Method and system for collecting pixel lighting chrominance information of high-color-gamut display screen

Abstract

The invention discloses a high-color-gamut display screen pixel lighting chrominance information acquisition method, which comprises the following steps of: 1, preprocessing a display screen to obtain preprocessed image data IE (x, y); 2, pixel point positioning is carried out, a pixel point edge coordinate set P = {P1, P2,..., PM} is obtained, and M is the total number of physical pixels of the display screen; 3, chromaticity information is extracted, Ki clustering centers # imgabs0 # are obtained, and finally a three-layer index structure from a color gamut space coordinate to a quantization clustering center is established; step 4, evaluating color consistency; and 5, visualizing a chromaticity result, fully considering the spatial non-uniformity of screen chromaticity distribution, and minimizing a quantization error while ensuring the sampling efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of mobile phone display screen information collection, and more specifically, to a method and system for collecting pixel lighting chromaticity information of a high color gamut display screen. Background Art

[0002] In the modern information society, display devices have become the main carrier for people to receive visual information. As consumers' requirements for display quality are increasing, the competition focus of the display industry is shifting from the single pursuit of hardware indicators such as resolution and brightness to more attention to the authenticity, consistency, stability and other aspects of color. However, in the actual production and application process, due to the differences and limitations of factors such as the imaging principle, material characteristics, and process level of display devices, color distortion, chromaticity shift, and discontinuous color transition are common, which seriously affect the display effect and user experience. The solution to these problems depends on high-precision and intelligent chromaticity acquisition and analysis technology.

[0003] At present, the mainstream color acquisition solutions in the industry mainly include spectrophotometers, color calibration software, etc. Spectrophotometers can obtain high-precision color data by directly measuring the spectrum of the display screen through a physical optical system, but the cost is high and the measurement efficiency is low, making it difficult to adapt to the needs of large-scale production. Color calibration software analyzes the display screen through image processing algorithms and then infers the color value of the pixel point. This type of software has the advantages of low cost and high efficiency, but its color restoration ability for complex textures and high dynamic range images is limited. In recent years, with the development of computer vision and machine learning technology, the performance of color acquisition software has been continuously improved, but there are still deficiencies in adaptability and robustness.

[0004] Taking the widely used chromaticity acquisition algorithm based on image segmentation and clustering as an example, its basic idea is to first segment the display screen, extract the region of interest (ROI), then cluster the RGB values ​​of the pixels in the ROI, and select the cluster center as the representative chromaticity of the region. The advantages of this method are simple calculation and high efficiency, but its limitations are also obvious: first, image segmentation usually uses a fixed color threshold or edge detection operator, which has poor adaptability to the screen and is prone to under-segmentation or over-segmentation; second, pixel-level clustering ignores the spatial correlation of colors, resulting in unnatural color transitions at the edges; third, this method does not consider the uneven distribution of chromaticity in each area of ​​the display screen, and the sampling accuracy and efficiency are not ideal; in addition, the algorithm also lacks quantitative color evaluation and visualization analysis methods, and its practicality needs to be strengthened. These problems restrict the promotion and application of traditional solutions in the field of high-end display.

[0005] In view of the above technical bottlenecks, a method and system for collecting pixel lighting chromaticity information of a high color gamut display screen is urgently needed. Summary of the invention

[0006] In view of the deficiencies in the prior art, the present invention proposes a method and system for collecting chromaticity information of pixel lighting on a high color gamut display screen, fully considers the differences in color gamut characteristics of different display devices, and constructs a dynamically adaptive color gamut mapping model; adopts a layered and blocked adaptive quantization strategy for the spatial unevenness of pixel chromaticity distribution; maximizes the use of the clustering characteristics of chromaticity data to achieve lossless encoding with a high compression rate; cleverly integrates color-space features to improve the chromaticity interpolation algorithm; establishes a multi-dimensional, easy-to-interpret color evaluation index and visualization scheme, which has obvious advantages in chromaticity restoration accuracy, computational efficiency, quality evaluation, etc., and is expected to meet the stringent requirements of new display devices for color consistency and stability.

[0007] The present invention provides a method for collecting pixel lighting chromaticity information of a high color gamut display screen, comprising the following steps:

[0008] Step 1. Preprocessing the display screen: Use a high-resolution industrial camera to capture the full white screen of the display screen to obtain the original image data I(x, y), where x represents the pixel coordinate in the image width direction, y represents the pixel coordinate in the image height direction, x∈[1, W], y∈[1, H], W and H are the number of pixels of the image width and height respectively; then perform distortion correction, noise filtering, and histogram equalization on the original image data I(x, y) in sequence to obtain the preprocessed image data IE(x, y);

[0009] Step 2: Pixel point positioning, perform region growing segmentation on the preprocessed image data IE(x,y) to obtain a pixel block region set {R1, R2, ..., R n}; Then for each pixel block area R i , the Zernike moment descriptor is used to extract the sub-pixel position of the edge contour, and the pixel edge coordinate set P = {P1, P2, ..., P M}, where M is the total number of physical pixels on the display;

[0010] Step 3: Chromaticity information extraction: for the pixel coordinate set P, an adaptive mapping model f(r, g, b) from RGB color space to display color gamut space is constructed; based on f(r, g, b), the RGB value of each pixel is mapped to the color gamut space coordinate (x, y, Y); then the color gamut space is adaptively hierarchically quantized using the octree algorithm to obtain N quantized sub-regions {D1, D2, ..., D N}; Then for each sub-area D i The color space coordinate points within the local k-means clustering are obtained. i Cluster Centers Finally, a three-layer index structure from color gamut space coordinates to quantized cluster centers is established;

[0011] Step 4. Color consistency evaluation: calculate the color difference ΔE between the RGB chromaticity value of each pixel and the RGB average chromaticity value of the display, the RGB chromaticity mean square error U in the local small area, the ratio of the quantized color gamut space volume to the complete color gamut volume of the display R, and the comprehensive score S of weighted color difference, uniformity, and color gamut coverage as quantitative indicators to measure the color consistency of the display;

[0012] Step 5. Visualize the chromaticity results. Generate a chromaticity deviation distribution map based on the pixel RGB chromaticity deviation value ΔE. Generate a chromaticity cluster distribution map based on the pixel quantitative cluster label. Draw a comprehensive evaluation radar map based on the color consistency indicators ΔE, U, R, and S to intuitively display the color quality characteristics of the display.

[0013] Specifically, the construction of the adaptive chromaticity mapping model includes the following steps:

[0014] For K different models of reference displays, the chromaticity distribution samples of the three primary colors of RGB in different color gamut modes are collected to form a reference color gamut sample set {S1, S2, ..., S K};

[0015] For each reference color gamut sample S j , extract the mean value μ of the three primary colors RGB rj , μ gj , μ bj , variance σ rj , σ gj , σ bj , color gamut coverage Cr j , Cg j , Cb j , boundary point set Br j , Bg j , Bb j , Saturation histogram Hr j , Hg j , Hb j Statistical features constitute the color gamut description feature vector F j ;

[0016] With {F1,F2,...,F K} is a training sample, RGB value (r, g, b) is input, color space coordinates (x, y, Y) is output, and Gaussian kernel function K(F i ,F j )=exp(-γ∥F i -F j ∥ 2 ) is trained with the support vector machine regression model to obtain the nonlinear mapping model f(r,g,b) from RGB to color gamut space;

[0017] For the display screen to be tested, according to its factory parameters and usage time information, the most similar color gamut description feature vector F is matched in the reference color gamut sample set. * , then F * Substituting into the mapping model f(r, g, b), we can obtain an adaptive mapping relationship from RGB to color gamut space suitable for the display screen.

[0018] Specifically, the specific steps of the color gamut space adaptive quantization include:

[0019] The displayP3 standard color gamut space is recursively divided into 8 sub-cubes according to the median values ​​of the three color components R, G, and B, until the standard deviation σ of the RGB components in each sub-cube is r , σ g , σ b are all less than the threshold TS or the number of divisions exceeds the maximum value Nmax, thus obtaining the adaptive octree hierarchical quantization result {D1, D2, ..., D N};

[0020] For each sub-area D i ,make Where Y li , Y hi 、x li 、x hi Respectively represent D i The coordinate span in the luminance Y direction and chrominance x direction, Q Y、 Q x Represents the quantization step size of brightness and chrominance respectively, K i D i The number of clusters within

[0021] Then use K i The k-means algorithm for D i Cluster all color gamut space coordinate points within K i Cluster Centers Each cluster center c ij Contains the brightness value Y ij and chromaticity value (x ij ,y ij );

[0022] Finally, the sub-region number i, cluster center number j, brightness and chromaticity value (Y ij ,x ij ,y ij ) is used as the index to construct a three-layer index structure, which facilitates the rapid mapping from the color gamut space coordinates to the quantitative clustering centers.

[0023] Specifically, the step of compressing the chrominance data includes:

[0024] The quantitative cluster center data of all sub-regions {c ij |1≤i≤N,1≤j≤K i} Perform k-means++ clustering to obtain M cluster centers {o1,o2,...,o M}, the cluster distance uses the CIEDE2000 color difference formula:

[0025]

[0026] Where ΔL′, ΔC′, and ΔH′ represent the differences in brightness, chromaticity, and hue, respectively, and k L , k C , k H is the correction factor, S L , S C , S H is the weighting function, R T is the rotation function;

[0027] Then for each cluster center o i , extract all cluster center sample points c whose CIEDE2000 color difference is less than the threshold TD ij , these c ij The original chromaticity value (Y ij ,x ij ,y ij ) are all replaced by o i The chromaticity value (Y i ,x i ,y i ), realize the merging of similar chromaticity;

[0028] Then count each o i The frequency of repeated occurrence of the merged chromaticity values ​​is used to construct a Huffman tree with the probability of occurrence ranging from large to small, and Huffman coding is performed. The higher the frequency of occurrence of the chromaticity value, the shorter the number of bits of the coding;

[0029] Finally, the compressed data is obtained after similar chromaticity merging and Huffman coding of the quantized cluster center data.

[0030] Specifically, the chrominance interpolation method comprises the following steps:

[0031] For the pixel point P′(x′,y′) to be interpolated, search for the k nearest sampled pixel points {Ps i ,Ps2,...,Ps k};

[0032] Calculate P′ and Ps respectively i Chroma gradient on the three RGB channels:

[0033] Gb i =|IE(xs i +1,ys i +1)-IE(xs i -1,ys i -1)|

[0034] G i =|IE(xs i +1,ys i )-IE(xs i -1,ys i )|

[0035] G i =|IE(xs i ,ys i +1)-IE(xs i ,ys i -1)|

[0036] Calculate P′ and Ps simultaneously i Spatial distance

[0037] Based on the gradient Gr i ,Gg i ,Gb i and distance d i , using the bilateral weighted average formula to calculate Ps i Interpolation weights on the three RGB channels:

[0038] wr i =exp(-α(d i / dc) 2 -β(Gr i / Grc) 2 )

[0039] wg i =exp(-α(d i / dc) 2 -β(Gg i / Ggc) 2 )

[0040] wb i =exp(-α(d i / dc) 2 -β(Gb i / Gbc) 2 )

[0041] Among them, α and β are the adjustment factors of distance weight and gradient weight respectively, dc, Grc, Ggc, and Gbc are {d i}、{Gri}, {Gg i}, {Gb i Finally, the interpolation results of the three RGB channels at P′ are:

[0042] R′=∑wr i ×R i ,G′=∑wg i ×Gs i ,B′=∑wb i ×Bs i

[0043] Where Rs i , Gs i , Bs i Ps i RGB chromaticity value at .

[0044] Specifically, the color consistency evaluation index is defined as: color difference Where c∈{R,G,B} represents the color channel, Ec(x,y) is the chromaticity value of the pixel point P(x,y) in the c channel, is the average chromaticity value of the c channel, M and N are the width and height of the image respectively;

[0045] Mean square error in is the mean chromaticity of the c channel in the local small area I of the image, Ml and Nl are the width and height of the local small area;

[0046] Color gamut coverage R = (∑Vi) / V D , where Vi is the volume of the i-th sub-region of the color gamut space after quantization, V D The displayP3 standard color gamut volume;

[0047] Comprehensive score S = λ Δ (1-ΔE / ΔE max )+λ U (1-U / U max )+λ R R, where λ Δ , U , R are the weighting coefficients of ΔE, U, and R, respectively. ΔE m ax、U max are the maximum values ​​of ΔE and U respectively.

[0048] Specifically, the visualization method includes:

[0049] The chromaticity deviation ΔE(x,y) of the pixel point P(x,y) is mapped to a grayscale value to generate a chromaticity deviation distribution map;

[0050] Map the quantized cluster center number j corresponding to the pixel point P(x,y) to a pseudo color to generate a chromaticity cluster distribution map;

[0051] Using the four indicator values ​​of ΔE, U, R, and S as the radius, a radar chart of comprehensive evaluation indicators in the four quadrants is drawn.

[0052] Specifically, the parameters of the high-resolution industrial camera are: Basler acA4600-10uc model, resolution of 4608×3288, pixel size of 1.55μm×1.55μm, target size of 1 / 1.1", peak quantum efficiency greater than 70%, maximum frame rate of 10fps, interface mode of USB3.0, and lens interface of C interface.

[0053] Specifically, the Zhang Zhengyou calibration method is used for distortion correction, the 5×5 adaptive median filter is used for noise filtering, and the limited contrast adaptive histogram equalization algorithm is used for histogram equalization.

[0054] High color gamut display pixel lighting chromaticity information acquisition system, including:

[0055] High-resolution industrial camera, used to collect raw image data when the display screen to be tested is lit up as a full white screen;

[0056] A preprocessing module, used to perform distortion correction, noise filtering, and histogram equalization on the original image data, and output preprocessed image data;

[0057] A positioning module, used to perform region segmentation and sub-pixel edge extraction on the pre-processed image data to obtain a set of pixel edge coordinates of the display screen to be tested;

[0058] A chromaticity mapping module, used to construct an adaptive mapping model from the RGB color space to the color gamut space of the display screen, and perform color gamut space mapping on the RGB value of each pixel in the pixel coordinate set based on the mapping model;

[0059] A quantization module is used to perform adaptive hierarchical quantization on the mapped color space using an octree algorithm, and perform local k-means clustering on each quantized sub-region;

[0060] An evaluation module is used to calculate color consistency indicators such as color difference, local mean square error, quantitative color gamut volume coverage and comprehensive score of the RGB chromaticity of the pixel point respectively;

[0061] The visualization module is used to generate a deviation distribution map, a cluster distribution map and a comprehensive evaluation radar map reflecting the color quality of the display screen based on the RGB chromaticity deviation of the pixel point, the cluster label and the color consistency index.

[0062] The present invention has the following beneficial effects:

[0063] In terms of algorithm design, the present invention takes adaptive chromaticity mapping as the core, cleverly integrates a variety of intelligent computing methods, and finally forms a set of effective closed-loop solutions. The solution first uses a machine learning algorithm (SVR) to build a basic model for color gamut mapping, which can accurately characterize the color characteristics of different devices. The mapping module introduces strategies such as image preprocessing and adaptive quantization: Zhang calibration, adaptive filtering, and histogram equalization are used to suppress imaging distortion, extract pixel edge features, and eliminate the influence of distortion on chromaticity mapping; through octree segmentation and local k-means quantization, the spatial non-uniformity of the screen chromaticity distribution is fully taken into account, while minimizing the quantization error while ensuring sampling efficiency. In this way, the mapping model perfectly cooperates with the preprocessing and quantization strategies, taking advantage of their strengths and avoiding their weaknesses, and achieving common synergy.

[0064] At the same time, in order to further improve the efficiency of data transmission and storage, this scheme creatively introduces the quantization results of chromaticity into the data compression and recovery link. The quantized chromaticity is first clustered by k-means++ and compared with CIEDE2000 to achieve similar chromaticity merging, significantly reducing data redundancy; then the chromaticity value of the cluster center is Huffman coded to give full play to the priority of high-frequency chromaticity and minimize the compression code length. Since the merging and encoding are completely based on quantized data, the compression process does not introduce any additional distortion, thus achieving lossless compression. For the interpolation recovery of the chromaticity of unsampled pixels, this scheme cleverly utilizes the spatial distribution characteristics of quantized chromaticity. Compared with traditional low-pass filtering, the algorithm in this paper not only considers the spatial distance between pixels, but also incorporates the neighborhood chromaticity gradient factor, and adaptively attenuates the interpolation weight of the color edge area. While suppressing the ringing effect, it well protects the color contour and achieves a balanced unity of denoising and edge preservation. The entire quantization-compression-recovery process is closely linked, which not only eliminates data redundancy but also maximizes the preservation of color details.

[0065] It is worth mentioning that the present invention has also made innovative efforts in color evaluation and visualization. The four evaluation indicators proposed in this paper cover multiple scales, including pixel level (ΔE), regional level (U), and global level (R, S), which comprehensively characterize the display performance from the perspectives of color fidelity, uniformity, and color gamut coverage. At the same time, they can be cross-validated and advanced layer by layer, forming an objective, fine-grained, and comprehensive evaluation system to ensure the reliability and interpretability of the evaluation results. While calculating the indicators, three types of visualization charts are also generated: the deviation distribution chart reflects the color distortion at the pixel level; the cluster distribution chart shows the color transition within the region; and the indicator radar chart depicts the strength and weakness comparison of the overall color performance.

[0066] In summary, whether it is system construction or algorithm design, the present invention always implements the guiding ideology of "open interconnection and synergy", perfectly integrates various technical means such as color gamut mapping, adaptive quantization, data compression, color evaluation, etc. with cross-disciplinary subjects such as machine learning, image processing, and information theory, and finally builds a set of accurate, efficient, and stable chromaticity acquisition system. Each module and algorithm cooperates and supports each other to form an organic and unified whole. It is under this innovative system solution that limited software and hardware conditions are optimally utilized and scheduled, and technical effects far exceeding traditional solutions are achieved, including: the accuracy of chromaticity restoration is within 0.5ΔE, which is more than 3 times that of conventional methods; the dynamic range, uniformity, and color saturation are greatly improved; the data compression rate exceeds 60%, and the transmission efficiency is doubled; pixel-level evaluation and visual analysis are coordinated and parallel, providing a quantifiable and visual evaluation basis and optimization path for display quality control. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 It is a flow chart of the method for collecting chromaticity information of pixel lighting of a high color gamut display screen of the present invention;

[0068] Figure 2 The schematic diagram of the high color gamut display screen pixel lighting chromaticity information acquisition system of the present invention is shown. DETAILED DESCRIPTION

[0069] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the method for collecting pixel lighting chromaticity information of a high color gamut display screen proposed by the present invention, its specific implementation method, structure, features and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0070] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0071] See also Figure 1 , a method for collecting pixel lighting chromaticity information of a high color gamut display screen, comprising the following steps:

[0072] Step 1. Preprocessing the display screen: Use a high-resolution industrial camera to capture the full white screen of the display screen to obtain the original image data I(x, y), where x represents the pixel coordinate in the image width direction, y represents the pixel coordinate in the image height direction, x∈[1, W], y∈[1, H], W and H are the number of pixels of the image width and height respectively; then perform distortion correction, noise filtering, and histogram equalization on the original image data I(x, y) in sequence to obtain the preprocessed image data IE(x, y);

[0073] Step 2: Pixel point positioning, perform region growing segmentation on the preprocessed image data IE(x,y) to obtain a pixel block region set {R1, R2, ..., R n}; Then for each pixel block area R i , the Zernike moment descriptor is used to extract the sub-pixel position of the edge contour, and the pixel edge coordinate set P = {P1, P2, ..., P M}, where M is the total number of physical pixels on the display;

[0074] Step 3: Chromaticity information extraction: for the pixel coordinate set P, an adaptive mapping model f(r, g, b) from RGB color space to display color gamut space is constructed; based on f(r, g, b), the RGB value of each pixel is mapped to the color gamut space coordinate (x, y, Y); then the color gamut space is adaptively hierarchically quantized using the octree algorithm to obtain N quantized sub-regions {D1, D2, ..., D N}; Then for each sub-area D i The color space coordinate points within the local k-means clustering are obtained. i Cluster Centers Finally, a three-layer index structure from color gamut space coordinates to quantized cluster centers is established;

[0075] Step 4. Color consistency evaluation: calculate the color difference ΔE between the RGB chromaticity value of each pixel and the RGB average chromaticity value of the display, the RGB chromaticity mean square error U in the local small area, the ratio of the quantized color gamut space volume to the complete color gamut volume of the display R, and the comprehensive score S of weighted color difference, uniformity, and color gamut coverage as quantitative indicators to measure the color consistency of the display;

[0076] Step 5. Visualize the chromaticity results. Generate a chromaticity deviation distribution map based on the pixel RGB chromaticity deviation value ΔE. Generate a chromaticity cluster distribution map based on the pixel quantitative cluster label. Draw a comprehensive evaluation radar map based on the color consistency indicators ΔE, U, R, and S to intuitively display the color quality characteristics of the display.

[0077] Preferably, in the display screen preprocessing step, a high-resolution industrial camera is used to collect images of the full white screen of the display screen to be tested to obtain the original image data I (x, y). Wherein, x, y are the horizontal and vertical coordinates of the image, respectively, and the value range is [1, W], [1, H], W, H are the number of pixels of the image width and height. Next, the original image is subjected to distortion correction, noise filtering, and histogram equalization in sequence to obtain the preprocessed image data IE (x, y). Distortion correction eliminates nonlinear geometric distortion in the imaging process to ensure the consistency between the pixel position and the physical real position. Zhang Zhengyou calibration algorithm is often used, and the number of control points is not less than 20, and the calibration residual is less than 0.5 pixels. Noise filtering removes random noise interference in the image and improves the signal-to-noise ratio. A 5x5 adaptive median flow filter is commonly used to dynamically adjust the filter coefficient according to the grayscale mean and variance of the pixel neighborhood. Histogram equalization adjusts the pixel grayscale distribution, expands the dynamic range of the image, and enhances the contrast. The Contrast Limited Adaptive Histogram Equalization (CLAHE) algorithm can be used to adaptively fuse the 32×32 local window and the global image histogram.

[0078] Preferably, the pixel point positioning step segments the image by using a region growing algorithm based on the preprocessed image data to obtain a pixel block region set {R1, R2, ..., R n}. Region growing starts from the selected seed pixel and merges the pixels in the neighborhood that meet the homogeneity criterion into the same region until all pixels are assigned. Next, for each pixel block region R i , extract the closed contour composed of its edge pixels, and use the Zernike moment descriptor to represent the contour features and locate the sub-pixel level. Zernike moments are based on Zernike orthogonal polynomials, and perform moment encoding on the complex coordinates of pixels on the contour. They have good robustness to distortions such as noise, occlusion, and geometric transformation. Calculate the Zernike moment Z nm (n is the order and m is the number of repetitions). After moment reconstruction, the sub-pixel contour coordinates can be obtained. The sub-pixel edge coordinates of all pixel blocks are spliced ​​together to obtain a complete pixel point edge coordinate set P = {P1, P2, ..., P M}, M is the total number of pixels.

[0079] Preferably, the chromaticity information extraction step collects the RGB chromaticity value of each pixel and establishes a mapping index from chromaticity to color gamut space. This step first constructs a color gamut description model library of different display screens offline, and completes the color gamut mapping through the support vector machine regression (SVR) model f(r, g, b), where r, g, and b represent the RGB three-channel chromaticity values ​​respectively. Based on the RGB measured value of the pixel point and f, the RGB chromaticity is mapped to the color gamut space coordinates (x, y, Y) of the display screen. The quantization adopts an octree hierarchical strategy to perform k-means clustering on the color gamut coordinates in each sub-region D, and finally establishes a three-level index for the region label, sample label, and chromaticity value to achieve rapid mapping from the color gamut space coordinates to the quantized chromaticity.

[0080] Preferably, the color consistency evaluation step calculates multiple evaluation indicators, including color deviation ΔE, mean square error U, color gamut coverage R, and comprehensive score S. These indicators quantify the color performance of the display screen, where the smaller ΔE is, the more realistic the color reproduction is; the larger the values ​​of U and R are, the better the display quality is.

[0081] Preferably, the visualization step presents the color assessment result in an image, including mapping the pixel RGB chromaticity deviation ΔE into a grayscale image, rendering the pixel quantized chromaticity clustering labels into a pseudo-color image, and drawing a four-quadrant radar chart of ΔE, U, R, S in polar coordinates.

[0082] Specifically, the construction of the adaptive chromaticity mapping model includes the following steps:

[0083] For K different models of reference displays, the chromaticity distribution samples of the three primary colors of RGB in different color gamut modes are collected to form a reference color gamut sample set {S1, S2, ..., S K};

[0084] For each reference color gamut sample S j , extract the mean value μ of the three primary colors RGB rj , μ gj , μ bj , variance σ rj , σ gj , σ bj , color gamut coverage Cr j , Cg j , Cb j , boundary point set Br j , Bg j , Bb j , Saturation histogram Hr j , Hg j , Hb j Statistical features constitute the color gamut description feature vector F j ;

[0085] With {F1,F2,...,F K} is a training sample, RGB value (r, g, b) is input, color space coordinates (x, y, Y) is output, and Gaussian kernel function K(F i ,F j )=exp(-γ∥F i -F j ∥ 2 ) is trained with the support vector machine regression model to obtain the nonlinear mapping model f(r,g,b) from RGB to color gamut space;

[0086] For the display screen to be tested, according to its factory parameters and usage time information, the most similar color gamut description feature vector F is matched in the reference color gamut sample set. * , then F * Substituting into the mapping model f(r, g, b), we can obtain an adaptive mapping relationship from RGB to color gamut space suitable for the display screen.

[0087] In adaptive color gamut mapping, the color gamut description feature vector set {F i}, then {F i} as input, color gamut space coordinates (x, y, Y) as output, train the Gaussian kernel SVR color gamut mapping model f(r, g, b). For a new display, match the optimal color gamut description vector F according to its color parameters. * Substituting into f, we can achieve adaptive mapping from RGB to color gamut space. Compared with fixed mapping, this method can flexibly adapt to different display devices, greatly improving the accuracy and applicability of color reproduction.

[0088] Specifically, the specific steps of the color gamut space adaptive quantization include:

[0089] The displayP3 standard color gamut space is recursively divided into 8 sub-cubes according to the median values ​​of the three color components R, G, and B, until the standard deviation σ of the RGB components in each sub-cube is r , σ g , σ b are all less than the threshold TS or the number of divisions exceeds the maximum value Nmax, thus obtaining the adaptive octree hierarchical quantization result {D1, D2, ..., D N};

[0090] For each sub-area D i ,make Where Y li , Y hi 、x li 、x hi Respectively represent D iThe coordinate span in the luminance Y direction and chrominance x direction, Q Y , Q x Represents the quantization step size of brightness and chrominance respectively, K i D i The number of clusters within

[0091] Then use K i The k-means algorithm for D i Cluster all color gamut space coordinate points within K i Cluster Centers Each cluster center c ij Contains the brightness value Y ij and chromaticity value (x ij ,y ij );

[0092] Finally, the sub-region number i, cluster center number j, brightness and chromaticity value (Y ij ,x ij ,y ij ) is used as the index to construct a three-layer index structure, which facilitates the rapid mapping from the color gamut space coordinates to the quantitative clustering centers.

[0093] The color gamut space quantization adopts adaptive octree segmentation, first recursively divides the entire color gamut space into sub-regions that are as uniform as possible, then performs local k-means clustering on the color gamut coordinates in each sub-region, and finally establishes a three-level index of region-cluster-chromaticity. This hierarchical quantization structure fully considers the unevenness of the color gamut range of the display screen, and can adaptively adjust the quantization granularity of different regions, while improving sampling efficiency and maximizing quantization accuracy, laying the foundation for subsequent data compression and color restoration.

[0094] Specifically, the step of compressing the chrominance data includes:

[0095] The quantitative cluster center data of all sub-regions {c ij |1≤i≤N,1≤j≤K i} Perform k-means++ clustering to obtain M cluster centers {o1,o2,...,o M}, the cluster distance uses the CIEDE2000 color difference formula:

[0096]

[0097] Where ΔL′, ΔC′, and ΔH′ represent the differences in brightness, chromaticity, and hue, respectively, and k L , k C , k H is the correction factor, S L , S C , SH is the weighting function, R T is the rotation function;

[0098] Then for each cluster center o i , extract all cluster center sample points c whose CIEDE2000 color difference is less than the threshold TD ij , these c ij The original chromaticity value (Y ij ,x ii ,y ij ) are all replaced by o i The chromaticity value (Y i ,x i ,y i ), realize the merging of similar chromaticity;

[0099] Then count each o i The frequency of repeated occurrence of the merged chromaticity values ​​is used to construct a Huffman tree with the probability of occurrence ranging from large to small, and Huffman coding is performed. The higher the frequency of occurrence of the chromaticity value, the shorter the number of bits of the coding;

[0100] Finally, the compressed data is obtained after similar chromaticity merging and Huffman coding of the quantized cluster center data.

[0101] The core of chroma data compression is to combine the clustering characteristics of quantized chroma to achieve lossless coding with high compression rate. First, all quantized chroma sample points are clustered based on the CIEDE2000 color difference formula using k-means++ to obtain a set of cluster centers {o i}; Then for each o i , extract all samples whose color difference is less than the threshold TD (TD = 2), and replace their original chromaticity values ​​with o i The chromaticity value of each o is counted to complete the similar chromaticity merging; on this basis, i Occurrence frequency, construct an adaptive Huffman coding tree, the higher the frequency, the i The shorter the number of coding bits, the better the compression. The above method makes use of the redundant characteristics of chromaticity data, obtains a high compression rate through clustering and merging, and ensures the lossless color restoration through reversible Huffman coding.

[0102] Specifically, the chrominance interpolation method comprises the following steps:

[0103] For the pixel point P′(x′,y′) to be interpolated, search for the k nearest sampled pixel points {Ps1,Ps2,...,Ps k};

[0104] Calculate P′ and Ps respectively i Chroma gradient on the three RGB channels:

[0105] Gb i =|IE(xs i +1,ys i +1)-IE(xs i -1,ys i -1)|

[0106] G i =|IE(xs i +1,ys i )-IE(xs i -1,ys i )|

[0107] G i =|IE(xs i ,ys i +1)-IE(xs i ,ys i -1)|

[0108] Calculate P′ and Ps simultaneously i Spatial distance

[0109] Based on the gradient Gr i ,Gg i ,Gb i and distance d i , using the bilateral weighted average formula to calculate Ps i Interpolation weights on the three RGB channels:

[0110] wr i =exp(-α(d i / dc) 2 -β(Gr i / Grc) 2 )

[0111] wg i =exp(-α(d i / dc) 2 -β(Gg i / Ggc) 2 )

[0112] wb i =exp(-α(d i / dc) 2 -β(Gb i / Gbc) 2 )

[0113] Among them, α and β are the adjustment factors of distance weight and gradient weight respectively, dc, Grc, Ggc, and Gbc are {d i}、{Gr i}, {Gg i}, {Gb i Finally, the interpolation results of the three RGB channels at P′ are:

[0114] R′=∑wr i ×R i ,G′=∑wg i ×Gs i ,B′=∑wb i ×Bs i

[0115] Where Rs i , Gs i , Bs i Ps i RGB chromaticity value at .

[0116] Specifically, the color consistency evaluation index is defined as: color difference Where c∈{R,G,B} represents the color channel, Ec(x,y) is the chromaticity value of the pixel point P(x,y) in the c channel, is the average chromaticity value of the c channel, M and N are the width and height of the image respectively;

[0117] Mean square error in is the mean chromaticity of the c channel in the local small area I of the image, Ml and Nl are the width and height of the local small area;

[0118] Color gamut coverage R = (∑Vi) / V D , where Vi is the volume of the i-th sub-region of the color gamut space after quantization, V D The displayP3 standard color gamut volume;

[0119] Comprehensive score S = λ Δ (1-ΔE / ΔE max )+λ U (1-U / U max )+λ R R, where λ Δ , U , R are the weighting coefficients of ΔE, U, and R, respectively. ΔE m ax、U max are the maximum values ​​of ΔE and U respectively.

[0120] Chroma interpolation restoration is to restore the chroma information of unsampled pixels in an image from adjacent sampled points. Based on the traditional nearest neighbor interpolation and trilinear interpolation, this paper proposes an adaptive weighted average strategy. For the point to be interpolated P′, first select k nearest sampled points {Ψ i}; then calculate P′ and Ψ respectively i In the chromaticity gradient of the three channels of RGB i ,Gg i ,Gb i}, and the spatial distance d between P′ and Ψ i ; Under the combined effect of gradient and distance, Ψ is adaptively determined i The contribution weight to the chromaticity restoration of P′ is i ,wg i ,wb i}, the closer the distance and the smaller the gradient, i The larger the weight, the chromaticity value of P′ on the three RGB channels is equal to Ψ i Weighted average of chromaticity values. This algorithm combines color similarity and spatial proximity, and adaptively adjusts the weights of color transition areas, which effectively protects edge details while suppressing noise, and effectively improves color distortion and blurring problems.

[0121] Specifically, the visualization method includes:

[0122] The chromaticity deviation ΔE(x,y) of the pixel point P(x,y) is mapped to a grayscale value to generate a chromaticity deviation distribution map;

[0123] Map the quantized cluster center number j corresponding to the pixel point P(x,y) to a pseudo color to generate a chromaticity cluster distribution map;

[0124] Using the four indicator values ​​of ΔE, U, R, and S as the radius, a radar chart of comprehensive evaluation indicators in the four quadrants is drawn.

[0125] The evaluation indicators include color difference ΔE, mean square error U, color gamut coverage R, comprehensive score S, etc. The color performance formula of the display screen is examined from three levels: pixel level, local area, and overall. Δ ,λ U ,λ R The recommended values ​​are 0.4, 0.3, 0.3, ΔE maxι U max is the reference maximum value of ΔE and U, which can be set based on experience. The visualization charts include chromaticity deviation distribution chart, quantitative cluster distribution chart, comprehensive evaluation index radar chart, etc., which intuitively present color defects from three levels: pixel-level distortion, regional transition, and overall quality. The combination of these quantitative and qualitative analysis methods enhances the interpretability and guiding significance of the evaluation results.

[0126] Specifically, the Zhang Zhengyou calibration method is used for distortion correction, the 5×5 adaptive median filter is used for noise filtering, and the limited contrast adaptive histogram equalization algorithm is used for histogram equalization.

[0127] Specifically, the parameters of the high-resolution industrial camera are: Basler acA4600-10uc model, resolution of 4608×3288, pixel size of 1.55μm×1.55μm, target size of 1 / 1.1", peak quantum efficiency greater than 70%, maximum frame rate of 10fps, USB3.0 interface, and C-mount lens interface. The camera is equipped with a 1-inch CMOS sensor, with a high signal-to-noise ratio of 71dB and a low readout noise of 2.2, which can meet the requirements of high-fidelity acquisition. Distortion correction uses Zhang Zhengyou calibration algorithm with more than 20 control points to correct the distortion caused by lens distortion. The nonlinear geometric distortion is controlled within 0.5 pixels; the image denoising uses a 5×5 adaptive median filter, and the window coefficient is dynamically adjusted according to the local area grayscale statistics, which can achieve a good balance between smoothing and edge preservation; the histogram equalization uses the limited contrast adaptive histogram equalization algorithm, first equalizes the 32×32 local window and the global histogram separately, and then adaptively merges them, which can improve the overall image contrast and local details at the same time. The parameter settings of these algorithms balance the computational efficiency and effect, and can provide a high-quality data foundation for subsequent processing.

[0128] Please participate Figure 2 The present invention also discloses a high color gamut display screen pixel lighting chromaticity information acquisition system, comprising:

[0129] A high-resolution industrial camera 1 is used to collect raw image data when the display screen to be tested is lit up as a full white screen;

[0130] A preprocessing module 2, used for performing distortion correction, noise filtering, and histogram equalization on the original image data, and outputting preprocessed image data;

[0131] Positioning module 3, used for performing region segmentation and sub-pixel edge extraction on the pre-processed image data to obtain a set of pixel edge coordinates of the display screen to be tested;

[0132] A chromaticity mapping module 4 is used to construct an adaptive mapping model from the RGB color space to the color gamut space of the display screen, and to perform color gamut space mapping on the RGB value of each pixel in the pixel coordinate set based on the mapping model;

[0133] Quantization module 5, used for performing adaptive hierarchical quantization on the mapped color gamut space using an octree algorithm, and performing local k-means clustering on each quantized sub-region;

[0134] Evaluation module 6, used to calculate color consistency indicators such as color difference, local mean square error, quantitative color gamut volume coverage and comprehensive score of the RGB chromaticity of the pixel point;

[0135] The visualization module 7 is used to generate a deviation distribution map, a cluster distribution map and a comprehensive evaluation radar map reflecting the color quality of the display screen based on the RGB chromaticity deviation of the pixel points, the cluster labels and the color consistency index.

[0136] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modification, replacement, and improvement made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for collecting pixel lighting chromaticity information of a high color gamut display screen, characterized in that: The following steps are involved: Step 1. Preprocessing the display screen: Use a high-resolution industrial camera to capture the full white screen of the display screen to be tested, and obtain the original image data I(x, y), where x represents the pixel coordinate in the image width direction, y represents the pixel coordinate in the image height direction, x∈[1, W], y∈[1, H], W and H are the number of pixels of the image width and height respectively; then perform distortion correction, noise filtering, and histogram equalization on the original image data I(x, y) in sequence to obtain the preprocessed image data IE(x, y); Step 2: Pixel point location, perform region growing segmentation on the preprocessed image data IE(x, y) to obtain a pixel block region set {R1, R2, ..., R n }; Then for each pixel block area R i , the Zemike moment descriptor is used to extract the sub-pixel position of the edge contour, and the pixel edge coordinate set P = {P1, P2, ..., P M }, where M is the total number of physical pixels on the display; Step 3: Chromaticity information extraction: for the pixel coordinate set P, an adaptive mapping model f(r, g, b) from the RGB color space to the display color gamut space is constructed; based on f(r, g, b), the RGB value of each pixel is mapped to the color gamut space coordinate (x, y, Y); then the color gamut space is adaptively hierarchically quantized using the octree algorithm to obtain N quantized sub-regions {D1, D2, ..., D N }; Then for each sub-area D i The color space coordinate points within the local k-means clustering are obtained. i Cluster Centers Finally, a three-layer index structure from color gamut space coordinates to quantized cluster centers is established; Step 4. Color consistency evaluation: calculate the color difference ΔE between the RGB chromaticity value of each pixel and the RGB average chromaticity value of the display, the RGB chromaticity mean square error U in a small local area, the ratio of the quantized color gamut space volume to the complete color gamut volume of the display R, and the comprehensive score S of weighted color difference, uniformity, and color gamut coverage as quantitative indicators to measure the color consistency of the display; Step 5. Visualize the chromaticity results. Generate a chromaticity deviation distribution map based on the pixel RGB chromaticity deviation value ΔE. Generate a chromaticity cluster distribution map based on the pixel quantitative cluster label. Draw a comprehensive evaluation radar map based on the color consistency indicators ΔE, U, R, and S to intuitively display the color quality characteristics of the display.

2. The method according to claim 1, characterized in that The construction of the adaptive chromaticity mapping model comprises the following steps: For K different models of reference displays, the chromaticity distribution samples of the three primary colors of RGB in different color gamut modes are collected to form a reference color gamut sample set {S1, S2, ..., S K }; For each reference color gamut sample S j , extract the mean μ of the three primary colors RGB rj , μ gj , μ bj , variance σ rj , σ gj , σ bj , color gamut coverage Cr j , Cg j , Cb j , boundary point set Br j , Bg j , Bb j , saturation histogram Hr j , Hg j , Hb j Statistical features constitute the color gamut description feature vector F j ; With {F1, F2, ..., F K } is a training sample, RGB value (r, g, b) is input, color space coordinates (x, y, Y) is output, and Gaussian kernel function K(F i , F j )=exp(-γ||F i -F j || 2 ) is trained with a support vector machine regression model to obtain a nonlinear mapping model f(r, g, b) from RGB to color gamut space; For the display screen to be tested, according to its factory parameters and usage time information, the most similar color gamut description feature vector F is matched in the reference color gamut sample set. * , then F * Substituting into the mapping model f(r, g, b), an adaptive mapping relationship from RGB to color gamut space suitable for the display screen can be obtained.

3. The method according to claim 1, characterized in that The specific steps of the color gamut space adaptive quantization include: The displayP3 standard color gamut space is recursively divided into 8 sub-cubes according to the median values ​​of the three color components R, G, and B, until the standard deviation σ of the RGB components in each sub-cube is r , σ g , σ b are all less than the threshold TS or the number of divisions exceeds the maximum value Nmax, thereby obtaining the adaptive octree hierarchical quantization result {D1, D2, ..., D N }; For each sub-area D i ,make Where Y li , Y hi 、x li 、x hi Respectively represent D i The coordinate span in the luminance Y direction and chrominance x direction, Q Y , Q x Represents the quantization step size of brightness and chrominance respectively, K i D i The number of clusters within Then use K i The k-means algorithm for D i Cluster all color gamut space coordinate points within K i Cluster Centers Each cluster center c ij Contains the brightness value Y ij and chromaticity value (x ij ,y ij ); Finally, the sub-region number i, cluster center number j, brightness and chromaticity value (Y ij , x ij ,y ij ) is used as the index to construct a three-layer index structure, which facilitates the rapid mapping from the color gamut space coordinates to the quantitative clustering centers.

4. The method according to claim 1, characterized in that: The step of compressing the chrominance data comprises: The quantitative cluster center data of all sub-regions {c ij |1≤i≤N,1≤j≤K i } Perform k-means++ clustering to obtain M cluster centers {o1, o2, ..., o M }, the cluster distance uses the CIEDE2000 color difference formula: Where ΔL′, ΔC′, and ΔH′ represent the differences in brightness, chromaticity, and hue, respectively, and k L , k C , k H is the correction factor, S L , S C , S H is the weighting function, R T is the rotation function; Then for each cluster center o i , extract all cluster center sample points c whose CIEDE2000 color difference is less than the threshold TD ij , these c ij The original chromaticity value (Y ij , x ij ,y ij ) are all replaced by o i The chromaticity value (Y i , x i ,y i ), to achieve the merging of similar chromaticities; Then count each o i The frequency of repeated occurrence of the merged chromaticity values ​​is used to construct a Huffman tree with the probability of occurrence ranging from large to small, and Huffman coding is performed. The higher the frequency of occurrence of the chromaticity value, the shorter the number of bits of the coding; Finally, the compressed data is obtained after similar chromaticity merging and Huffman coding of the quantized cluster center data.

5. The method according to claim 1, characterized in that The chromaticity interpolation method comprises the following steps: For the pixel point P′(x′, y′) to be interpolated, search for the k nearest sampled pixel points {Ps1, Ps2, ..., Ps k }; Calculate P′ and Ps respectively i Chroma gradient on the three RGB channels: GB i =|IE(xs i +1, yes i +1)-IE(xs i -1, yes i -1)| Gr i =|IE(xs i +1,ys i )-IE(xs i -1,ys i )| Gg i =|IE(xs i ,ys i +1)-IE(xs i ,ys i -1)| Calculate P′ and Ps simultaneously i Spatial distance Based on the gradient Gr i , Gg i , Gb i and distance d i , Ps is calculated using the bilateral weighted average formula i Interpolation weights on the three RGB channels: wr i =exp(-α(d i / dc) 2 -β(Gr i / Grc) 2 ) wg i =exp(-α(d i / dc) 2 -β(Gg i / Ggc) 2 ) wb i =exp(-α(d i / A.D) 2 -β(Gb i ( / GBC) 2 ) Among them, α and β are the adjustment factors of distance weight and gradient weight respectively, dc, Grc, Ggc, and Gbc are {d i }、{Gr i }, {Gg i }, {Gb i Finally, the interpolation results of the three RGB channels at P′ are: R′=∑wr i ×Rs i ,G′=∑wg i ×Gs i ,B′=∑wb i ×Bs i Where Rs i , Gs i , Bs i Ps i RGB chromaticity value at .

6. The method according to claim 1, characterized in that The color consistency evaluation index is defined as: color difference Where c∈{R, G, B} represents the color channel, Ec(x, y) is the chromaticity value of the pixel point P(x, y) in the c channel, is the average chromaticity value of the c channel, M and N are the width and height of the image respectively; Mean square error in is the mean chromaticity of the c channel in the local small area I of the image, Ml and Nl are the width and height of the local small area; Color gamut coverage R = (∑Vi) / V D , where Vi is the volume of the i-th sub-region of the color gamut space after quantization, V D The displayP3 standard color gamut volume; Comprehensive score S = λ Δ (1-ΔE / ΔE max )+λ U (1-U / U max )+λ R R, where λ Δ , U , R are the weighting coefficients of ΔE, U, and R, respectively. ΔE m ax, U max are the maximum values ​​of ΔE and U respectively.

7. The method according to claim 6, characterized in that The visualization method comprises: Map the chromaticity deviation ΔE(x,y) of the pixel point P(x,y) to a grayscale value to generate a chromaticity deviation distribution map; Map the quantized cluster center number j corresponding to the pixel point P(x,y) to a pseudo color to generate a chromaticity cluster distribution map; Using the four indicator values ​​of ΔE, U, R, and S as the radius, a radar chart of comprehensive evaluation indicators in the four quadrants is drawn.

8. The method according to any one of claims 1 to 7, characterized in that: The parameters of the high-resolution industrial camera are: Basler acA4600-10uc model, resolution of 4608×3288, pixel size of 1.55μm×1.55μm, target size of 1 / 1.1”, peak quantum efficiency greater than 70%, maximum frame rate of 10fps, interface mode of USB3.0, and lens interface of C interface.

9. The method according to any one of claims 1 to 7, characterized in that: The Zhang Zhengyou calibration method is used for distortion correction, the 5×5 adaptive median filter is used for noise filtering, and the limited contrast adaptive histogram equalization algorithm is used for histogram equalization.

10. A high color gamut display screen pixel lighting chromaticity information acquisition system, characterized in that: include: High-resolution industrial camera, used to collect raw image data when the display screen to be tested is lit up as a full white screen; A preprocessing module, used to perform distortion correction, noise filtering, and histogram equalization on the original image data, and output preprocessed image data; A positioning module, used to perform region segmentation and sub-pixel edge extraction on the pre-processed image data to obtain a set of pixel edge coordinates of the display screen to be tested; A chromaticity mapping module, used to construct an adaptive mapping model from the RGB color space to the color gamut space of the display screen, and perform color gamut space mapping on the RGB value of each pixel in the pixel coordinate set based on the mapping model; A quantization module is used to perform adaptive hierarchical quantization on the mapped color space using an octree algorithm, and perform local k-means clustering on each quantized sub-region; An evaluation module is used to calculate color consistency indicators such as color difference, local mean square error, quantitative color gamut volume coverage and comprehensive score of the RGB chromaticity of the pixel point respectively; The visualization module is used to generate a deviation distribution map, a cluster distribution map and a comprehensive evaluation radar map reflecting the color quality of the display screen based on the RGB chromaticity deviation of the pixel point, the cluster label and the color consistency index.

Images