OLED screen color development optimization system and method based on color management engine

Abstract

The invention discloses an OLED screen color development optimization system and method based on a color management engine, and relates to the technical field of OLED display, and the method comprises the following steps: collecting ambient light related data through an OLED screen integrated sensor to form an ambient light characteristic data set, and calling an equipment characteristic file to extract constraint parameters to form an equipment capability constraint data set; the method comprises the following steps: acquiring original data of a display picture, processing through a pre-trained lightweight semantic segmentation model, extracting features, and integrating to form a picture content chromaticity and brightness data set; constructing a deep learning regression model by taking the environment light characteristic data set, the equipment capability constraint data set and the picture content chromaticity and brightness data set as input, outputting a color development degradation index after training, and determining a high-priority region in combination with a region priority weight; and setting an optimization target and a constraint condition, calling a multi-target constraint optimization algorithm to solve a high-priority region compensation parameter, and issuing the high-priority region compensation parameter to a driving module to execute color development optimization.

Description

Technical Field

[0001] This invention relates to the field of OLED display technology, specifically to an OLED screen color optimization system and method based on a color management engine. Background Technology

[0002] As a typical representative of self-emissive display technology, OLED screens have been widely used in smart terminal devices such as mobile phones, automotive displays, and tablets due to their advantages of high contrast, wide color gamut, and thinness and flexibility. The color management engine is one of the core technology modules of OLED screens, mainly ensuring color consistency through device characteristic calibration and color space conversion. Meanwhile, ambient light has a significant impact on display performance. In strong light environments, the reflection and scattering of ambient light alter the screen's apparent brightness and color perception, which is a common environmental adaptation challenge in the display technology field.

[0003] Existing OLED screen color optimization solutions mostly focus on color consistency calibration under normal lighting conditions, without designing specific solutions for the color degradation characteristics in strong light environments. Faced with the problems of color saturation attenuation and contrast reduction caused by strong light, existing technologies often adopt a crude optimization approach of increasing overall brightness, without considering the differentiated needs of core areas such as text and faces; furthermore, they fail to fully leverage the hardware characteristics of OLED screens, which limit peak brightness, easily leading to insufficient optimization effects or exceeding power consumption and lifespan constraints, thus failing to balance visual experience under strong light with safe device operation. Summary of the Invention

[0004] The purpose of this invention is to provide an OLED screen color optimization system and method based on a color management engine to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for optimizing the color display of an OLED screen based on a color management engine, the method comprising the following steps: Step 1: Collect ambient light-related data by integrating sensors into the OLED screen to form an ambient light characteristic dataset, and extract constraint parameters from the device characteristic file to form a device capability constraint dataset; Step 2: Obtain the original data of the display screen, process it through a pre-trained lightweight semantic segmentation model and extract features, and integrate it to form a color and brightness dataset of the screen content; Step 3: Construct a deep learning regression model using the ambient light characteristics dataset, device capability constraint dataset, and image content chromaticity and brightness dataset as input. After training, output the color degradation index and determine the high priority area by combining the regional priority weight. Step 4: Set the optimization objectives and constraints, call the multi-objective constraint optimization algorithm to solve the compensation parameters of the high-priority region, and send them to the driver module to perform color optimization.

[0006] In step 1, the optical characteristics of the current environment and the hardware boundaries of the device are obtained to provide basic data for subsequent degradation prediction and optimization parameter solution, and to avoid the optimization scheme from exceeding the capabilities of the device hardware. By integrating a spectral sensor, a brightness sensor, and an angle sensor into an OLED screen, ambient light spectral distribution data S(λ) (where λ represents the wavelength of visible light) and ambient brightness value L are collected, respectively. env 1. Ambient light incident angle θ; 2. Acquire ambient light reflection brightness L on the screen surface. ref This data is integrated to form an ambient light characteristics dataset. The color management engine calls the device characteristic file built into the OLED screen, reads the device constraint parameters, and forms a device capability constraint dataset, specifically including: the upper limit of panel peak brightness L. max Panel aging parameter A panel Real-time power consumption constraint P limit .

[0007] In step 2, salient and non-salient regions in the image are identified, and the color and brightness features of each region are extracted to provide input for the image content dimension for degradation prediction, thereby achieving precise optimization of the partitions. The color management engine obtains the raw image data of the current frame from the OLED screen display buffer and inputs it into the pre-trained lightweight semantic segmentation model. The pre-trained model uses the PP-LiteSeg algorithm (lightweight semantic segmentation technology), and the training set includes common display content such as text, faces, landscapes, and UI elements. The model extracts features and fuses them at multiple scales to output a pixel-level region mask M, thereby classifying image regions and distinguishing between salient regions (such as text, faces, and high-detail texture regions) and non-salient regions (such as solid color background regions). The semantic segmentation results are post-processed to remove noise (using the morphological erosion-dilation algorithm to remove isolated noise points in the segmentation results and ensure clear region boundaries). Output: region mask M, location coordinates of each region, region type label and area percentage of each region. Based on the semantic segmentation results, all pixels within each region are traversed. Color space conversion technology (using the standard RGB to HSV conversion formula: H represents hue, S=Max(R,G,B)-Min(R,G,B) / Max(R,G,B), V=Max(R,G,B) / 255) is used to convert the original pixel color data into HSV color space data: h represents hue, s represents saturation, and v represents brightness. Statistical analysis methods (calculating the arithmetic mean and variance) are then used to aggregate and calculate the HSV data within the region, obtaining region-level features, including the mean brightness L of each region. mean , Brightness variance, Saturation meanmean Saturation variance; Based on a pixel-level traversal process, the coordinates and corresponding HSV values ​​of each pixel are recorded. The HSV data of all pixels within a single region are arranged by coordinates to form a distribution matrix within the region, yielding pixel-level features, specifically including: the pixel-level brightness distribution matrix L of each region. local (x,y), pixel-level chromaticity distribution matrix C local (x,y); where x and y are pixel coordinates; The results of semantic segmentation are correlated and weighted according to a preset priority rule (preset priority rule: weight W for salient regions). p =1.0, weight W for non-significant regions p =0.5 (the rules are stored in the configuration file of the color management engine), to obtain semantic association features, specifically including: type labels of each region and region priority weights; By integrating region-level features, pixel-level features, and semantic association features, a color saturation and brightness dataset of the image content is obtained.

[0008] In step 3, the degree of color degradation in each region under strong light is predicted, the optimization needs of each region are quantified, high-priority regions are identified, and optimization resources are tilted towards key regions. Using ambient light characteristics dataset, device capability constraint dataset, and image content chromaticity and brightness dataset as inputs, and color degradation quantification index measured by a professional color analyzer (such as Konica Minolta CA-410) as sample labels, a deep learning regression model is constructed (the model adopts a CNN+Transformer hybrid architecture, with the input layer dimension adapted to the feature dimensions of the three datasets, and the output layer consisting of three fully connected neurons, corresponding to the three degradation indices respectively); the model outputs the color degradation quantification index. The training model learns the mapping relationship between inputs and outputs; the AdamW optimizer is used in the training process. The specific quantitative index of color degradation includes: saturation attenuation ΔS. loss ΔS loss Normalized values ​​between 0 and 1, where 0 represents no attenuation and 1 represents complete attenuation, are used to predict the degree of saturation reduction in images under strong light; contrast loss value ΔC loss ΔC loss The normalized value is between 0 and 1, where 0 represents no loss and 1 represents complete loss. It is used to predict the degree of contrast reduction caused by the combined effect of dark-field reflected light and bright-field peak limitation under strong light; the degree of loss of detail discrimination ΔD loss ΔD loss The normalized value is 0-1, where 0 indicates no decrease and 1 indicates complete blur. It is used to predict the degree of blurring of details under strong light. Integrate the trained lightweight model into the color management engine; By combining the region priority weights obtained from semantic segmentation with the degradation prediction values ​​output by the model, the comprehensive degradation index I for each region is calculated: I = W p ×(W C ×ΔC loss +W S ×ΔS loss +W D ×ΔD loss ); where W p W represents the region priority weight; C W S W D These represent the weights for contrast loss, saturation decay, and detail blurring, respectively. The weights are set based on the fact that contrast and saturation have a greater impact on visual experience than detail recognition. This weight allocation has been verified through a large number of user visual experiments. Collect all semantic region I values, sort the regions in descending order to form a region optimization priority list; set a filtering threshold T. I When the I value of a certain semantic region is greater than T I When this occurs, the region is defined as a high-priority region; the filtering threshold T I Through multiple experiments, it was determined that high-priority areas should cover the content that users are most concerned about. In step 4, within the hardware constraints, the optimal compensation parameters are solved for the high-priority areas to achieve accurate repair of color degradation while ensuring safe operation of the equipment. Set optimization goals: improve the perceptible saturation of high-priority areas to the preset standard, restore the image contrast to the baseline level outside of strong light environments, and meet the preset clarity threshold for detail recognition; Set constraints: The actual luminance of all pixels shall not exceed the upper limit of the panel's peak luminance L. max The overall power consumption of the screen does not exceed the real-time power consumption constraint P. limit The cumulative values ​​of pixel emission duration and brightness do not exceed the lifetime decay threshold. The lifespan degradation threshold is based on panel aging parameter A. panel Dynamic adjustment; Lifespan degradation threshold = New panel lifespan threshold × (1-A) panel ); The color management engine calls a multi-objective constraint optimization algorithm (the multi-objective constraint optimization algorithm adopts a non-dominated sorting genetic algorithm, with a population size of 100 and 50 iterations), inputting a list of region optimization priorities, color degradation prediction results, and equipment capability constraint dataset, and solves for the core optimization parameters for high-priority regions: Regional brightness compensation coefficient: based on the contrast loss value ΔC of high-priority areas. lossCalculate the brightness compensation coefficient kL for each high-priority area, kL = 1 + ΔC loss ×(L max -L mean ) / L mean ; Hue and saturation compensation parameters: based on the saturation attenuation ΔS in high-priority regions. loss Calculate the saturation enhancement coefficient kS for this region, kS=1 / (1-ΔS) loss Based on the color shift prediction results (the color shift prediction results come from the additional output of the deep learning regression model, which predicts the hue shift Δh under strong light), determine the corresponding hue correction value. Temporal pulse strategy parameters: combining the brightness requirements of high-priority areas with panel aging parameters A panel The pulse width and frequency of PWM dimming are determined (using a PWM dimming parameter matching algorithm, the target brightness and duty cycle are linearly positively correlated), so that the pixel emission duty cycle matches the brightness compensation requirements, and the single emission time of a single pixel does not exceed the lifetime decay threshold. The color management engine encapsulates the brightness compensation coefficient, hue and saturation compensation parameters, and temporal pulse strategy parameters of high-priority areas obtained from multi-objective constraint optimization into a dedicated optimization parameter package for high-priority areas, and sends it to the display driver module of the OLED screen (the display driver module is the built-in driver chip of the OLED screen). After receiving the parameters, the display driver module performs color rendering and brightness adjustment for each pixel in the high-priority area according to the position coordinates of the high-priority area: it corrects the pixel color attributes according to the hue and saturation compensation parameters, adjusts the pixel emission brightness according to the brightness compensation coefficient, and controls the pixel emission timing in combination with the temporal pulse strategy parameters to complete the color optimization of the high-priority area.

[0009] An OLED screen color optimization system based on a color management engine, the system includes an ambient light acquisition module, an image analysis module, a degradation prediction module, and an optimization execution module; The ambient light acquisition module is used to collect ambient light-related data through sensors integrated into the OLED screen to form an ambient light characteristic dataset, and to extract constraint parameters from the device characteristic file to form a device capability constraint dataset. The image analysis module is used to acquire the original data of the display image, process it through a pre-trained lightweight semantic segmentation model and extract features, and integrate it to form an image content chroma and brightness dataset. The degradation prediction module is used to construct a deep learning regression model with ambient light characteristics dataset, device capability constraint dataset, and image content chromaticity and brightness dataset as input. After training, it outputs a color degradation index and determines high-priority areas by combining regional priority weights. The optimization execution module is used to set optimization objectives and constraints, call a multi-objective constraint optimization algorithm to solve for compensation parameters in high-priority regions, and send the results to the driving module to perform color optimization.

[0010] The ambient light acquisition module includes an ambient light acquisition unit and a device constraint extraction unit; The ambient light acquisition unit is used to acquire ambient light-related data through sensors integrated into the OLED screen to form an ambient light characteristic dataset; the device constraint extraction unit is used to call the device characteristic file, extract device constraint parameters, and form a device capability constraint dataset.

[0011] The image analysis module includes an image acquisition unit, a semantic segmentation unit, and a feature extraction unit; The image acquisition unit is used to acquire the original image data of the current frame from the OLED screen display buffer; the semantic segmentation unit is used to process the image through a pre-trained lightweight semantic segmentation model to classify the image regions and output region masks and type labels; the feature extraction unit is used to extract the chromaticity and brightness features of the image content based on the semantic segmentation results to form an image content chromaticity and brightness dataset.

[0012] The degradation prediction module includes a model training unit, a degradation prediction unit, and a region priority unit. The model training unit is used to construct a deep learning regression model and train the model to learn the mapping relationship between the input and the color degradation index; the degradation prediction unit is used to use the trained model to predict the color degradation index, including saturation decay, contrast loss and decrease in detail recognition; the region priority unit is used to combine the region priority weight and the degradation prediction value to calculate the comprehensive degradation index and determine the high priority region.

[0013] The optimization execution module includes an optimization target setting unit, an optimization solution unit, and a parameter distribution unit; The optimization target setting unit is used to set the target and constraints for color optimization; the optimization solution unit is used to call a multi-objective constraint optimization algorithm to solve for the compensation parameters of high-priority regions; the parameter distribution unit is used to encapsulate the optimization parameters and distribute them to the display driver module to perform color optimization.

[0014] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention identifies high-priority regions such as text and faces through semantic segmentation, and accurately predicts the degree of saturation attenuation and contrast loss in each region under strong light by combining a deep learning regression model. It can focus on key areas to carry out color optimization, avoiding the waste of resources caused by overall optimization in existing solutions. This invention calculates the brightness compensation coefficient based on the contrast loss value through a linear mapping algorithm, and controls the compensation range by combining the upper limit constraint of the panel's peak brightness. Under the premise that the peak brightness of the OLED screen is limited, it can accurately offset the problem of increased brightness in dark fields caused by strong light reflection, and effectively restore the image contrast of high-priority areas. This invention is based on the existing color management engine architecture and OLED screen hardware modules, without the need for additional hardware equipment, and can be directly integrated into existing display systems, improving the compatibility and industrial applicability of the solution. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating an OLED screen color optimization system based on a color management engine, according to the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Example: Figure 1 As shown, the present invention provides a technical solution, a method for optimizing the color display of an OLED screen based on a color management engine, the method comprising the following steps: Step 1: Collect ambient light-related data by integrating sensors into the OLED screen to form an ambient light characteristic dataset, and extract constraint parameters from the device characteristic file to form a device capability constraint dataset; Step 2: Obtain the original data of the display screen, process it through a pre-trained lightweight semantic segmentation model and extract features, and integrate it to form a color and brightness dataset of the screen content; Step 3: Construct a deep learning regression model using the ambient light characteristics dataset, device capability constraint dataset, and image content chromaticity and brightness dataset as input. After training, output the color degradation index and determine the high priority area by combining the regional priority weight. Step 4: Set the optimization objectives and constraints, call the multi-objective constraint optimization algorithm to solve the compensation parameters of the high-priority region, and send them to the driver module to perform color optimization.

[0018] In step 1, the optical characteristics of the current environment and the hardware boundaries of the device are obtained to provide basic data for subsequent degradation prediction and optimization parameter solution, and to avoid the optimization scheme from exceeding the capabilities of the device hardware. By integrating a spectral sensor, a brightness sensor, and an angle sensor into an OLED screen, ambient light spectral distribution data S(λ) (where λ represents the wavelength of visible light) and ambient brightness value L are collected, respectively. env 1. Ambient light incident angle θ; 2. Acquire ambient light reflection brightness L on the screen surface. ref This data is integrated to form an ambient light characteristics dataset. The color management engine calls the device characteristic file built into the OLED screen, reads the device constraint parameters, and forms a device capability constraint dataset, specifically including: the upper limit of panel peak brightness L. max Panel aging parameter A panel Real-time power consumption constraint P limit .

[0019] In step 2, salient and non-salient regions in the image are identified, and the color and brightness features of each region are extracted to provide input for the image content dimension for degradation prediction, thereby achieving precise optimization of the partitions. The color management engine obtains the raw image data of the current frame from the OLED screen display buffer and inputs it into the pre-trained lightweight semantic segmentation model. The pre-trained model uses the PP-LiteSeg algorithm (lightweight semantic segmentation technology), and the training set includes common display content such as text, faces, landscapes, and UI elements. The model extracts features and fuses them at multiple scales to output a pixel-level region mask M, thereby classifying image regions and distinguishing between salient regions (such as text, faces, and high-detail texture regions) and non-salient regions (such as solid color background regions). The semantic segmentation results are post-processed to remove noise (using the morphological erosion-dilation algorithm to remove isolated noise points in the segmentation results and ensure clear region boundaries). Output: region mask M, location coordinates of each region, region type label and area percentage of each region. Based on the semantic segmentation results, all pixels within each region are traversed. Color space conversion technology (using the standard RGB to HSV conversion formula: H represents hue, S=Max(R,G,B)-Min(R,G,B) / Max(R,G,B), V=Max(R,G,B) / 255) is used to convert the original pixel color data into HSV color space data: h represents hue, s represents saturation, and v represents brightness. Statistical analysis methods (calculating the arithmetic mean and variance) are then used to aggregate and calculate the HSV data within the region, obtaining region-level features, including the mean brightness L of each region. mean , Brightness variance, Saturation mean meanSaturation variance; Based on a pixel-level traversal process, the coordinates and corresponding HSV values ​​of each pixel are recorded. The HSV data of all pixels within a single region are arranged by coordinates to form a distribution matrix within the region, yielding pixel-level features, specifically including: the pixel-level brightness distribution matrix L of each region. local (x,y), pixel-level chromaticity distribution matrix C local (x,y); where x and y are pixel coordinates; The results of semantic segmentation are correlated and weighted according to a preset priority rule (preset priority rule: weight W for salient regions). p =1.0, weight W for non-significant regions p =0.5 (the rules are stored in the configuration file of the color management engine), to obtain semantic association features, specifically including: type labels of each region and region priority weights; By integrating region-level features, pixel-level features, and semantic association features, a color saturation and brightness dataset of the image content is obtained.

[0020] In step 3, the degree of color degradation in each region under strong light is predicted, the optimization needs of each region are quantified, high-priority regions are identified, and optimization resources are tilted towards key regions. Using ambient light characteristics dataset, device capability constraint dataset, and image content chromaticity and brightness dataset as inputs, and color degradation quantification index measured by a professional color analyzer (such as Konica Minolta CA-410) as sample labels, a deep learning regression model is constructed (the model adopts a CNN+Transformer hybrid architecture, with the input layer dimension adapted to the feature dimensions of the three datasets, and the output layer consisting of three fully connected neurons, corresponding to the three degradation indices respectively); the model outputs the color degradation quantification index. The training model learns the mapping relationship between inputs and outputs; the AdamW optimizer is used in the training process. The specific quantitative index of color degradation includes: saturation attenuation ΔS. loss ΔS loss Normalized values ​​between 0 and 1, where 0 represents no attenuation and 1 represents complete attenuation, are used to predict the degree of saturation reduction in images under strong light; contrast loss value ΔC loss ΔC loss The normalized value is between 0 and 1, where 0 represents no loss and 1 represents complete loss. It is used to predict the degree of contrast reduction caused by the combined effect of dark-field reflected light and bright-field peak limitation under strong light; the degree of loss of detail discrimination ΔD loss ΔD loss The normalized value is 0-1, where 0 indicates no decrease and 1 indicates complete blur. It is used to predict the degree of blurring of details under strong light. Integrate the trained lightweight model into the color management engine; By combining the region priority weights obtained from semantic segmentation with the degradation prediction values ​​output by the model, the comprehensive degradation index I for each region is calculated: I = W p ×(W C ×ΔC loss +W S ×ΔS loss +W D ×ΔD loss ); where W p W represents the region priority weight; C W S W D These represent the weights for contrast loss, saturation decay, and detail blurring, respectively. The weights are set based on the fact that contrast and saturation have a greater impact on visual experience than detail recognition. This weight allocation has been verified through a large number of user visual experiments. Collect all semantic region I values, sort the regions in descending order to form a region optimization priority list; set a filtering threshold T. I When the I value of a certain semantic region is greater than T I When this occurs, the region is defined as a high-priority region; the filtering threshold T I Through multiple experiments, it was determined that high-priority areas should cover the content that users are most concerned about. In step 4, within the hardware constraints, the optimal compensation parameters are solved for the high-priority areas to achieve accurate repair of color degradation while ensuring safe operation of the equipment. Set optimization goals: improve the perceptible saturation of high-priority areas to the preset standard, restore the image contrast to the baseline level outside of strong light environments, and meet the preset clarity threshold for detail recognition; Set constraints: The actual luminance of all pixels shall not exceed the upper limit of the panel's peak luminance L. max The overall power consumption of the screen does not exceed the real-time power consumption constraint P. limit The cumulative values ​​of pixel emission duration and brightness do not exceed the lifetime decay threshold. The lifespan degradation threshold is based on panel aging parameter A. panel Dynamic adjustment; Lifespan degradation threshold = New panel lifespan threshold × (1-A) panel ); The color management engine calls a multi-objective constraint optimization algorithm (the multi-objective constraint optimization algorithm adopts a non-dominated sorting genetic algorithm, with a population size of 100 and 50 iterations), inputting a list of region optimization priorities, color degradation prediction results, and equipment capability constraint dataset, and solves for the core optimization parameters for high-priority regions: Regional brightness compensation coefficient: based on the contrast loss value ΔC of high-priority areas. loss Calculate the brightness compensation coefficient kL for each high-priority area, kL = 1 + ΔCloss ×(L max -L mean ) / L mean ; Hue and saturation compensation parameters: based on the saturation attenuation ΔS in high-priority regions. loss Calculate the saturation enhancement coefficient kS for this region, kS=1 / (1-ΔS) loss Based on the color shift prediction results (the color shift prediction results come from the additional output of the deep learning regression model, which predicts the hue shift Δh under strong light), determine the corresponding hue correction value. Temporal pulse strategy parameters: combining the brightness requirements of high-priority areas with panel aging parameters A panel The pulse width and frequency of PWM dimming are determined (using a PWM dimming parameter matching algorithm, the target brightness and duty cycle are linearly positively correlated), so that the pixel emission duty cycle matches the brightness compensation requirements, and the single emission time of a single pixel does not exceed the lifetime decay threshold. The color management engine encapsulates the brightness compensation coefficient, hue and saturation compensation parameters, and temporal pulse strategy parameters of high-priority areas obtained from multi-objective constraint optimization into a dedicated optimization parameter package for high-priority areas, and sends it to the display driver module of the OLED screen (the display driver module is the built-in driver chip of the OLED screen). After receiving the parameters, the display driver module performs color rendering and brightness adjustment for each pixel in the high-priority area according to the position coordinates of the high-priority area: it corrects the pixel color attributes according to the hue and saturation compensation parameters, adjusts the pixel emission brightness according to the brightness compensation coefficient, and controls the pixel emission timing in combination with the temporal pulse strategy parameters to complete the color optimization of the high-priority area.

[0021] An OLED screen color optimization system based on a color management engine, the system includes an ambient light acquisition module, an image analysis module, a degradation prediction module, and an optimization execution module; The ambient light acquisition module is used to collect ambient light-related data through sensors integrated into the OLED screen to form an ambient light characteristic dataset, and to extract constraint parameters from the device characteristic file to form a device capability constraint dataset. The image analysis module is used to acquire the original data of the display image, process it through a pre-trained lightweight semantic segmentation model and extract features, and integrate it to form an image content chroma and brightness dataset. The degradation prediction module is used to construct a deep learning regression model with ambient light characteristics dataset, device capability constraint dataset, and image content chromaticity and brightness dataset as input. After training, it outputs a color degradation index and determines high-priority areas by combining regional priority weights. The optimization execution module is used to set optimization objectives and constraints, call a multi-objective constraint optimization algorithm to solve for compensation parameters in high-priority regions, and send the results to the driving module to perform color optimization.

[0022] The ambient light acquisition module includes an ambient light acquisition unit and a device constraint extraction unit; The ambient light acquisition unit is used to acquire ambient light-related data through sensors integrated into the OLED screen to form an ambient light characteristic dataset; the device constraint extraction unit is used to call the device characteristic file, extract device constraint parameters, and form a device capability constraint dataset.

[0023] The image analysis module includes an image acquisition unit, a semantic segmentation unit, and a feature extraction unit; The image acquisition unit is used to acquire the original image data of the current frame from the OLED screen display buffer; the semantic segmentation unit is used to process the image through a pre-trained lightweight semantic segmentation model to classify the image regions and output region masks and type labels; the feature extraction unit is used to extract the chromaticity and brightness features of the image content based on the semantic segmentation results to form an image content chromaticity and brightness dataset.

[0024] The degradation prediction module includes a model training unit, a degradation prediction unit, and a region priority unit. The model training unit is used to construct a deep learning regression model and train the model to learn the mapping relationship between the input and the color degradation index; the degradation prediction unit is used to use the trained model to predict the color degradation index, including saturation decay, contrast loss and decrease in detail recognition; the region priority unit is used to combine the region priority weight and the degradation prediction value to calculate the comprehensive degradation index and determine the high priority region.

[0025] The optimization execution module includes an optimization target setting unit, an optimization solution unit, and a parameter distribution unit; The optimization target setting unit is used to set the target and constraints for color optimization; the optimization solution unit is used to call a multi-objective constraint optimization algorithm to solve for the compensation parameters of high-priority regions; the parameter distribution unit is used to encapsulate the optimization parameters and distribute them to the display driver module to perform color optimization.

[0026] In this embodiment, under direct sunlight outdoors, architectural construction drawings are viewed on an OLED tablet. The drawings contain key information such as layout lines, dimension annotations, axis numbers, and building outlines. However, in this strong backlighting scenario, the OLED tablet screen exhibits significant technical problems: the contrast of the layout lines decreases; the saturation of the dimension annotation text is significantly reduced, blurring the text-to-background boundary; and some detailed information such as axis numbers is lost, making them difficult to quickly identify. The present invention provides an OLED screen color display optimization method based on a color management engine to improve the above-mentioned problems: Step 1: The spectrum sensor, brightness sensor, and angle sensor integrated in the flat OLED screen start working simultaneously: the spectrum sensor collects ambient light spectral distribution data S(λ), and the brightness sensor collects ambient brightness values ​​L. env =9000cd / m², the angle sensor collects the ambient light incident angle θ=32°, and simultaneously collects the ambient light reflected brightness L on the screen surface through the sensor. ref =138cd / m 2 The above data will be integrated to form an ambient light characteristics dataset; The color management engine calls the tablet's built-in device characteristic file, reads device constraint parameters, and forms a device capability constraint dataset. The specific parameters are: panel peak brightness limit L. max =1500cd / m 2 Panel aging parameter A panel =0.07 (mild aging state), real-time power consumption constraint P limit =17W, define the hardware boundaries to avoid the optimization scheme from exceeding the equipment's carrying capacity; Step 2: The color management engine obtains the raw image data (2560×1600 resolution) of the current construction drawing from the display buffer of the flat OLED screen and inputs it into a pre-trained lightweight semantic segmentation model. The training set of this model includes common professional content such as architectural construction drawings and engineering documents. Through feature extraction and multi-scale feature fusion, it outputs a pixel-level region mask M to achieve accurate classification of image regions: layout lines, dimension annotation text, and axis numbers are classified as salient regions (requiring key optimization), while blank backgrounds and secondary fill color blocks are classified as non-salient regions (not requiring key optimization).

[0027] The semantic segmentation results are processed by morphological erosion-dilation algorithm to remove noise, and isolated noise points (such as line edge noise and noise around the text) are removed. The output region mask M, the position coordinates of each region, the region type label, and the area ratio of each region (23% for salient regions and 77% for non-salient regions).

[0028] Based on the semantic segmentation results, all pixels within each region are traversed, and the original RGB color data is converted to HSV color space data using the standard RGB-to-HSV conversion formula (H is hue, S=(Max(R,G,B)-Min(R,G,B)) / Max(R,G,B), V=Max(R,G,B) / 255). Statistical analysis methods calculating the arithmetic mean and variance are then used to aggregate and calculate the HSV data within each region, yielding region-level features: the mean brightness L of salient regions. mean =390cd / m², luminance variance =89, saturation mean S mean=0.51, saturation variance =0.11; simultaneously record the coordinates and corresponding HSV values ​​of each pixel, arrange the HSV data of all pixels in a single region according to coordinates to form a distribution matrix, and obtain the pixel-level features, that is, the pixel-level brightness distribution matrix L of each salient region. local (x,y), pixel-level chromaticity distribution matrix C local (x, y) (x, y are pixel coordinates); Associating the region type labels output by semantic segmentation with preset priority rules (salience region weight W) p1 =1.0, weight W for non-significant regions p2 =0.5 (the rules are stored in the color management engine configuration file), to obtain semantic association features (including region type labels and region priority weights). Finally, the region-level features, pixel-level features, and semantic association features are integrated to form the image content chroma and brightness dataset.

[0029] Step 3: Using the ambient light characteristics dataset, device capability constraint dataset, and image content chromaticity and brightness dataset as input, and the color degradation quantification index measured by the Konica Minolta CA-410 professional color analyzer as sample labels, a deep learning regression model with a CNN+Transformer hybrid architecture is constructed. The model's output layer consists of three fully connected neurons, corresponding to the three types of color degradation quantification indices. The AdamW optimizer is used during training to enable the model to learn the mapping relationship between input and output. The trained model outputs a quantitative index of color degradation in each region: the contrast loss value ΔC in the area of ​​arranged lines. loss =0.74, saturation decay ΔS loss =0.60, the degree of decrease in detail recognition ΔD loss =0.56; ΔC of the dimension annotation text area loss =0.76、ΔS loss =0.63、ΔD loss =0.59; ΔC in the non-significant background region loss =0.33、ΔS loss =0.27、ΔD loss =0.24; Combining regional priority weights, according to the formula I=W p ×(W C ×ΔC loss +W S ×ΔS loss +W D ×ΔD loss ) Calculate the comprehensive degradation index (where W) C =0.4, W S =0.4, W D=0.2, this weight allocation has been verified by user visual experiments): Line layout area I=1.0×(0.4×0.74+0.4×0.60+0.2×0.56)=0.648; Dimension text area I=1.0×(0.4×0.76+0.4×0.63+0.2×0.59)=0.674; Background area I=0.5×(0.4×0.33+0.4×0.27+0.2×0.24)=0.144; Set the filtering threshold T I =0.3, the I value of the layout line area, dimension annotation text area, and axis numbering area are all greater than T. I The region is identified as a high-priority optimization region; the background region's I value is less than T. I It is not included in the key optimization scope; Step 4: Set optimization goals: Increase the perceptible saturation of high-priority areas to the preset standard (S≥0.9), restore the image contrast to the baseline level outside of strong light environments (contrast ratio≥42:1), and ensure that detail recognition meets the preset clarity threshold (ΔD). loss ≤0.1); Set constraints: the actual luminous intensity of all pixels does not exceed the upper limit of the panel's peak brightness of 1500 cd / m². 2 The overall power consumption of the screen does not exceed the real-time power consumption constraint of 17W; the cumulative value of pixel emission duration and brightness does not exceed the lifespan decay threshold (lifespan decay threshold = new panel lifespan threshold × (1-A)). panel = new panel lifespan threshold × 0.93). The color management engine calls a non-dominated sorting genetic algorithm (population size 100, iterations 50) to solve for the core optimization parameters for high-priority regions, taking into account the region optimization priority list, color degradation prediction results, and device capability constraint dataset. Area brightness compensation coefficient: based on the contrast loss value ΔC of the area where lines are arranged. loss =0.74, according to formula k L =1+ΔC loss ×(L max -L mean ) / L mean Calculate and obtain k L =1 + 0.74 × (1500 - 390) / 390 ≈ 3.11, after compensation, the brightness of this area = 390 × 3.11 = 1212.9 cd / m² 2 (not exceeding L) max =1500cd / m 2 ); Hue and saturation compensation parameters: saturation attenuation ΔS based on the dimensioned text region loss =0.63, according to formula k S=1 / (1-ΔS loss ) Calculate and obtain k S =1 / (1-0.63)≈2.70, after adjustment the saturation of this area S=0.51×2.70=1.377; based on the hue offset Δh=+2.8° of the additional output of the deep learning regression model, determine the corresponding hue correction value; Temporal pulse strategy parameters: combining the brightness requirements of high-priority areas with panel aging parameters A panel =0.07, using the PWM dimming parameter matching algorithm (target brightness and duty cycle are linearly positively correlated), the pulse width of PWM dimming is determined to be 76% and the frequency is 1.2kHz, so that the pixel emission duty cycle matches the brightness compensation requirements and the single emission time of a single pixel does not exceed the lifetime decay threshold. The color management engine encapsulates the aforementioned optimization parameters into a dedicated optimization parameter package for high-priority areas and sends it to the built-in display driver chip of the OLED flat panel screen. After receiving the parameters, the display driver module performs precise optimization on each pixel within the high-priority area according to its position coordinates: it corrects the pixel color attributes based on hue and saturation compensation parameters, adjusts the pixel luminous brightness according to the brightness compensation coefficient, and controls the pixel luminous timing sequence in combination with temporal pulse strategy parameters to complete the color optimization of the high-priority area.

[0030] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for optimizing the color display of an OLED screen based on a color management engine, characterized in that: The method includes the following steps: Step 1: Collect ambient light-related data by integrating sensors into the OLED screen to form an ambient light characteristic dataset, and extract constraint parameters from the device characteristic file to form a device capability constraint dataset; Step 2: Obtain the original data of the display screen, process it through a pre-trained lightweight semantic segmentation model and extract features, and integrate it to form a color and brightness dataset of the screen content; Step 3: Construct a deep learning regression model using the ambient light characteristics dataset, device capability constraint dataset, and image content chromaticity and brightness dataset as input. After training, output the color degradation index and determine the high priority area by combining the regional priority weight. Step 4: Set the optimization objectives and constraints, call the multi-objective constraint optimization algorithm to solve the compensation parameters of the high-priority region, and send them to the driver module to perform color optimization.

2. The OLED screen color optimization method based on a color management engine according to claim 1, characterized in that: In step 1, a spectral sensor, a brightness sensor, and an angle sensor are integrated into the OLED screen to collect the spectral distribution data S(λ) of ambient light and the ambient brightness value L, respectively. env 1. Ambient light incident angle θ; 2. Acquire ambient light reflection brightness L on the screen surface. ref This data is integrated to form an ambient light characteristics dataset. The color management engine calls the device characteristic file built into the OLED screen, reads the device constraint parameters, and forms a device capability constraint dataset, specifically including: the upper limit of panel peak brightness L. max Panel aging parameter A panel Real-time power consumption constraint P limit .

3. The OLED screen color optimization method based on a color management engine according to claim 2, characterized in that: In step 2, the color management engine obtains the raw image data of the current frame from the OLED screen display buffer and inputs it into the pre-trained lightweight semantic segmentation model; The model outputs a pixel-level region mask M through feature extraction and multi-scale feature fusion, thereby achieving image region classification and distinguishing between salient and non-salient regions. After noise removal processing of the semantic segmentation results, the output includes: region mask M, coordinates of each region, region type label, and area percentage of each region. Based on the semantic segmentation results, all pixels within each region are traversed, and the original pixel color data is converted into HSV color space data using color space conversion technology: h represents hue, s represents saturation, and v represents brightness. Statistical analysis methods are then used to aggregate and calculate the HSV data within the region to obtain region-level features, including the average brightness L of each region. mean , Brightness variance, Saturation mean mean Saturation variance; Based on a pixel-level traversal process, the coordinates and corresponding HSV values ​​of each pixel are recorded. The HSV data of all pixels within a single region are arranged by coordinates to form a distribution matrix within the region, yielding pixel-level features, specifically including: the pixel-level brightness distribution matrix L of each region. local (x,y), pixel-level chromaticity distribution matrix C local (x,y); where x and y are pixel coordinates; The semantic segmentation output is associated with the semantic features, and weights are set according to preset priority rules to obtain semantic association features, which include: type labels of each region and region priority weights. By integrating region-level features, pixel-level features, and semantic association features, a color saturation and brightness dataset of the image content is obtained.

4. The OLED screen color optimization method based on a color management engine according to claim 3, characterized in that: In step 3, a deep learning regression model is constructed using the ambient light characteristics dataset, the device capability constraint dataset, and the image content chromaticity and brightness dataset as inputs, and the color degradation quantitative index measured by a professional color analyzer as the sample label. The model outputs a quantitative index of color degradation. The training model learns the mapping relationship between inputs and outputs; The specific quantitative index of color degradation includes: saturation attenuation ΔS. loss Used to predict the degree of saturation loss in an image under strong light; contrast loss value ΔC loss : Used to predict the degree of contrast reduction caused by the combined effect of dark-field reflected light and bright-field peak limitation under strong light; the degree of reduction in detail resolution ΔD loss Used to predict the degree of blurring of details under strong light; Integrate the trained lightweight model into the color management engine; By combining the region priority weights obtained from semantic segmentation with the degradation prediction values ​​output by the model, the comprehensive degradation index I for each region is calculated: I = W p ×(W C ×ΔC loss +W S ×ΔS loss +W D ×ΔD loss ); where W p W represents the region priority weight; C W S W D These represent the weights for contrast loss, saturation decay, and detail blurring, respectively. Collect all semantic region I values, sort the regions in descending order to form a region optimization priority list; set a filtering threshold T. I When the I value of a certain semantic region is greater than T I When this occurs, the region is defined as a high-priority region.

5. The OLED screen color optimization method based on a color management engine according to claim 4, characterized in that: In step 4, the optimization goals are set as follows: the perceptible saturation of high-priority areas is increased to the preset standard, the image contrast is restored to the baseline level outside of strong light environments, and the detail recognition meets the preset clarity threshold. Set constraints: The actual luminance of all pixels shall not exceed the upper limit of the panel's peak luminance L. max The overall power consumption of the screen does not exceed the real-time power consumption constraint P. limit The cumulative values ​​of pixel emission duration and brightness do not exceed the lifetime decay threshold. The lifespan degradation threshold is based on panel aging parameter A. panel Dynamic adjustment; The color management engine calls a multi-objective constrained optimization algorithm, taking into account a list of region optimization priorities, color degradation prediction results, and a device capability constraint dataset, and solves for the core optimization parameters for high-priority regions. Regional brightness compensation coefficient: based on the contrast loss value ΔC of high-priority areas. loss Calculate the brightness compensation coefficient for each high-priority region; Hue and saturation compensation parameters: based on the saturation attenuation ΔS in high-priority regions. loss Calculate the saturation enhancement coefficient for this region; determine the corresponding hue correction value based on the color shift prediction results. Temporal pulse strategy parameters: combining the brightness requirements of high-priority areas with panel aging parameters A panel Determine the pulse width and frequency of PWM dimming to match the pixel emission duty cycle with the brightness compensation requirements and control the single emission duration of a single pixel to not exceed the lifespan decay threshold. The color management engine encapsulates the brightness compensation coefficient, hue and saturation compensation parameters, and temporal pulse strategy parameters of the high-priority areas obtained from multi-objective constraint optimization into a dedicated optimization parameter package for the high-priority areas, and then sends it to the display driver module of the OLED screen. After receiving the parameters, the display driver module performs color rendering and brightness adjustment on each pixel in the high-priority area according to the position coordinates of the high-priority area: it corrects the pixel color attributes according to the hue and saturation compensation parameters, adjusts the pixel luminous brightness according to the brightness compensation coefficient, and controls the pixel luminous timing in combination with the temporal pulse strategy parameters to complete the color optimization of the high-priority area.

6. An OLED screen color optimization system based on a color management engine, applied to the OLED screen color optimization method based on a color management engine as described in any one of claims 1-5, characterized in that: The system includes an ambient light acquisition module, an image analysis module, a degradation prediction module, and an optimization execution module; The ambient light acquisition module is used to collect ambient light-related data through sensors integrated into the OLED screen to form an ambient light characteristic dataset, and to extract constraint parameters from the device characteristic file to form a device capability constraint dataset. The image analysis module is used to acquire the original data of the display image, process it through a pre-trained lightweight semantic segmentation model and extract features, and integrate it to form an image content chroma and brightness dataset. The degradation prediction module is used to construct a deep learning regression model with ambient light characteristics dataset, device capability constraint dataset, and image content chromaticity and brightness dataset as input. After training, it outputs a color degradation index and determines high-priority areas by combining regional priority weights. The optimization execution module is used to set optimization objectives and constraints, call a multi-objective constraint optimization algorithm to solve for compensation parameters in high-priority regions, and send the results to the driving module to perform color optimization.

7. The OLED screen color optimization system based on a color management engine according to claim 6, characterized in that: The ambient light acquisition module includes an ambient light acquisition unit and a device constraint extraction unit; The ambient light acquisition unit is used to acquire ambient light-related data through sensors integrated into the OLED screen to form an ambient light characteristic dataset; the device constraint extraction unit is used to call the device characteristic file, extract device constraint parameters, and form a device capability constraint dataset.

8. The OLED screen color optimization system based on a color management engine according to claim 7, characterized in that: The image analysis module includes an image acquisition unit, a semantic segmentation unit, and a feature extraction unit; The image acquisition unit is used to acquire the original image data of the current frame from the OLED screen display buffer; the semantic segmentation unit is used to process the image through a pre-trained lightweight semantic segmentation model to classify the image regions and output region masks and type labels; the feature extraction unit is used to extract the chromaticity and brightness features of the image content based on the semantic segmentation results to form an image content chromaticity and brightness dataset.

9. The OLED screen color optimization system based on a color management engine according to claim 8, characterized in that: The degradation prediction module includes a model training unit, a degradation prediction unit, and a region priority unit; The model training unit is used to construct a deep learning regression model and train the model to learn the mapping relationship between the input and the color degradation index; the degradation prediction unit is used to use the trained model to predict the color degradation index, including saturation decay, contrast loss and decrease in detail recognition; the region priority unit is used to combine the region priority weight and the degradation prediction value to calculate the comprehensive degradation index and determine the high priority region.

10. The OLED screen color optimization system based on a color management engine according to claim 9, characterized in that: The optimization execution module includes an optimization target setting unit, an optimization solution unit, and a parameter distribution unit; The optimization target setting unit is used to set the target and constraints for color optimization; the optimization solution unit is used to call a multi-objective constraint optimization algorithm to solve for the compensation parameters of high-priority regions; the parameter distribution unit is used to encapsulate the optimization parameters and distribute them to the display driver module to perform color optimization.

Images