Median of the numerical distribution, where the preset multiples It is a constant determined by experience or through analysis and calibration of a large number of sample display images (including known qualified and unqualified areas). It is intended to distinguish normal color fluctuations from significant variations that indicate defects. For example, if a batch of 100 display images with known conditions are analyzed and it is found that the color channel a standard deviation exceeds 2.8 times the median, it usually corresponds to color unevenness or mura defects that are perceptible to the human eye. Then this preset multiple is It can be set to 2.8. For example, if the median of the calculated local chromaticity standard deviation data is 0.5 (in CIELAB chromaticity a channel units), the standard deviation threshold will be set to The calculated 1.4 is the standard deviation threshold used for subsequent identification of high-variability regions.

[0028] According to the standard deviation threshold set in the previous step, for example, 1.4, and using the previously obtained local chromaticity standard deviation data (the data is a matrix, in which each element represents the chromaticity a channel standard deviation in the corresponding sliding window), the chromaticity a channel standard deviation of each sliding window is judged one by one, and at the same time, a binary image with the same size as the original RGB pixel data of the display optical image is initialized, called the candidate high-variance area marking map, and all pixel values ​​of this image are initially set to 0 (representing the normal area), and each standard deviation value in the local chromaticity standard deviation data is traversed. If the current standard deviation value is greater than or equal to the set standard deviation threshold (for example, the standard deviation of a window is 1.5, which is greater than or equal to 1.4), all pixels in the display image area actually covered by the sliding window are marked. Point, the corresponding pixel position in the candidate high variation region label map is marked as 1 (representing the candidate high variation pixel area). If a pixel is covered by multiple overlapping sliding windows exceeding the threshold, its value in the candidate high variation region label map is still 1. If the standard deviation of the chromaticity a channel of the current sliding window is less than the standard deviation threshold (for example, the standard deviation of a window is 0.8, which is less than 1.4), then the area covered by the sliding window does not change its state in the candidate high variation region label map (that is, it remains 0, or if it has been marked as 1 by other high variation windows, it remains 1). After completing the standard deviation judgment and marking of all sliding windows, the final candidate high variation region label map is a binary image, in which a pixel with a value of 1 indicates that it belongs to at least one sliding window coverage with high chromaticity variation.

[0029] Based on the candidate high-variability region labeling map, which is a binary image, pixels marked as 1 represent potential chromaticity high-variability points. These marked points are then clustered to extract and merge spatially adjacent and continuous candidate high-variability pixel regions. This process uses a connected component labeling algorithm. First, a list is initialized to store the final results, which is used to store the positioning information of each independent chromaticity high-variability region, and a "visited" labeling map of the same size as the candidate high-variability region labeling map is created. All elements are initialized to "unvisited". Then, starting from the upper left corner pixel of the candidate high-variability region labeling map, it is scanned row by row and column by column. When a pixel with a label value of 1 and a status of "unvisited" is encountered, it is regarded as a seed point of a new candidate high-variability region. Starting from the seed point, a graph traversal algorithm such as breadth-first search (BFS) or depth-first search (DFS) is used to check Find all pixels connected by 8-neighborhood (i.e., pixels sharing edges or corners are considered adjacent) and with the same label value of 1, and mark these found pixels as "visited". All sets of pixels marked as 1 connected in this way together constitute an independent, continuous candidate high-variance pixel region. For each independent region formed in this way, further determine its boundary information, for example, record the coordinates of the upper left and lower right corners of the minimum circumscribed rectangle of the region, or record the coordinate set of all pixels constituting the region, and calculate its geometric center, area and other attributes. After this information (such as boundary coordinates, pixel list, area, etc.) is integrated, it is recorded as a region location information and added to the list of region positioning. Repeat this scanning and region extraction process until all pixels in the candidate high-variance region label map have been visited. The final list is the chromatic high-variance region positioning.

[0030] The steps for obtaining the average chromaticity parameters of the reference area are: Based on the chromaticity high variation region positioning, the continuous display area not marked by the chromaticity high variation region positioning is identified in the smoothed chromaticity channel image, and the areas that overlap or are adjacent to the chromaticity high variation region positioning are excluded to form candidate display areas; Based on the candidate display areas, the chromaticity value changes within each candidate display area are counted and compared, the standard deviation of the pixel chromaticity values ​​within each area is calculated, and the candidate display area with the lowest standard deviation and the smallest internal chromaticity value change is determined to form an optical uniformity reference block; Based on the optical uniformity reference block, the chromaticity a channel value and the chromaticity b channel value of each pixel in the reference block are traversed and extracted, and all the extracted chromaticity a channel values ​​and chromaticity b channel values ​​are averaged respectively. The average chromaticity a channel value and the average chromaticity b channel value are calculated and recorded to obtain the average chromaticity parameter of the reference area.

[0031] Specifically, based on the chromaticity high variation region positioning, the positioning information includes a series of coordinates or pixel sets of regions identified as chromaticity unevenness. First, an initial mask with the same size as the image is generated on the smoothed chromaticity channel map, in which all pixels are initially marked as "available". Then, each marked region in the chromaticity high variation region positioning is traversed, and all pixels in these regions are marked as "unavailable" in the initial mask. Furthermore, in order to exclude neighboring regions that may be affected by the high variation region, a morphological dilation operation is performed on each marked chromaticity high variation region. The dilation distance is set to 5 pixels, for example. This distance is set based on experience to ensure that the selected reference region is far away from the edge influence of any known defect. For example, if a high variation region The boundary of the different region is at row 100. After dilation of 5 pixels, rows 95 to 105 may be considered as adjacent regions and excluded. All pixel areas covered by the dilation operation are also marked as "unavailable" in the initial mask. After completing the above exclusion operation, connected component analysis is performed on all pixel areas still marked as "available" in the initial mask to identify spatially continuous "available" areas. Each such connected component is regarded as a potential candidate display area, and these areas are screened to eliminate areas with too small an area. For example, a minimum area threshold is set to 5000 pixels. This threshold is set based on the total pixels of the display screen and the representativeness of the expected reference area. Only continuous "available" areas with an area larger than this threshold will be retained to form candidate display areas.

[0032] Based on the candidate display areas, i.e., a series of continuous display blocks that meet the size requirements and are not contaminated by the high chromaticity variation area and its adjacent areas, the chromaticity uniformity within each candidate display area is quantitatively evaluated. The specific operation is as follows: for a certain candidate display area, all pixels in the area are first traversed, and the chromaticity a channel value and chromaticity b channel value of each pixel are extracted from the smoothed chromaticity channel map. Based on these extracted values, the average chromaticity a channel value and the average chromaticity b channel value of all pixels in the area are calculated to obtain the average chromaticity point of the area. Then, for each pixel in the area, its chromaticity is calculated. The Euclidean distance between the pixel and the average chromaticity point of the region represents the degree of deviation of a single pixel from the average chromaticity of the region. The deviation distances of all pixels in the region are collected, and the standard deviation of these deviation distances is calculated. This standard deviation is used as an indicator to measure the change in chromaticity values ​​within the candidate display region. This calculation process is repeated for all candidate display regions to obtain the chromaticity standard deviation value corresponding to each candidate display region. Finally, the chromaticity standard deviation values ​​of all candidate display regions are compared, and the candidate display region with the lowest standard deviation value is selected. For example, if the standard deviation of candidate region A is 0.3, the standard deviation of candidate region B is 0.25, and the standard deviation of candidate region C is 0.35, then candidate region B is selected because its internal chromaticity value change is the smallest. This selected region is the optical uniformity reference block.

[0033] Based on the optical uniformity reference block, which is the area with the smallest internal chromaticity variation selected from all candidate display areas, the coordinate set of all pixel points contained in the optical uniformity reference block in the smoothed chromaticity channel map is first determined. Then, two accumulator variables are initialized to accumulate the sum of the chromaticity a channel value and the sum of the chromaticity b channel value respectively. At the same time, a counter is initialized to record the total number of pixels in the block. Then, each pixel point in the optical uniformity reference block is traversed. For the currently traversed pixel point, its corresponding chromaticity a channel value and chromaticity b channel value are accurately extracted from the smoothed chromaticity channel map. The extracted chromaticity a channel value is added to the accumulator of the sum of chromaticity a channel values, and the extracted chromaticity b channel value is added to the accumulator of the sum of chromaticity b channel values. At the same time, the counter is increased by one. After traversing all pixels in the optical uniformity reference block, the sum of chromaticity a channel values ​​is divided by the total number of pixels (that is, the value of the counter) to obtain the average chromaticity a channel value of the reference block. Similarly, the sum of chromaticity b channel values ​​is divided by the total number of pixels to obtain the average chromaticity b channel value of the reference block. These two calculated average values ​​together constitute the average chromaticity parameter of the reference area.

[0034] The steps to obtain the relative chromaticity deviation spectrum are: Based on the average chromaticity parameter of the reference area, the chromaticity a channel value and the chromaticity b channel value of each pixel are read pixel by pixel from the smoothed chromaticity channel image, and the column coordinate value and row coordinate value of the corresponding pixel in the image are extracted to obtain a pixel attribute set containing the chromaticity channel and the pixel position. According to the pixel attribute set, the relative chromaticity deviation value between each pixel and the average chromaticity parameter of the reference area is calculated. The calculation formula is: ; in, For the The relative chromaticity deviation value of each pixel, For the The chroma a channel value of each pixel, For the The chroma b channel value of each pixel, is the average value of the chroma a channel of the reference area, is the average value of the chroma b channel of the reference area, and Respectively The column coordinate value and row coordinate value of each pixel point, and are the column and row coordinate values ​​of the geometric center of the reference area, is the total number of columns of the chromaticity channel map after smoothing, is the total number of rows of the chromaticity channel map after smoothing, is the balance factor used to adjust the weight of spatial distance influence; Based on the relative chromaticity deviation value of each pixel, the relative chromaticity deviation value is filled into the corresponding position of the two-dimensional matrix of the same dimension as the smoothed chromaticity channel map according to the row and column coordinate order of each pixel in the image to generate a relative chromaticity deviation map.

[0035] Specifically, based on the average chromaticity parameter of the reference area and the previously obtained smoothed chromaticity channel map, the specific operation is to initialize a set for storing all pixel attributes, which will contain the chromaticity a channel value, chromaticity b channel value, column coordinate value and row coordinate value of each pixel in the image, and traverse each pixel position in the smoothed chromaticity channel map. The smoothed chromaticity channel map is composed of two independent two-dimensional matrices, which respectively store the smoothed chromaticity a channel value and the smoothed chromaticity b channel value of each pixel. For the first pixel in the image, pixel points, whose positions in the image are determined by their column coordinate values ​​(e.g. ) and row coordinate values ​​(e.g. ) is uniquely determined, first from the chromaticity a channel numerical matrix of the smoothed chromaticity channel image, according to and Read the corresponding chroma a channel value (denoted as ), then from the chromaticity b channel numerical matrix of the smoothed chromaticity channel image, according to the same and Read the corresponding chroma b channel value (denoted as ), the obtained , , and These four values ​​are stored as a data unit (such as a structure or tuple) in a predefined pixel attribute set. This process is performed sequentially for all pixels in the smoothed chromaticity channel image (from column coordinate 0 to the total number of columns minus 1, and row coordinate 0 to the total number of rows minus 1) until the relevant attributes of each pixel in the image are read and stored, finally forming a pixel attribute set containing the chromaticity channel and pixel position.

[0036] formula: The benefit of the formula is that it comprehensively considers the difference between the pixel point and the average chromaticity of the reference area in the CIELAB chromaticity space, as well as the distance between the pixel point and the geometric center of the reference area in physical space, and uses the balance factor The formula can more flexibly adapt to the detection requirements of different types of display defects. It can capture both local sudden color differences and large-scale gradual color non-uniformity, thus providing a relative color deviation metric that is more consistent with human visual perception. parameter The steps to obtain are: Representative The chromaticity a channel value of each pixel is directly extracted from the "pixel attribute set containing chromaticity channel and pixel position" generated in the previous step. Specifically, for any pixel in the image , the corresponding chroma a channel value is the value in CIELAB space after smoothing by Gaussian filter For example, for a pixel located at a specific position in an image, the chroma a channel value recorded in the "pixel attribute set" is 12.0.

[0037] parameter The steps to obtain are: Representative The chroma b channel value of each pixel is obtained in the same way as Similarly, it is also extracted from the "pixel attribute set containing chromaticity channels and pixel positions", which corresponds to the pixel point in the smoothed CIELAB space. Value, for example, for the above For the same pixel, the recorded chroma b channel value is -8.0.

[0038] parameter The steps to obtain are: The chromaticity a channel average value for the reference area is derived from the "reference area average chromaticity parameter" calculated in the previous step. Specifically, the chromaticity a channel average value for all pixels in the selected "optical uniformity reference block" is calculated. For example, the chromaticity a channel average value for a 100x100 pixel "optical uniformity reference block" is 10.0.

[0039] parameter The steps to obtain are: The average value of the chromaticity b channel of the reference area is also derived from the "average chromaticity parameter of the reference area". Its calculation method is the same as The difference is that it is the average value of the chromaticity b channel of all pixels in the "optical uniformity reference block". For example, for the same 100x100 pixel "optical uniformity reference block", the calculated chromaticity b channel average value is -5.0.

[0040] parameter The steps to obtain are: Representative The column coordinate value of a pixel point (usually corresponding to the width direction of the image). This value is extracted from the "pixel attribute set containing chromaticity channels and pixel positions" and represents the position index of the pixel point in the horizontal direction of the image. It usually starts counting from 0. For example, the column coordinate value of a pixel point is 300.

[0041] parameter The steps to obtain are: Representative The row coordinate value of the pixel point (usually corresponding to the height direction of the image) is obtained in the same way as Similarly, it is extracted from the "pixel attribute set containing chroma channels and pixel positions" to represent the position index of the pixel in the vertical direction of the image, usually counting starting from 0. For example, the row coordinate value of a pixel is 400.

[0042] parameter The steps to obtain are: Represents the geometric center column coordinate value of the reference area. This value is obtained by calculating the average value of the column coordinates of all pixels in the previously determined "optical uniformity reference block". If the "optical uniformity reference block" contains pixels, and the column coordinates of each pixel are ,but For example, an “optical uniformity reference block” consists of all pixels with column coordinates from 50 to 150 and row coordinates from 100 to 200, and the column coordinate value of its geometric center is .

[0043] parameter The steps to obtain are: Represents the geometric center row coordinate value of the reference area, and the calculation method is the same as Similarly, the average value of the row coordinates of all pixels in the “optical uniformity reference block” is calculated. If the row coordinate of each pixel is ,but For the “optical uniformity reference block” in the above example, the row coordinate value of its geometric center is .

[0044] parameter The steps to obtain are: Represents the total number of columns of the smoothed chromaticity channel image, that is, the width of the image (in pixels). This value is an inherent property of the display's optical image and is determined during image acquisition or preprocessing. For example, if the resolution of a display is 1920x1080, then .

[0045] parameter The steps to obtain are: Represents the total number of rows of the smoothed chroma channel map, that is, the height of the image (in pixels), and Similarly, it is an inherent property of the image. For a display with a resolution of 1920x1080, .

[0046] parameter The steps to obtain are: The balancing factor used to adjust the weight of spatial distance influence is determined by experimental testing and optimization of a series of sample images containing typical display defects. The goal is to achieve the best match between the calculated relative chromaticity deviation value and the perceived severity of display defects by the human eye. The specific method is to prepare a group (for example, 30) of display screen images with known defect types and levels (assessed by professional inspectors), and use different The relative chromaticity deviation map is calculated by multiplying the values ​​(e.g., from 10 to 200, with a step size of 5), extracting the statistical features of the map (such as the maximum deviation value, the area ratio of high deviation areas, etc.), and performing correlation analysis with the defect levels assessed manually. The one with the highest correlation or the best classification accuracy is selected. For example, the balance factor determined by this method for a specific model of OLED display detection The value is 75.

[0047] Calculation process: Taking a specific pixel as an example, the parameter values ​​are as follows: ; ; ; ; ; ; ; ; ; ; ; Calculate the chromaticity difference part: ; ; ; ; Sum of squared chromaticity differences ; Calculate the spatial distance part: ; ; Normalized column coordinate difference ; Normalized row coordinate difference ; Normalized squared column coordinate differences ; Normalized squared row coordinate difference ; Normalized sum of squared spatial distances ; Weighted spatial distance term: Weighted spatial distance ; Calculate the total relative colorimetric deviation value : ; ; ; ; The results show that for the selected pixel point, the relative chromaticity deviation value after comprehensively considering the chromaticity difference and spatial position difference is 4.222864. This value quantitatively represents the degree of perceptual difference of the pixel point relative to the reference uniform area. The larger the value, the more significant the difference in chromaticity of the pixel point and the reference area, or the greater the potential impact on overall visual uniformity due to its spatial position far from the reference center and the existence of a certain chromaticity difference.

[0048] Based on the relative chromaticity deviation value calculated for each pixel, the specific operation is to first create a new two-dimensional matrix whose dimension is the same as the total number of rows of the previously processed smoothed chromaticity channel map ( ) and the total number of columns ( ) is exactly the same, this new matrix is ​​used to store the final relative chromaticity deviation map, and the initial values ​​of all elements are set to zero or a specific background value. Then, the matrix containing each pixel obtained from the previous step is and its corresponding relative chromaticity deviation value In the data set, the information of each pixel is read one by one. For each pixel , get its column coordinate value in the original image and row coordinate values , and the calculated relative chromaticity deviation value , then, the The values ​​​​are filled in the newly created two-dimensional matrix corresponding to The position of the matrix Row, No. The column elements are assigned values This process is performed once for all pixels in the image to ensure that each pixel position is filled with its corresponding relative chromaticity deviation value. After all pixel data are filled, the two-dimensional matrix constitutes a relative chromaticity deviation map.

[0049] The steps to obtain the Fourier spectrum of the chrominance channel are: Based on the smoothed chromaticity channel map, the chromaticity b channel values ​​corresponding to all pixels of the display light-emitting unit array are extracted row by row and column by column, and the chromaticity b channel values ​​are sequentially stored in a matrix according to their original row and column positions to form a chromaticity b channel value matrix; According to the chrominance b channel value matrix, a two-dimensional Fourier transform is performed on the matrix, and the chrominance b channel values ​​of all pixels are frequency-domain mapped along the row and column directions of the matrix respectively to obtain an intermediate frequency domain representation matrix in complex form, thereby forming a two-dimensional frequency domain mapping matrix; Based on the two-dimensional frequency domain mapping matrix, the amplitude of each frequency position is calculated respectively and stored in the spectrum matrix according to the corresponding frequency position. The distribution state of the amplitude is reconstructed through the frequency coordinates in the two-dimensional space to obtain the Fourier spectrum of the chrominance channel.

[0050] Specifically, based on the smoothed chromaticity channel map, the specific operation is to first determine the size of the smoothed chromaticity channel map, that is, the height (total number of rows) and width (total number of columns) of the image, and then create a new two-dimensional matrix whose dimensions are exactly the same as the size of the smoothed chromaticity channel map to specifically store the value of the chromaticity b channel. Then, from the chromaticity b channel data of the smoothed chromaticity channel map (usually an independent two-dimensional matrix), starting from the first row and first column of the pixel, scan row by row, and then scan column by column in each row, extract the chromaticity b channel value corresponding to each pixel point in turn. For example, for the pixel with coordinates (row index) in the image , column index ) pixel, read its value from the b channel data of the smoothed chromaticity channel image, and store this value in the same (row index) of the newly created two-dimensional matrix , column index ) position, and this process continues until the chromaticity b channel values ​​of all pixels in the smoothed chromaticity channel image are copied to this new two-dimensional matrix. The new matrix completely retains the chromaticity b channel information of each pixel in the original image spatial structure, thereby forming a chromaticity b channel value matrix.

[0051] According to the chromaticity b channel value matrix, which contains the chromaticity b channel values ​​of all pixels on the display screen in the spatial domain, a two-dimensional discrete Fourier transform is performed on it. This transform maps the image data in the spatial domain to the frequency domain. The two-dimensional Fourier transform can be completed by first performing a one-dimensional Fourier transform on each row of the matrix to obtain an intermediate result matrix, and then performing a one-dimensional Fourier transform on each column of the intermediate result matrix. Alternatively, the integrated two-dimensional fast Fourier transform (FFT) algorithm can be directly used for efficient calculation. For each pixel in the matrix, Chroma b channel value , its two-dimensional Fourier transform result Is a complex number, representing the frequency at a specific space (in Represents the horizontal frequency, Represents the amplitude and phase of the signal component on the vertical direction frequency), each of which is obtained after transformation The value is a complex number in the form ,in is the real part, is the imaginary part, is the imaginary unit, and all these complex values ​​are converted to their corresponding frequency coordinates Organized to form a complex matrix of the same size as the original chrominance b channel numerical matrix. This complex matrix is ​​the intermediate frequency domain representation matrix, also called the two-dimensional frequency domain mapping matrix.

[0052] Based on the two-dimensional frequency domain mapping matrix, each element in the matrix Are all complex numbers, representing the corresponding frequency points The Fourier coefficients of , then we need to calculate the amplitude of each frequency position, for any frequency position The complex Fourier coefficients at , whose magnitude (also called modulus) is calculated as its real part The square and imaginary part of The square root of the sum of the squares is , perform this amplitude calculation operation on each complex element in the two-dimensional frequency domain mapping matrix, and obtain a real-valued matrix of the same size as the original matrix. Each element of the matrix is ​​the amplitude of the corresponding frequency position. Then, create a new two-dimensional matrix called the spectrum matrix, which has the same dimension as the two-dimensional frequency domain mapping matrix, and calculate the amplitude of each frequency position. Store the corresponding frequency coordinates in the spectrum matrix In order to facilitate observation and analysis, the spectrum matrix is ​​usually centered, that is, the zero-frequency component of the spectrum (DC component, corresponding to the average brightness or chrominance of the image) is moved to the center of the spectrum graph, and the amplitude in the spectrum matrix is ​​calculated according to its two-dimensional spatial frequency coordinates (the horizontal axis represents the horizontal frequency). , the vertical axis represents the vertical frequency ) for visualization or storage, the distribution of the amplitude in the frequency space is reconstructed, and the final amplitude distribution diagram is the chromaticity channel Fourier spectrum.

[0053] The steps for obtaining candidate frequency points of periodic artifacts are as follows: Based on the Fourier spectrum of the chrominance channel, traverse each frequency point in the spectrum graph, extract its horizontal frequency index, vertical frequency index and corresponding amplitude information, count the amplitude sets of all frequency points, and calculate the mean and standard deviation of the amplitude set. Filter out frequency points with amplitudes greater than the mean plus one standard deviation as preliminary suspected frequency peaks, and generate a candidate set of frequency peaks; Based on the frequency peak candidate set and the pixel chromaticity deviation value of the relative chromaticity deviation map, the frequency space deviation verification strength is calculated using the following formula: ; in, For the The frequency space deviation verification strength of the frequency peaks, The relative chromaticity deviation spectrum Rank The relative chromaticity deviation value of the column pixel, and Respectively The horizontal frequency index and vertical frequency index of the frequency peak in the Fourier spectrum of the chrominance channel, is the total number of rows in the graph, is the total number of columns in the atlas, is an imaginary unit; Based on the frequency space deviation verification strength, the frequency space deviation verification strength is traversed and compared with the set verification strength reference threshold, and all frequency peaks whose frequency space deviation verification strength is higher than the verification strength reference threshold are identified to obtain the candidate frequency points of periodic artifacts.

[0054] Specifically, based on the Fourier spectrum of the chrominance channel, the spectrum is a one-dimensional matrix, in which each element represents the amplitude at a specific spatial frequency. First, each frequency point in the spectrum is traversed, and for each frequency point, its horizontal frequency index in the spectrum is extracted (for example, denoted as ), longitudinal frequency index (e.g. ) and the amplitude information corresponding to the frequency point (for example, ), collect the amplitudes of all frequency points (usually excluding the DC component, that is, the zero frequency point, because it represents the average chrominance of the image and is insensitive to periodic artifacts) to form an amplitude set. Then, calculate the average value (mean) and standard deviation of all amplitudes in the amplitude set. Then, set a screening condition, that is, the amplitude needs to be greater than the mean plus one standard deviation, and compare each amplitude in the amplitude set with this condition. If the amplitude of a certain frequency point is greater than the mean plus one standard deviation, the amplitude is filtered. If the condition is met, for example, the calculated mean is 0.5 and the standard deviation is 0.2, then the screening condition is that the amplitude is greater than If the amplitude of a frequency point is 0.8, the frequency point is considered to be a preliminary significant frequency peak, and its horizontal frequency index is recorded. and longitudinal frequency index , the index pairs of all frequency points that pass this filter condition are collected to generate a set of frequency peak candidates.

[0055] formula: The benefit of the formula is that it verifies whether the frequency point actually corresponds to the periodic chromaticity deviation in the image space by calculating the Fourier component size of the relative chromaticity deviation spectrum at a specific candidate frequency point. The formula combines the chromaticity deviation information in the spatial domain ( ) and candidate peak information in the frequency domain ( ), effectively linking the frequency characteristics of periodic artifacts with their actual spatial manifestations on the image, improving the accuracy and robustness of periodic artifact detection and avoiding the false peaks introduced by relying solely on the spectrum amplitude; parameter The steps to obtain are: The relative color deviation map generated in the previous step is Rank The relative chromaticity deviation value of the column pixel point, for example, The display will give you a Relative color deviation spectrum, where each The value of reflects the degree of chromaticity deviation of the corresponding pixel, for example, .

[0056] parameter The steps to obtain are: For the The horizontal frequency index of the frequency peak in the Fourier spectrum of the chrominance channel. The index value comes from the "frequency peak candidate set" generated in the previous step. Each element in the set is a coordinate pair of a frequency peak. , for example, select the first frequency peaks, whose lateral frequency index is 10.

[0057] parameter The steps to obtain are: For the The vertical frequency index of the frequency peak in the Fourier spectrum of the chrominance channel is also derived from the first frequency peak in the "frequency peak candidate set". The coordinate pair of the frequency peak, for example, the selected frequency peaks, whose vertical frequency index is 5.

[0058] parameter The steps to obtain are: is the total number of rows of the relative chromaticity deviation map, that is, the height of the display optical image (in pixels). This value is determined when the original RGB pixel data of the display optical image is obtained. For example, if the height of a display is 1080 pixels, then .

[0059] parameter The steps to obtain are: is the total number of columns in the relative chromaticity deviation map, that is, the width of the optical image on the display (in pixels), which is obtained in the same way as For example, if the width of a display is 1920 pixels, .

[0060] Calculation process: have , select a frequency peak candidate, such as frequency peaks, whose transverse frequency index , longitudinal frequency index .

[0061] Relative colorimetric deviation chart for: ; ; ; ; Total number of rows , total number of columns .

[0062] Calculate the phase part of the exponential term : for : ,but ; for : ,but ; for : ,but ; for : ,but ; Calculate the summation : ; ; ; ; Calculating absolute values : ; Calculate frequency space deviation verification strength : ; ; ; This result shows that for the selected frequency peaks , and its frequency space deviation verification strength is 0.125.

[0063] Verify the strength based on the frequency space deviation. This strength is a value calculated for each frequency peak in the "frequency peak candidate set". First, a verification strength reference threshold is set. The basis for setting the threshold value is to perform statistical analysis on a large number of sample images of display screens containing known periodic artifacts and sample images of normal display screens that do not contain such artifacts. Specifically, the frequency space deviation verification strength of all frequency peaks of these two types of samples is calculated respectively, and the distribution difference of the verification strength values ​​of these two types of samples is observed. For example, by drawing the receiver operating characteristic (ROC) curve or analyzing its probability density function, an intensity value that can effectively distinguish the presence or absence of artifacts is selected as the reference threshold. For example, if the analysis shows that the verification strength of the vast majority of periodic artifacts is higher than 0.6, while the verification strength of the frequency peaks that occasionally appear in normal images is generally lower than 0.3, the verification strength reference threshold can be set to 0.5. Then, the frequency space deviation verification strength of all frequency peaks previously calculated is traversed. , each The value is compared with the verification strength reference threshold (for example, 0.5). The value is greater than the verification strength reference threshold, then the frequency peak (i.e. its corresponding horizontal frequency index and longitudinal frequency index ) are marked as candidate frequency points of periodic artifacts and collected together. All frequency peaks that pass this verification process together constitute the final set of candidate frequency points of periodic artifacts.

[0064] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A display screen detection method based on image processing, characterized in that: The following steps are involved: Obtaining original RGB pixel data of the display screen optical image, converting the original RGB pixel data into luminance channel values ​​and chroma a channel values ​​and chroma b channel values ​​in the CIELAB space, integrating the chroma a channel values ​​and chroma b channel values, applying a Gaussian filter for smoothing, and obtaining a smoothed chroma channel image; Based on the chromaticity a channel values ​​of the smoothed chromaticity channel image, a sliding window is set on the display panel pixel information to calculate the standard deviation of the chromaticity a channel values ​​within the window to obtain local chromaticity standard deviation data; based on the local chromaticity standard deviation data, a standard deviation threshold is set to identify pixel areas that exceed the standard deviation threshold, and obtain the location of the high chromaticity variation area; Based on the smoothed chromaticity channel map, selecting an optical uniformity reference block of the display area that is not marked by the chromaticity high variation area location and calculating an average chromaticity value to obtain an average chromaticity parameter of the reference area; and based on the average chromaticity parameter of the reference area, calculating and obtaining a relative chromaticity deviation map; Based on the chromaticity b-channel values ​​of the smoothed chromaticity channel image, a two-dimensional Fourier transform operation is performed on the chromaticity b-channel values ​​of the display light-emitting unit array to obtain a chromaticity channel Fourier spectrum. Based on the chromaticity channel Fourier spectrum, frequency peaks that are different from the reference spectrum are detected to obtain candidate frequency points of periodic artifacts.

2. The display screen detection method based on image processing according to claim 1, characterized in that: The steps for obtaining the smoothed chromaticity channel image are as follows: Obtain the original RGB pixel data of the display screen's optical image, collect the entire screen's pixel matrix, and store the image data into the RGB three-channel matrix according to the red channel value, green channel value, and blue channel value of each pixel point to form the original RGB pixel data; According to the original RGB pixel data, the RGB values ​​are mapped to the CIEXYZ color space, and then the luminance channel value, chroma a channel value, and chroma b channel value of each pixel in the CIELAB space are calculated based on the mapping relationship of the CIEXYZ color space to obtain the CIELAB space channel value; According to the CIELAB spatial channel values, the chromaticity a channel values ​​and the chromaticity b channel values ​​of all pixels are extracted, the chromaticity a channel values ​​and the chromaticity b channel values ​​are aligned and combined into a two-dimensional vector matrix according to the pixel positions, and the two-dimensional vector matrix is ​​convolved and smoothed using a two-dimensional Gaussian kernel to form a smoothed chromaticity channel map.

3. The display screen detection method based on image processing according to claim 1, characterized in that: The steps for obtaining the local chromaticity standard deviation data are: Based on the chromaticity a channel value of the smoothed chromaticity channel image, with each pixel in the display panel pixel information as the center point, a square sliding window of a fixed size is set, so that the window slides row by row and column by column to cover each pixel position on the smoothed chromaticity channel image, forming a sliding window positioning matrix; According to the sliding window positioning matrix, traverse and extract the chroma a channel values ​​corresponding to all pixels covered by the window in each sliding window, calculate the average value of all the extracted chroma a channel values ​​within each sliding window, and record the average value to form a chroma a channel average value matrix within the sliding window; Based on the chroma a channel average value matrix in the sliding window, taking the average value in each sliding window as a benchmark, traverse all the extracted chroma a channel values ​​in each sliding window, calculate the difference between each chroma a channel value and the average value in the corresponding sliding window, calculate the average of the squares of the differences and take the square root to obtain the local chroma standard deviation data.

4. The display screen detection method based on image processing according to claim 1, characterized in that: The steps for obtaining the location of the high chromaticity variation region are as follows: Based on the local chromaticity standard deviation data, the numerical distribution of the standard deviation of the chromaticity a channel values ​​within all sliding windows is counted, the numerical distribution median in the local chromaticity standard deviation data is extracted, and the numerical distribution median is used as a basic reference to set a standard deviation threshold for identifying high variation areas to form a standard deviation threshold; According to the standard deviation threshold, each value in the local chromaticity standard deviation data is judged one by one. If the standard deviation of the chromaticity a channel value is greater than or equal to the standard deviation threshold, the corresponding sliding window coverage area is marked as a candidate high-variance pixel area, otherwise it is marked as a normal area, forming a candidate high-variance area marking map; Based on the candidate high-variability region label map, adjacent and continuous candidate high-variability pixel regions are extracted and merged, region boundaries are determined, and region position information is established to obtain the chromaticity high-variability region positioning.

5. The display screen detection method based on image processing according to claim 1, characterized in that: The steps for obtaining the average chromaticity parameter of the reference area are: Based on the chromaticity high variation region location, identifying continuous display regions not marked by the chromaticity high variation region location in the smoothed chromaticity channel image, excluding regions that overlap or are adjacent to the chromaticity high variation region location, to form candidate display regions; According to the candidate display areas, statistics are collected and compared on the changes in chromaticity values ​​within each candidate display area, the standard deviation of the chromaticity values ​​of the pixels within each area is calculated, and the candidate display area with the lowest standard deviation and the smallest change in internal chromaticity values ​​is determined to form an optical uniformity reference block; Based on the optical uniformity reference block, the chromaticity a channel value and the chromaticity b channel value of each pixel point in the reference block are traversed and extracted, all the extracted chromaticity a channel values ​​and chromaticity b channel values ​​are averaged respectively, and the average chromaticity a channel value and the average chromaticity b channel value are calculated and recorded to obtain the average chromaticity parameter of the reference area.

6. The display screen detection method based on image processing according to claim 1, characterized in that: The steps for obtaining the relative chromaticity deviation spectrum are: Based on the average chromaticity parameter of the reference area, the chromaticity a channel value and the chromaticity b channel value of each pixel point are read pixel by pixel from the smoothed chromaticity channel image, and the column coordinate value and the row coordinate value of the corresponding pixel point in the image are extracted to obtain a pixel point attribute set including the chromaticity channel and the pixel position; Calculating the relative chromaticity deviation between each pixel and the average chromaticity parameter of the reference area according to the pixel attribute set; Based on the relative chromaticity deviation value of each pixel, the relative chromaticity deviation value is filled into the corresponding position of the two-dimensional matrix of the same dimension as the smoothed chromaticity channel map according to the row and column coordinate order of each pixel in the image to generate a relative chromaticity deviation map.

7. The display screen detection method based on image processing according to claim 1, characterized in that: The steps for obtaining the Fourier spectrum of the chrominance channel are as follows: Based on the smoothed chromaticity channel map, extracting chromaticity b channel values ​​corresponding to all pixel points of the display light-emitting unit array row by row and column by column, and storing the chromaticity b channel values ​​in a matrix according to their original row and column positions to form a chromaticity b channel value matrix; According to the chroma b channel value matrix, performing a two-dimensional Fourier transform on the matrix, performing frequency domain mapping on the chroma b channel values ​​of all pixels along the row and column directions of the matrix, obtaining an intermediate frequency domain representation matrix in complex form, and forming a two-dimensional frequency domain mapping matrix; Based on the two-dimensional frequency domain mapping matrix, the amplitude of each frequency position is calculated respectively and stored in the spectrum matrix according to the corresponding frequency position. The distribution state of the amplitude is reconstructed through the frequency coordinates in the two-dimensional space to obtain the chrominance channel Fourier spectrum.

8. The display screen detection method based on image processing according to claim 1, characterized in that: The steps for obtaining the candidate frequency points of periodic artifacts are as follows: Based on the Fourier spectrum of the chrominance channel, traverse each frequency point in the spectrum graph, extract its horizontal frequency index, vertical frequency index and corresponding amplitude information, count the amplitude sets of all frequency points, and calculate the mean and standard deviation of the amplitude sets, screen out frequency points with amplitudes greater than the mean plus one standard deviation as preliminary suspected frequency peaks, and generate a candidate set of frequency peaks; Calculating the frequency space deviation verification strength based on the frequency peak candidate set and the pixel chromaticity deviation value of the relative chromaticity deviation map; Based on the frequency-space deviation verification strength, the frequency-space deviation verification strength is traversed and compared with the set verification strength reference threshold, and all frequency peaks whose frequency-space deviation verification strength is higher than the verification strength reference threshold are identified to obtain candidate periodic artifact frequency points.

9. The display screen detection system according to any one of claims 1 to 8, characterized in that: include: An image preprocessing module obtains the original RGB pixel data of the display screen optical image, converts the original RGB pixel data into the brightness channel value and the chroma a channel value and chroma b channel value in the CIELAB space, integrates the chroma a channel value and the chroma b channel value, applies a Gaussian filter to smooth, and obtains a smoothed chroma channel image; A chromaticity variation detection module sets a sliding window on the display panel pixel information, based on the chromaticity a channel values ​​of the smoothed chromaticity channel image, to calculate the standard deviation of the chromaticity a channel values ​​within the window, thereby obtaining local chromaticity standard deviation data. Based on the local chromaticity standard deviation data, a standard deviation threshold is set to identify pixel areas exceeding the standard deviation threshold, thereby obtaining the location of areas with high chromaticity variation; a chromaticity deviation calculation module, based on the smoothed chromaticity channel map, selecting an optical uniformity reference block of the display area that is not marked by the chromaticity high variation area location, calculating an average chromaticity value, obtaining an average chromaticity parameter of the reference area, and calculating a relative chromaticity deviation map based on the average chromaticity parameter of the reference area; The periodic artifact detection module performs a two-dimensional Fourier transform operation on the chromaticity b channel values ​​of the display light-emitting unit array based on the chromaticity b channel values ​​of the smoothed chromaticity channel image to obtain a chromaticity channel Fourier spectrum. Based on the chromaticity channel Fourier spectrum, the module detects frequency peaks that are different from the reference spectrum to obtain candidate frequency points of periodic artifacts.

Images