A method and system for evaluating skin color restoration perceptual color difference of a display device

Abstract

The application discloses a kind of display equipment skin color restoration perceived color difference evaluation method and system, including obtaining reference image and to be evaluated image;Obtain preset reference white point coordinate;Reference image and to be evaluated image are converted from RGB color space to YCbCr color space, and skin color region pixel is extracted according to skin color threshold value;Skin color region RGB value is converted into CIE XYZ value and CIELAB value;The difference between the brightness of two images skin color region, chroma difference, hue difference, average brightness and average hue;Based on the fitting relationship of visual color difference and the average brightness, average hue of image pair of subjective visual evaluation experimental data, determine brightness direction weight parameter, chroma direction weight parameter and hue direction weight parameter, construct skin color difference evaluation model;Color difference parameter and weight coefficient are input into model, to obtain evaluation perceived color difference value, and output skin color restoration quality evaluation result.The application can be used for display equipment skin color restoration evaluation, white balance effect evaluation and related image processing scene.

Description

Technical Field

[0001] This invention belongs to the fields of display quality evaluation, color image analysis and color science technology, and specifically relates to a method and system for evaluating the perceived color difference in skin tone reproduction of display devices. Background Technology

[0002] Color difference describes the difference between two colors, typically calculated using the Euclidean distance between two color coordinates in a uniform color space to represent the degree of difference between the two stimuli. In industrial production, digital imaging, and display quality evaluation, color consistency is often judged based on the magnitude of the color difference. The CIELAB color difference formula recommended by the International Commission on Illumination (CIE) is widely used due to its ease of calculation and broad applicability. However, related research indicates that the universal color difference formula still suffers from insufficient visual uniformity in certain color regions, leading to a discrepancy between the calculated color difference and human subjective perception.

[0003] To improve prediction accuracy, academia and industry have proposed and applied various color difference correction methods. For example, based on CIELAB, a series of formulas with equal weighting factors for lightness, chroma, and hue, such as CIE94 and CIEDE2000, have been introduced to improve the predictive performance of overall color difference. Among them, CIEDE2000 performs better in overall color difference evaluation and is widely used in digital imaging, display evaluation, and equipment calibration.

[0004] However, for portrait images, especially facial skin tones, skin tone is one of the most important color elements affecting the naturalness, realism, and viewing comfort of an image. Compared to the colors of general objects, observers are usually more sensitive to changes in facial skin tone, and skin tone deviations are more easily detected. Although existing universal color difference formulas have good comprehensive color difference prediction capabilities across the entire color gamut, they are insufficient to accurately reflect the skin tone reproduction effect in portrait images and its impact on the overall visual effect.

[0005] In summary, the existing technology lacks a perceptual color difference assessment method and system that is oriented towards skin color areas, can more accurately reflect the differences in skin color presentation between two images, and is more consistent with the subjective perception of the human eye, so as to meet the actual needs of skin color reproduction assessment, white balance effect assessment and related image processing scenarios of display devices. Summary of the Invention

[0006] The purpose of this invention is to address the problems described in the background art by proposing a method and system for evaluating the perceived color difference in skin tone reproduction on display devices. Regarding the need for color difference evaluation of facial skin tones in human images, the human visual system considers factors such as brightness, chroma, and hue when judging skin tone differences. This invention constructs a skin tone-oriented perceived color difference evaluation model to achieve a skin tone reproduction effect evaluation on display devices that better aligns with the subjective perception characteristics of the human eye.

[0007] The technical solution of the present invention is to provide a method for evaluating the perceived color difference in skin tone reproduction of a display device, comprising the following steps:

[0008] Step 1: Obtain the reference image and the image to be evaluated;

[0009] Step 2: Obtain the preset reference white point tristimulus values ​​XYZw;

[0010] Step 3: Convert the reference image and the image to be evaluated from the RGB color space to the YCbCr color space, extract the skin color region pixels in each image based on the preset skin color threshold, and generate the corresponding skin color region mask.

[0011] Step 4: Based on the skin color region mask, obtain the RGB values ​​of each pixel within the skin color region in both images;

[0012] Step 5: Based on the RGB and CIE XYZ feature model of the display device, convert the RGB values ​​of skin color into CIE XYZ values;

[0013] Step 6: Calculate the CIELAB value of each image based on the preset reference white point tristimulus value XYZw;

[0014] Step 7: Calculate the brightness difference between the reference image and the image to be evaluated. Color difference Tone difference Average brightness Average tone ;

[0015] Step 8, based on average brightness and average tone Determine the fitting factor for the lightness direction and hue direction fitting factor And further determine the brightness direction weight parameters. Chroma Direction Weighting Parameter and hue direction weight parameters ;

[0016] Step 9: Input the parameters into the constructed skin color difference assessment model and calculate the perceived color difference value. ;

[0017] Step 10: Based on the perceived color difference value, output the skin tone reproduction quality assessment result.

[0018] Furthermore, the specific form of the skin color difference assessment model is as follows:

[0019]

[0020] in, To evaluate the perceived color difference value, For power exponent correction parameters, These are the lightness direction weighting parameters, chroma direction weighting parameters, and hue direction weighting parameters, respectively.

[0021] Furthermore, the brightness direction weight parameter in step 8 Chroma Direction Weighting Parameter and hue direction weight parameters satisfy:

[0022]

[0023] in, , and These are the basic weight parameters for brightness, chroma, and hue, respectively. and These are the effective weighting factors for the brightness and hue directions, respectively.

[0024] Furthermore, the effective weighting factors satisfy:

[0025]

[0026] in, and These are the lightness direction fitting factor and the hue direction fitting factor, respectively. and To fit the weighting factors, and The method of least squares is used for optimization.

[0027] Furthermore, the brightness direction fitting factor and hue direction fitting factor The fitting amount in the lightness direction respectively Fitting amount of hue direction Normalization yields the following result:

[0028]

[0029] in, For subjective visual assessment of sample logs, For the first The visual color difference values ​​of each sample pair, where the value fitted in the lightness direction is... From average brightness The quadratic function is determined, and the hue direction fitting amount is... From average hue The linear function is determined, and the relationship between visual color difference and average brightness is established based on subjective visual evaluation experimental data. Average tone The fitting relationship.

[0030] Furthermore, the brightness direction fitting amount Fitting amount of hue direction They respectively satisfy:

[0031]

[0032] in, , , , and These are the function coefficients.

[0033] Furthermore, in step 3, the preset skin tone threshold is the skin tone threshold range in the YCbCr color space, and the skin tone threshold range is set according to the statistical distribution range of the preset skin tone sample library.

[0034] Moreover, in step 5, the feature model is a gain-bias-gamma model.

[0035] Furthermore, in step 6, the CIELAB value calculation uses the CIE 1976 LAB color space.

[0036] Furthermore, in step 9, the fitting parameters , , The skin color reproduction quality assessment results in step 10 are determined by optimization using the least squares method, including the perceived color difference value, quality grade, and whether it is qualified or unqualified.

[0037] This invention also provides a skin tone reproduction perception color difference evaluation system for display devices, including an image acquisition module, a reference white point acquisition module, a skin tone region extraction module, an RGB value extraction module, a color space conversion module, a color difference calculation module, a perception color difference evaluation module, and an evaluation output module; wherein, the image acquisition module is used to acquire a reference image and an image to be evaluated; the reference white point acquisition module is used to acquire preset reference white point tristimulus values. The skin color region extraction module converts the reference image and the image to be evaluated from the RGB color space to the YCbCr color space, and extracts skin color region pixels based on a preset skin color threshold to generate a corresponding skin color region mask; the RGB value extraction module obtains the RGB values ​​of each pixel within the skin color region in the two images; the color space conversion module converts the skin color RGB values ​​to CIE XYZ values ​​based on the RGB and CIE XYZ feature model of the display device, and converts them based on a preset reference white point tristimulus value. Calculate CIELAB values; the color difference calculation module is used to calculate the brightness difference between the reference image and the image to be evaluated. Color difference Tone difference Average brightness and average tone The perceptual color difference assessment module is used to input the parameters into a pre-built skin color difference assessment model to obtain the perceptual color difference value. The evaluation output module is used to output skin tone reproduction quality evaluation results based on the perceived color difference value.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] This invention establishes a dedicated perceptual color difference assessment method for facial skin tone. By extracting skin tone regions from both a reference image and the image to be evaluated, and combining a display device color feature model and a skin tone color difference assessment model, it quantifies the difference in skin tone presentation between the reference image and the image to be evaluated. Simultaneously, based on the comprehensive color difference parameters, it further introduces fitting factors and weight modulation mechanisms related to average brightness and average hue, enabling the assessment results to more accurately reflect the subjective perception of skin tone differences by the human eye. This improves the relevance, accuracy, and practicality of skin tone reproduction quality assessment for display devices. This invention can be applied to skin tone reproduction assessment, white balance effect assessment, and related image processing scenarios for display devices. Attached Figure Description

[0040] Figure 1 This is a flowchart of the method of the present invention.

[0041] Figure 2 This is a schematic diagram of the subjective visual evaluation experiment interface in an embodiment of the present invention.

[0042] Figure 3 This is a schematic diagram of the system modules of the present invention. Detailed Implementation

[0043] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0044] like Figure 1 As shown in the figure, this embodiment provides a method for evaluating the perceived color difference in skin tone reproduction of a display device, including the following steps.

[0045] 1) Obtain a reference image and an image to be evaluated. The reference image is used as a benchmark for skin color restoration evaluation, and the image to be evaluated is used as the target image to be evaluated.

[0046] In this embodiment, the reference image is a standard reference face image, and the image to be evaluated is a face image output by a display device, used to evaluate the deviation of the skin tone reproduction of the image output by the display device relative to the reference image.

[0047] In this embodiment, the reference image and the image to be evaluated are images of the same object acquired under different light source conditions, or images of the same image obtained after different white balance processing, used to evaluate the effect of different light source conditions or different white balance parameters on the skin color rendering effect.

[0048] 2) Obtain preset reference white point tristimulus values ;

[0049] In this embodiment, the preset reference white point tristimulus value can be preset according to the target white point of the display device.

[0050] 3) Convert the reference image and the image to be evaluated from the RGB color space to the YCbCr color space, extract the skin color region pixels in each image based on the preset skin color threshold, and generate the corresponding skin color region mask;

[0051] In this embodiment, the skin tone threshold is a preset skin tone threshold range in the YCbCr color space. This threshold range can be set based on the distribution characteristics of the color gamut range labeled as Asian skin tone in a certain skin database within the YCbCr color space.

[0052] 4) Based on the skin color region mask, obtain the RGB values ​​of each pixel within the skin color region in both images;

[0053] In this embodiment, the skin color region mask is used to limit the effective area for subsequent color parameter calculation. Color parameter conversion and difference analysis are performed only on the skin color region, thereby avoiding interference from non-skin color information such as background area, hair area, and clothing area on the subsequent evaluation results.

[0054] 5) Based on the RGB and CIE XYZ feature models of display devices, convert skin color RGB values ​​to CIE XYZ values;

[0055] In this embodiment, the characterization model is a gain-bias-gamma model, which maps the RGB values ​​of skin tone obtained in step 4 to CIE XYZ tristimulus values. This model is simple and easy to implement, and is commonly used for basic display calibration. The construction of the gain-bias-gamma (GOG) model relies on two assumptions: channel independence and constant chromaticity. The main calculation formulas are shown below:

[0056]

[0057] in,

[0058]

[0059] and,

[0060]

[0061] in, , , Indicates black spots on the monitor ( The corresponding tristimulus values ​​can be obtained by measuring a spectroradiometer. These represent the chromaticity values ​​corresponding to pure red, pure green, and pure blue blocks, respectively. , , These represent the sum of the tristimulus values ​​corresponding to the brightest pure color block. , , These represent the normalized luminance functions for the corresponding red, green, and blue channels, respectively, including... , , , , , Six parameters to be determined. During the model training phase, the parameters can be determined based on the pre-set RGB values ​​and the measured CIE XYZ values ​​(10° observer view). , , , , , Optimization is performed to determine the forward GOG model and the reverse GOG model for the corresponding device.

[0062] 6) Based on preset reference white point tristimulus values Calculate the CIELAB value for each image;

[0063] In this embodiment, the CIE 1976 LAB color space conversion formula is used, based on the preset reference white point tristimulus values ​​obtained in step 2. Convert the CIE XYZ values ​​to CIELAB values, and then obtain the corresponding chroma. With color tone The color space conversion method described above is a publicly available method, and will not be elaborated upon in this invention.

[0064] 7) Calculate the brightness difference between the reference image and the image to be evaluated. Color difference Tone difference Average brightness and average tone ;

[0065] In an embodiment, Used to represent the difference in brightness between corresponding skin tone areas in two images. Used to represent chroma differences Used to indicate tonal differences. Average brightness. and average tone It is then calculated from the color parameters of the corresponding skin color regions in the reference image and the image to be evaluated.

[0066] 8) Based on average brightness and average tone Determine the fitting factor for the lightness direction and hue direction fitting factor And further determine the brightness direction weight parameters. Chroma Direction Weighting Parameter and hue direction weight parameters ;

[0067] In the embodiment, visual color difference and average brightness are first established based on subjective visual evaluation experimental data. Average tone The fitting relationship. Among them, the fitting value in the brightness direction. From average brightness The quadratic function is determined, and the hue direction fitting amount is... From average hue The linear function is determined, specifically satisfying:

[0068]

[0069] in, , , , and The coefficients are obtained by fitting experimental data through subjective visual evaluation. Multiple sets of image sample pairs were collected, and the visual color difference of each sample pair was obtained. The average brightness and average hue of the corresponding image pairs were also obtained. Based on these experimental data, the quadratic function relationship between visual color difference and average brightness, and the linear function relationship between visual color difference and average hue were fitted respectively, and finally the values ​​of these 5 coefficients were obtained.

[0070] In this embodiment, the values ​​of each coefficient are as follows: =−0.339, =39.55, =−1152, =−0.0912, =6.92.

[0071] Furthermore, the brightness direction fitting factor and hue direction fitting factor Fitted by lightness direction respectively Fitting amount of hue direction Normalization yields the following result:

[0072]

[0073] in, For subjective visual assessment of sample logs, For the first Visual color difference values ​​for each sample pair.

[0074] Furthermore, the effective weighting factors satisfy:

[0075]

[0076] in, and To fit the weighting factors, and The method of least squares is used for optimization.

[0077] Based on this, the brightness direction weight parameter Chroma Direction Weighting Parameter and hue direction weight parameters satisfy:

[0078]

[0079] in, , and These are the basic weight parameters for brightness, chroma, and hue, respectively. and These are the effective weighting factors for the brightness and hue directions, respectively.

[0080] In the embodiment, the fitting parameters , , The value is determined by optimization using the least squares method.

[0081] 9) Input the parameters into the constructed skin color difference assessment model and calculate the perceived color difference value. ;

[0082] In this embodiment, the skin color difference assessment model specifically satisfies the following relationship:

[0083]

[0084] in, To evaluate the perceived color difference value, For power exponent correction parameters, , and These are the lightness direction weight parameter, chroma direction weight parameter, and hue direction weight parameter, respectively.

[0085] The above model can comprehensively reflect the impact of brightness difference, chroma difference, and hue difference on skin color perception differences, thereby obtaining skin color difference assessment results that are more in line with the subjective perception rules of the human eye.

[0086] 10) Output the skin tone reproduction quality assessment result based on the perceived color difference value;

[0087] In this embodiment, the skin color reproduction evaluation result includes at least an assessment of the perceived color difference value; optionally, a quality level or pass / fail judgment may also be output based on a preset evaluation threshold. The smaller the assessment of the perceived color difference value, the smaller the deviation of the skin color reproduction of the image to be evaluated relative to the reference image, and the higher the skin color reproduction quality; the larger the assessment of the perceived color difference value, the greater the deviation of the skin color reproduction of the image to be evaluated relative to the reference image, and the lower the skin color reproduction quality.

[0088] To further demonstrate the technical advantages of the method described in this invention in terms of the consistency of color difference assessment in display devices for reproducing human eye-perceived color differences, a subjective visual evaluation experiment was conducted. The experimental interface is shown below. Figure 2 As shown, the consistency between the visual color difference obtained from the subjective visual evaluation experiment and the perceived color difference value evaluated by the model output in step 9) is measured by calculating the standardized residual sum of squares index (STRESS index), and compared with the color difference evaluation method recommended by CIE.

[0089] A total of 20 observers were recruited (10 males and 10 females; average age: 25.2 years). All observers were students who had passed the Ishihara color blindness test and possessed normal color vision. The observers' ages ranged from 18 to 28 years old. After passing the test, they completed a questionnaire outside the laboratory containing information such as gender and age. Upon entering the laboratory, observers changed into gray lab coats to prevent glare and color interference, and were asked to turn off their mobile phones and other light-emitting devices.

[0090] The experimental results are shown in Table 1. The method of this invention was compared with CIELAB (color difference symbol is...). CIE94 (color difference symbol is) ) and CIEDE2000 (color difference symbol is The results show that the STRESS value corresponding to the method of the present invention is lower than that of the compared CIELAB, CIE94 and CIEDE2000 methods, indicating that the method of the present invention has good consistency with the subjective visual evaluation results.

[0091] Table 1. STRESS index of different color difference assessment models based on visual color difference dataset

[0092]

[0093] The results fully demonstrate that the color difference evaluation model for skin tone reproduction perception of display devices constructed in this invention has high accuracy and stability.

[0094] like Figure 3 As shown, the present invention also provides a color difference assessment system for skin tone reproduction perception in display devices, comprising the following modules:

[0095] The image acquisition module is used to acquire reference images and images to be evaluated.

[0096] The reference white point acquisition module is used to acquire preset reference white point tristimulus values ​​XYZw.

[0097] The skin color region extraction module is used to convert the reference image and the image to be evaluated from the RGB color space to the YCbCr color space, and extract skin color region pixels based on a preset skin color threshold to generate the corresponding skin color region mask.

[0098] The RGB value extraction module is used to obtain the RGB values ​​of each pixel in the skin color region of two images;

[0099] The color space conversion module is used to convert the RGB values ​​of each pixel in the skin color area to CIE XYZ values ​​based on the RGB and CIE XYZ feature GOG model of the display device, and convert the CIE XYZ values ​​to CIELAB values ​​based on the preset reference white point tristimulus value XYZw.

[0100] The color difference calculation module is used to calculate the brightness difference between the reference image and the image to be evaluated. Color difference Tone difference Average brightness and average tone ;

[0101] The perceived color difference assessment module is used to input the parameters into a pre-built skin color difference assessment model to obtain the perceived color difference value. ;

[0102] The evaluation output module is used to output the skin tone reproduction quality evaluation result based on the evaluated perceived color difference value.

[0103] The specific implementation process of each module is the same as the corresponding steps in the aforementioned method implementation, and will not be repeated here.

[0104] This invention is intended to cover all such substitutions, modifications, and variations falling within the broad scope of the claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for evaluating perceived color difference in skin tone reproduction in a display device, characterized in that, Includes the following steps: Step 1: Obtain the reference image and the image to be evaluated; Step 2: Obtain the preset reference white point tristimulus values ​​XYZw; Step 3: Convert the reference image and the image to be evaluated from the RGB color space to the YCbCr color space, extract the skin color region pixels in each image based on the preset skin color threshold, and generate the corresponding skin color region mask. Step 4: Based on the skin color region mask, obtain the RGB values ​​of each pixel within the skin color region in both images; Step 5: Based on the RGB and CIE XYZ feature model of the display device, convert the RGB values ​​of skin color into CIE XYZ values; Step 6: Calculate the CIELAB value of each image based on the preset reference white point tristimulus value XYZw; Step 7: Calculate the brightness difference between the reference image and the image to be evaluated. Color difference Tonal difference Average brightness Average tone ; Step 8, based on average brightness and average tone Determine the fitting factor for the lightness direction and hue direction fitting factor And further determine the brightness direction weight parameters. Chroma Direction Weighting Parameter and hue direction weight parameters ; Step 9: Input the parameters into the constructed skin color difference assessment model and calculate the perceived color difference value. ; Step 10: Based on the perceived color difference value, output the skin tone reproduction quality assessment result.

2. The method for evaluating color difference perception in skin tone reproduction of a display device according to claim 1, characterized in that: The specific form of the skin color difference assessment model in step 9 is as follows: ； in, To evaluate the perceived color difference value, For power exponent correction parameters, These are the lightness direction weight parameter, chroma direction weight parameter, and hue direction weight parameter, respectively.

3. The method for evaluating color difference in skin tone reproduction perception of a display device according to claim 1, characterized in that: The formula for calculating the weight in step 8 is as follows: ; in, , and These are the basic weight parameters for brightness, chroma, and hue, respectively. and These are the effective weighting factors for the brightness and hue directions, respectively.

4. The method for evaluating color difference perception in skin tone reproduction of a display device according to claim 3, characterized in that: Effective weighting factors satisfy: ； in, and These are the fitting factors for the lightness and hue directions, respectively. and To fit the weighting factors, and The least squares method is used for optimization.

5. The method for evaluating color difference perception in skin tone reproduction of a display device according to claim 3, characterized in that: Brightness direction fitting factor and hue direction fitting factor Fitted by lightness direction respectively Fitting amount of hue direction Normalization yields the following result: ； in, For subjective visual assessment of sample logs, For the first Visual color difference values ​​for each sample pair; lightness direction fit value From average brightness The quadratic function is determined, and the hue direction fitting amount is... From average hue The linear function is determined, and the relationship between visual color difference and average brightness is established based on subjective visual evaluation experimental data. Average tone The fitting relationship.

6. The method for evaluating color difference perception in skin tone reproduction of a display device according to claim 5, characterized in that: The brightness direction fitting amount Fitting amount of hue direction They respectively satisfy: ； in, , , , and These are the function coefficients.

7. The method for evaluating color difference perception in skin tone reproduction of a display device according to claim 2, characterized in that: In step 9, the fitting parameters of the skin color difference assessment model The least squares method is used for optimization.

8. The method for evaluating color difference perception in skin tone reproduction of a display device according to claim 1, characterized in that: The preset skin tone threshold in step 3 is the skin tone threshold range in the YCbCr color space, and the skin tone threshold range is set according to the statistical distribution range of the preset skin tone sample library. The feature model in step 5 is a gain-bias-gamma model; The CIELAB value calculation in step 6 uses the CIE 1976 LAB color space.

9. The method for evaluating color difference perception in skin tone reproduction of a display device according to claim 1, characterized in that: The skin tone reproduction quality assessment results in step 10 include the perceived color difference value, quality grade, and pass / fail determination.

10. A color difference assessment system for skin tone reproduction perception in a display device, used to implement the color difference assessment method for skin tone reproduction perception in a display device according to any one of claims 1-9, characterized in that, Includes the following modules: The image acquisition module is used to acquire reference images and images to be evaluated. The reference white point acquisition module is used to acquire preset reference white point tristimulus values ​​XYZw. The skin color region extraction module is used to convert the reference image and the image to be evaluated from the RGB color space to the YCbCr color space, and extract skin color region pixels based on a preset skin color threshold to generate the corresponding skin color region mask. The RGB value extraction module is used to obtain the RGB values ​​of each pixel in the skin color region of two images; The color space conversion module is used to convert the RGB values ​​of each pixel in the skin color area to CIE XYZ values ​​based on the RGB and CIE XYZ feature GOG model of the display device, and convert the CIE XYZ values ​​to CIELAB values ​​based on the preset reference white point tristimulus value XYZw. The color difference calculation module is used to calculate the brightness difference between the reference image and the image to be evaluated. Color difference Tonal difference Average brightness and average tone ; The perceived color difference assessment module is used to input the parameters into a pre-built skin color difference assessment model to obtain the perceived color difference value. ; The evaluation output module is used to output the skin tone reproduction quality evaluation result based on the evaluated perceived color difference value.

Images