A method and system for adjusting underwear design parameters based on image analysis

Abstract

The application discloses a kind of underwear design parameter adjustment method and system based on image analysis, it is related to image analysis technical field, including the following steps: constructing optical data under standard illumination condition under reference object in fitting site;Acquire the underwear image of fitting person in the fitting site, and simultaneously acquire the local reference image of the reference object;Identify the pixel value difference between the local reference image and the optical data;Based on the pixel value difference, the pixel gain and pixel offset required by underwear image are calculated;Based on the pixel gain and pixel offset, each pixel in the underwear image is corrected;After the correction, the underwear image is processed by picture analysis;Based on the underwear image feature, adjust underwear design parameter.The application can effectively correct the image deviation caused by illumination condition change, ensure the accuracy of image data, so as to improve the accuracy and reliability of underwear design parameter adjustment.

Description

Technical Field

[0001] This application relates to the field of image analysis technology, and more specifically, to a method and system for adjusting underwear design parameters based on image analysis. Background Technology

[0002] Traditional methods for adjusting lingerie design parameters often face challenges in long-term operation due to external environmental factors, such as changes in lighting conditions, which can interfere with the accuracy of image data. These subtle, cumulative changes can cause the system's output measurement results and adjustment suggestions to deviate from actual needs, thus affecting automation efficiency and user satisfaction. In the field of lingerie design and customization, to achieve accurate assessment of the wearer's body shape and lingerie fit, high-resolution images are used to capture human body shape and lingerie wearing status, and then automatically calculate and recommend optimal design parameters. During the initial deployment of the system, a dedicated fitting room was carefully designed and equipped with professional-grade lighting equipment. This equipment is designed to provide a highly standardized lighting environment, characterized by stable light color and uniform brightness distribution, ensuring that during image acquisition, whether it is the wearer's skin texture, the color details of the lingerie, or the subtle wrinkles on the body, they can be recorded by the camera sensor with high fidelity. This standardized lighting is the foundation for the system's accurate image analysis, ensuring the repeatability and consistency of image data in color and brightness, thus allowing subsequent image segmentation, feature extraction, and size measurement to be performed on a reliable benchmark.

[0003] However, all physical devices have an inherent lifespan and performance degradation. With the image analysis system used for extended periods and at high frequency in daily operation, the originally professional lighting fixtures in the fitting room, especially their internal light-emitting elements, inevitably experience natural decay and shift in the color and brightness uniformity of the emitted light due to continuous energy conversion and heat accumulation. This decay is not a sudden failure, but a slow and continuous accumulation process. For example, the phosphor in LED beads may gradually age due to high temperatures, causing the emitted light color to shift towards a specific wavelength, or its luminous flux to gradually decrease, resulting in a decrease in overall brightness. This slow change in lighting characteristics directly leads to subtle but persistent deviations between the overall color representation and local brightness distribution of the underwear and human body images acquired by the system at different time points and the standard images recorded during the system's initial calibration. These deviations may appear to the human eye as slight changes in tone or brightness, but for an image processing system that relies on precise pixel values ​​for analysis, they constitute significant inconsistencies in input data.

[0004] Of particular note is that while the reflectivity of lingerie fabrics to different colors of light is an inherent property, the color and texture information presented by these fabrics under attenuated light sources differs from the performance expected and learned by the system under standard light. For example, a piece of lingerie that appears "warm beige" under standard light may be captured by the image sensor as a "cool gray" beige under attenuated light sources with a bluish tint. Simultaneously, the subtle details of its fabric texture may become blurred or distorted due to uneven brightness. The core judgment rules used for image understanding within this image analysis system—such as the color value range used to distinguish skin and fabric when segmenting images, and the texture recognition information used to analyze fabric stretch to assess fit—are all rigorously trained and optimized based on initial standard lighting conditions. These rules were initially designed assuming that the color and brightness distribution of the input image conforms to specific statistical characteristics. When the system encounters non-standard color and texture representations caused by attenuated light sources, these pre-set judgment rules begin to misjudge. For example, a color threshold originally used to identify skin boundaries may misidentify parts of the underwear edge as skin under a color-biased light source, or misinterpret subtle shadows on the skin as wrinkles in the underwear. Similarly, a texture feature extractor used to assess fabric stretch may fail to accurately capture the deformation details of the fabric in images with uneven brightness, leading to decreased accuracy in underwear boundary recognition and reduced reliability in assessing the fit of underwear to the human body. Summary of the Invention

[0005] This application discloses a method and system for adjusting underwear design parameters based on image analysis, which can effectively correct image deviations caused by changes in lighting conditions, ensure the accuracy of image data, thereby improving the precision and reliability of underwear design parameter adjustment, and overcoming the problem of inaccurate image analysis caused by illumination attenuation in the prior art.

[0006] The technical solution of this application is as follows:

[0007] In a first aspect, this application discloses a method for adjusting underwear design parameters based on image analysis, specifically including the following steps:

[0008] Construct optical data of underground reference objects in the fitting room under standard lighting conditions;

[0009] Capture images of the underwear worn by the person trying on clothes in the fitting room, and simultaneously capture partial reference images of the reference object;

[0010] Identify the pixel value differences between the local reference image and the optical data;

[0011] The required pixel gain and pixel offset for the underwear image are calculated based on the pixel value difference;

[0012] Based on the pixel gain and pixel offset, each pixel in the underwear image is corrected to obtain the corrected underwear image;

[0013] The corrected underwear image was subjected to image analysis processing to obtain underwear image features;

[0014] Adjust underwear design parameters based on the characteristics of the underwear image.

[0015] Furthermore, identifying pixel value differences between the partial reference image and optical data includes:

[0016] Identify edge markers on the reference object and extract the reference image region on the local reference image based on the edge markers;

[0017] Based on this reference image area, the average pixel value of each standard color block and grayscale gradient area is obtained;

[0018] Calculate the pixel value difference between the average pixel value and the optical data.

[0019] Based on the above, the pixel gain and pixel offset required for the lingerie image are calculated based on the pixel value difference, including:

[0020] Get the current pixel value of each color channel in the lingerie image;

[0021] Configuring pixel gain and pixel offset for each color channel based on pixel value differences and current pixel values.

[0022] Furthermore, each pixel in the lingerie image is corrected based on pixel gain and pixel offset, resulting in the corrected lingerie image, which includes:

[0023] Identify the raw RGB value of each pixel in an underwear image;

[0024] The original RGB value of each pixel is adjusted based on pixel gain and pixel offset to obtain the adjusted RGB value;

[0025] Determine whether the adjusted RGB values ​​are within the valid range of color values;

[0026] If the adjusted RGB value is determined to be within the valid range of color values, then the adjusted RGB value will be used as the corrected RGB value.

[0027] If it is determined that the adjusted RGB value is not within the valid range of color values, the original RGB value corresponding to the adjusted RGB value is corrected to the threshold node value based on the proximity relationship between the adjusted RGB value and the threshold node value. The threshold node value is 0 or 255.

[0028] Furthermore, image analysis processing is performed on the corrected underwear image to obtain the following underwear image features:

[0029] Based on image segmentation technology, the skin area and underwear area of ​​the person trying on clothes are segmented to obtain the segmented underwear image;

[0030] Extract underwear image features from the segmented underwear image.

[0031] Furthermore, based on image segmentation technology, the skin area and underwear area of ​​the person trying on clothes are segmented to obtain the segmented underwear image, including:

[0032] Preprocessing of fabric optical properties is performed on the pixel data in the corrected underwear image;

[0033] The image segmentation parameters are adjusted based on the preprocessing results of the fabric's optical properties to obtain the adjusted image segmentation parameters.

[0034] Based on the preprocessing results of the fabric's optical properties and the adjusted image segmentation parameters, the skin area and underwear area of ​​the person trying on the clothes are segmented to obtain the segmented underwear image.

[0035] Furthermore, the preprocessing of the pixel data in the corrected underwear image to reflect the fabric's optical properties includes:

[0036] Estimate the translucency of underwear fabric;

[0037] The contribution of the skin color and texture beneath the underwear to the apparent color of the underwear fabric is calculated based on the degree of transparency.

[0038] By subtracting the contribution ratio from the apparent color of the underwear fabric, the pure inherent optical properties of the underwear fabric can be inverted.

[0039] The pixel data of the underwear area is adjusted based on the inherent optical properties of the pure material.

[0040] Furthermore, estimating the translucency of lingerie fabric includes:

[0041] Local skin features are obtained by sampling the skin within the area covered by the underwear.

[0042] Adjust the semi-transparency estimation parameters based on local skin characteristics and the material properties of the underwear fabric;

[0043] Estimate the local semi-transparency of the underwear fabric based on the adjusted semi-transparency estimation parameters.

[0044] Furthermore, the contribution of skin color and texture beneath the underwear to the apparent color of the underwear fabric is calculated based on translucency, including:

[0045] Obtain transmission spectral data of underwear fabric at different incident angles;

[0046] Obtain the translucency response curves of underwear fabric at different wavelengths;

[0047] The directional translucency of the underwear fabric is calculated based on the local geometric orientation and incident angle of light, combined with transmission spectral data.

[0048] The wavelength-dependent translucency of the underwear fabric is calculated based on the local color information of the underwear fabric and the translucency response curve.

[0049] The local comprehensive translucency of the underwear fabric is obtained based on directional translucency and wavelength-dependent translucency.

[0050] Based on the local integrated semi-transparency and the color and texture information of the skin beneath the underwear, the contribution ratio of the skin color and texture beneath the underwear to the apparent color of the underwear fabric is calculated.

[0051] Secondly, this application also discloses an image analysis-based system for adjusting underwear design parameters, the system comprising:

[0052] The reference object setting module is used to construct optical data of underground reference objects in the fitting room under standard lighting conditions;

[0053] The image acquisition module is used to acquire images of the underwear worn by the person trying on clothes in the fitting area, and at the same time acquire partial reference images of the reference object;

[0054] The illumination difference calculation module is used to identify the pixel value difference between the local reference image and the optical data;

[0055] The pixel compensation calculation module is used to calculate the required pixel gain and pixel offset for the lingerie image based on the pixel value difference;

[0056] The pixel correction module is used to correct each pixel in the underwear image based on the pixel gain and pixel offset to obtain the corrected underwear image;

[0057] The image processing module is used to perform image analysis and processing on the corrected underwear image to obtain underwear image features;

[0058] The parameter adjustment module is used to adjust the underwear design parameters based on the features of the underwear image.

[0059] This application discloses an image analysis-based method for adjusting underwear design parameters. It constructs optical data of an underground reference object in a fitting room under standard lighting conditions, and acquires images of the underwear worn by the person trying on the garment and partial reference images of the reference object. The method then identifies pixel value differences between the partial reference images and the optical data. Based on these pixel value differences, it calculates the required pixel gain and pixel offset for the underwear image and corrects each pixel in the underwear image to obtain a corrected underwear image. Finally, it performs image analysis processing on the corrected underwear image to obtain underwear image features, and adjusts the underwear design parameters based on these features. This method effectively solves the problem in existing technologies where changes in lighting conditions in the fitting room lead to a decrease in image data accuracy, thus affecting the accuracy of underwear design parameter adjustments. By introducing a reference object for real-time lighting calibration, this application can accurately compensate for color and brightness deviations caused by light source attenuation or offset, ensuring that image data remains consistent with standard conditions under any lighting conditions. This allows subsequent image segmentation, feature extraction, and size measurement to be performed on a reliable benchmark, significantly improving the accuracy of underwear boundary recognition and the reliability of underwear-body fit assessment, ultimately enhancing the accuracy of underwear design parameter adjustments and user satisfaction. Attached Figure Description

[0060] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0061] Figure 1 This is a flowchart of the method for adjusting underwear design parameters based on image analysis in an embodiment of the present invention;

[0062] Figure 2 This is a flowchart of a method in this invention to correct each pixel in an underwear image based on pixel gain and pixel offset to obtain a corrected underwear image.

[0063] Figure 3 This is a schematic diagram of the structure of the image analysis-based underwear design parameter adjustment system in an embodiment of the present invention. Detailed Implementation

[0064] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0065] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0066] Specifically, Figure 1 The flowchart of the image analysis-based underwear design parameter adjustment method according to an embodiment of the present invention is shown, including the following steps:

[0067] S101. Construct optical data of the underground reference object in the fitting room under standard lighting conditions;

[0068] The optical data here refers to the digital representation of the visual characteristics of a reference object, such as color, brightness, and texture, under standard, ideal lighting conditions. This data serves as a benchmark for subsequent comparison with actually acquired images to quantify changes in the lighting environment.

[0069] The implementation environment of this application is typically a fitting room equipped with image acquisition equipment (such as a high-resolution camera), computing processing units (such as a computer or a dedicated image processor), and display or output devices. A reference object is placed within the fitting room, and its position and optical characteristics are accurately recorded during system initialization.

[0070] First, constructing optical data of reference objects in the fitting room under standard lighting conditions is fundamental to the entire method. This step involves placing one or more reference objects with known optical properties within the fitting room, such as standard color charts or grayscale gradient boards, and scanning or photographing them under standard, ideal lighting conditions using a high-precision spectrometer or a calibrated camera. This yields detailed data on pixel values, brightness distribution, color saturation, and other parameters across different color channels. This data is stored as a baseline for subsequent lighting correction. For example, a calibration board containing red, green, and blue primary color blocks and a series of grayscale levels can be used as a reference. Under standard lighting, the average RGB pixel values ​​for each color block and grayscale region are recorded.

[0071] S102. Acquire an image of the underwear worn by the person trying on clothes in the fitting area, and simultaneously acquire a partial reference image of the reference object;

[0072] Secondly, images of the underwear worn by the person trying on the garment are captured in the fitting area, along with partial reference images of the reference objects. While the person is trying on the underwear, an image acquisition device (such as a digital camera) captures an image of the underwear. Simultaneously, to monitor lighting conditions in real time, the image acquisition device also captures partial reference images of the reference objects placed in the fitting area. This ensures that the underwear image and the reference object image are acquired under the same lighting conditions, providing synchronized data for subsequent lighting correction. For example, while capturing a full-body image of the person trying on the garment, a small, pre-defined reference object area is also included in the camera's field of view; the image of this area serves as the partial reference image.

[0073] S103. Identify the pixel value difference between the partial reference image and the optical data;

[0074] The pixel value difference here refers to the numerical deviation between the pixel value of a reference object in a local reference image and the preset optical data. This difference directly reflects the degree of deviation between the current lighting conditions and the standard lighting conditions.

[0075] Next, the pixel value differences between the partial reference image and the optical data are identified. This step aims to quantify the deviation between the current lighting conditions and standard lighting conditions. The pixel value differences can be calculated by comparing the pixel values ​​of the reference object in the real-time acquired partial reference image with pre-constructed standard optical data. This difference can be the average pixel value difference of the color channels, or it can be a difference in brightness or contrast. For example, the pixel value differences of the R, G, and B channels are obtained by comparing the average RGB values ​​of the standard color patches in the partial reference image with the average RGB values ​​of the corresponding color patches in the standard optical data channel by channel.

[0076] S104. Calculate the pixel gain and pixel offset required for the underwear image based on the pixel value difference;

[0077] Pixel gain and pixel offset are two key parameters used to correct image pixel values. Pixel gain is primarily used to adjust the overall brightness or contrast of an image, while pixel offset is used to adjust the overall tone or color balance. Through the combined effect of these two parameters, image distortion caused by changes in lighting can be effectively compensated.

[0078] Then, based on the pixel value differences, the required pixel gain and pixel offset for the lingerie image are calculated. According to the pixel value differences identified in the previous step, the system calculates the pixel gain and pixel offset needed to correct the lingerie image. Pixel gain is mainly used to adjust the overall brightness or contrast of the image to compensate for changes in light intensity; pixel offset is used to adjust the overall hue or color balance of the image to compensate for changes in light color. The calculation of these parameters aims to make the corrected lingerie image visually and numerically closer to an image under standard lighting conditions. For example, if the red channel pixel values ​​of a locally referenced image are generally low, a positive red channel pixel gain is calculated to increase the brightness of the red component in the lingerie image.

[0079] S105. Based on the pixel gain and pixel offset, each pixel in the underwear image is corrected to obtain the corrected underwear image;

[0080] Subsequently, each pixel in the lingerie image is corrected based on the pixel gain and pixel offset to obtain the corrected lingerie image. Using the pixel gain and pixel offset calculated in the previous step, the original RGB value of each pixel in the lingerie image is adjusted. The correction process typically involves linear or non-linear transformations of the pixel values ​​for each color channel to compensate for the effects of lighting variations. After correction, the color and brightness distribution of the lingerie image will more closely resemble the effect of shooting under standard lighting conditions, thus providing a more accurate data basis for subsequent image analysis. For example, for each pixel in the lingerie image, its original RGB value will be adjusted according to the calculated gain and offset, for example: R_corrected = R_original Gain_R + Offset_R.

[0081] S106. Perform image analysis processing on the corrected underwear image to obtain underwear image features;

[0082] The lingerie image features here refer to the quantitative information extracted from the corrected lingerie image that describes the lingerie's shape, color, texture, fit, and other aspects. These features form the basis for adjusting lingerie design parameters.

[0083] Furthermore, the corrected underwear image is subjected to image analysis processing to obtain underwear image features. After obtaining the corrected underwear image, various image processing techniques can be used to analyze it to extract features related to underwear design. These features may include the underwear's color, texture, shape, size, fit to the body, and wrinkles. For example, edge detection algorithms can be used to identify the underwear's outline, color histograms can be used to analyze the underwear's color distribution, or texture analysis algorithms can be used to assess the fabric's stretchability.

[0084] S107. Adjust the underwear design parameters based on the underwear image features.

[0085] Finally, the lingerie design parameters are adjusted based on the lingerie image features. Based on the lingerie image features extracted in the previous step, the system can automatically or assist the designer in adjusting the lingerie's design parameters. These parameters may include cup size, shoulder strap length, band size, fabric selection, and style details. In this way, it ensures that the lingerie design better suits the wearer's body shape and wearing needs, improving the lingerie's comfort and aesthetics. For example, if the image features indicate that the bra cups have gaps, the system may suggest adjusting the cup size or shape parameters.

[0086] The core of the image analysis-based method for adjusting underwear design parameters proposed in this application lies in effectively addressing the impact of changes in lighting conditions in the fitting room on the accuracy of image analysis through a series of refined image processing steps. By introducing reference objects and constructing optical data and acquiring local reference images, this application can effectively identify and quantify changes in lighting conditions in the fitting room, thereby performing precise pixel correction on the underwear images, ensuring the accuracy of subsequent image analysis, and ultimately achieving reliable adjustment of underwear design parameters. This overcomes the image data distortion problem caused by light attenuation in existing technologies.

[0087] The image analysis-based method for adjusting lingerie design parameters proposed in this application effectively identifies and quantifies changes in lighting conditions in fitting rooms by introducing a reference object and constructing its optical data and acquiring local reference images. Specifically, by comparing the pixel value differences between the local reference image and standard optical data, the required pixel gain and pixel offset of the lingerie image can be accurately calculated. Subsequently, each pixel in the lingerie image is corrected based on these parameters, resulting in a corrected image that is visually and numerically closer to the standard lighting conditions. This correction mechanism ensures the accuracy of subsequent image analysis and processing, enabling the lingerie image features extracted from the corrected image to truly reflect the wearing status and design requirements of the lingerie. Ultimately, based on these accurate lingerie image features, the system can reliably adjust lingerie design parameters, thereby overcoming the image data distortion problem caused by light attenuation in existing technologies and significantly improving the automation efficiency and user satisfaction of lingerie design parameter adjustment.

[0088] Compared with existing technologies, the core innovation of this application lies in the introduction of a reference object and its optical data, and the use of these for real-time monitoring of lighting conditions and image correction. Traditional methods often assume constant lighting conditions or only use simple white balance adjustments to cope with changes in lighting, but these methods are difficult to accurately compensate for the complex color shifts and brightness unevenness caused by light source attenuation over long-term operation. This application constructs optical data of an offline reference object in the fitting room under standard lighting conditions and acquires local reference images of the reference object in real time, enabling precise identification of pixel value differences between the current lighting conditions and standard conditions. Based on this difference, the system can calculate the pixel gain and pixel offset required for the underwear image and perform fine-grained correction on each pixel in the underwear image. This correction mechanism ensures the accuracy of subsequent image analysis and processing, allowing the underwear image features extracted from the corrected image to truly reflect the wearing status and design requirements of the underwear. For example, in a yellowish lighting environment, traditional methods may cause the underwear image to appear yellowish overall, leading the system to misjudge the underwear color or skin area. However, this application, through reference object correction, can restore the image to the standard color, thereby avoiding misjudgment. Ultimately, based on these accurate lingerie image features, the system can reliably adjust lingerie design parameters, significantly improving the automation efficiency of lingerie design parameter adjustment and user satisfaction, and providing more reliable technical support for personalized lingerie customization.

[0089] It should be noted that the step of identifying the pixel value difference between the local reference image and the optical data includes: identifying edge marker points on the reference object, and extracting the reference image region on the local reference image based on the edge marker points; obtaining the average pixel value of each standard color block and grayscale gradient region based on the reference image region; and calculating the pixel value difference between the average pixel value and the optical data.

[0090] Specifically, when identifying the pixel value differences between the aforementioned local reference image and the aforementioned optical data, edge markers on the reference object can first be identified. These edge markers can be specific geometric shapes or color regions preset on the reference object, serving to accurately locate the reference object in the image. Based on these edge markers, a reference image region containing standard color blocks and grayscale gradient areas can be extracted from the local reference image. This reference image region is a crucial part for subsequent illumination correction, and its accurate extraction is essential for subsequent pixel value analysis.

[0091] Furthermore, after obtaining the reference image area, each standard color patch and grayscale gradient region within that area can be analyzed to obtain its average pixel value. Standard color patches typically have known theoretical pixel values ​​under standard lighting conditions, while grayscale gradient regions provide pixel value references for different brightness levels. By calculating the average pixel values ​​of these regions, the actual optical response of the reference object under the current lighting conditions can be obtained, thus providing fundamental data for the quantification of lighting deviations.

[0092] The obtained average pixel value is then compared with the optical data of a pre-constructed underground reference object in the fitting room under standard lighting conditions to calculate the pixel value difference between the two. This pixel value difference accurately reflects the deviation between the current lighting conditions of the fitting room and the standard lighting conditions, providing a quantitative basis for subsequent pixel gain and pixel offset calculations.

[0093] This application's solution first identifies edge markers on a reference object, ensuring accurate extraction of the reference image area and avoiding errors introduced by inaccurate reference object positioning. By calculating the average pixel values ​​of standard color blocks and grayscale gradient areas, the impact of current lighting conditions on image pixel values ​​can be comprehensively and meticulously captured. Therefore, comparing these actual measurements with standard optical data allows for precise quantification of lighting deviations, providing reliable foundational data for subsequent pixel gain and pixel offset calculations, thereby ensuring the accuracy of lingerie image correction.

[0094] In this application, the calculation of the required pixel gain and pixel offset for the underwear image based on the pixel value difference includes: obtaining the current pixel value of each color channel in the underwear image; and configuring the pixel gain and pixel offset for each color channel based on the pixel value difference and the current pixel value.

[0095] Specifically, obtaining the current pixel value of each color channel in the lingerie image means that before calculating pixel gain and pixel offset, the system analyzes each pixel in the lingerie image to obtain its specific values ​​in each color channel, such as red (R), green (G), and blue (B). These current pixel values ​​reflect the original color information and brightness distribution of the image at the time of acquisition. The purpose is to provide real-time, targeted benchmark data for subsequent gain and offset configuration.

[0096] The configuration of pixel gain and pixel offset for each color channel based on pixel value differences and current pixel values ​​can be understood as follows: after obtaining information on lighting differences in the fitting room (i.e., pixel value differences) and the color channel pixel values ​​of the underwear image itself, the system will comprehensively consider these two aspects to determine the gain and offset to be applied to each color channel. For example, if the current pixel value of a certain color channel is already too high, even if the pixel value difference indicates that increased brightness is needed, a relatively small gain or even a negative offset may be applied to avoid overexposure. Conversely, if the current pixel value is too low, a larger gain or positive offset may be needed. This configuration method aims to achieve more refined correction that better meets the actual image requirements, ensuring that the corrected image has natural and accurate colors. In practical applications, linear or non-linear mapping functions can be used, taking pixel value differences and current pixel values ​​as input and outputting the corresponding pixel gain and pixel offset, for example, through lookup tables, mathematical models, or machine learning algorithms for calculation and configuration.

[0097] The proposed solution first obtains the current pixel value of each color channel in the lingerie image, allowing subsequent pixel gain and pixel offset configurations to fully consider the image's color and brightness distribution. Because the calculation of gain and offset considers not only the pixel value difference between the reference object and standard lighting conditions but also the real-time pixel state of the lingerie image itself, the configured pixel gain and pixel offset are more targeted and accurate. This comprehensive calculation method avoids over-correction or under-correction problems that may arise from relying solely on pixel value differences, ensuring that the correction parameters can better adapt to the specific conditions of different images, thus providing more accurate basic data for subsequent image correction steps.

[0098] In some preferred embodiments, a specific example is given below. Suppose that in a fitting room, due to warm ambient lighting, the captured lingerie image has an overall yellowish tint. By identifying the pixel value difference between a reference object and standard optical data, the system initially determines that the blue channel needs enhancement, while the red and green channels need appropriate suppression. However, if the lingerie image itself already has some areas with high pixel values ​​in the blue channel, directly applying a uniform gain might cause overexposure in these areas. The solution in this application first obtains the current pixel value of each color channel in the lingerie image. For example, for the blue channel, the system finds that the current pixel value in some areas is close to saturation. At this point, when configuring pixel gain and pixel offset based on pixel value differences and current pixel values, the system will appropriately adjust or limit the gain of the blue channel according to the actual situation of these high pixel value areas, and even apply a small negative offset in some cases to avoid overexposure. Simultaneously, a larger gain will be applied to areas with lower pixel values ​​in the blue channel. In this way, pixel gain and pixel offset can be dynamically adjusted according to the current pixel value of each color channel, ensuring that the corrected lingerie image not only eliminates ambient lighting deviations, but also has a more uniform and natural color distribution, avoiding color distortion in local areas.

[0099] In response, Figure 2 This invention illustrates a method for correcting each pixel in an underwear image based on pixel gain and pixel offset to obtain a corrected underwear image, as shown in the embodiment of the invention. The method specifically includes:

[0100] S201. Identify the original RGB value of each pixel in the underwear image;

[0101] S202. Adjust the original RGB value of each pixel based on the pixel gain and pixel offset to obtain the adjusted RGB value;

[0102] S203. Determine whether the adjusted RGB value is within the valid range of color values;

[0103] S204. If it is determined that the adjusted RGB value is within the valid range of color values, then the adjusted RGB value is used as the corrected RGB value.

[0104] S205. If it is determined that the adjusted RGB value is not within the valid range of color values, then the original RGB value corresponding to the adjusted RGB value is corrected to the threshold node value according to the proximity relationship between the adjusted RGB value and the threshold node value.

[0105] It should be noted that the threshold node value is 0 or 255.

[0106] Specifically, identifying the raw RGB value of each pixel in an underwear image refers to obtaining the initial values ​​of each pixel in the red (R), green (G), and blue (B) color channels. These raw RGB values ​​are typically in the range of 0 to 255. The raw RGB value of each pixel is then adjusted based on pixel gain and pixel offset to obtain the adjusted RGB value. This can be understood as applying pre-calculated gain and offset to the raw pixel value of each color channel to compensate for differences in lighting. For example, the adjusted RGB value can be expressed by the formula: "Adjusted value = Original value". The calculation uses "pixel gain + pixel offset". In practical applications, determining whether the adjusted RGB value is within the valid range of color values ​​means checking whether the adjusted value of each color channel is between 0 and 255. If the adjusted RGB value is within the range of 0 to 255, it is considered valid and directly used as the corrected RGB value. Further, if the adjusted RGB value is determined to be outside the valid range of color values, the original RGB value corresponding to the adjusted RGB value is corrected to the threshold node value based on the proximity relationship between the adjusted RGB value and the threshold node value. The threshold node value is usually 0 or 255. Specifically, if the adjusted value is less than 0, it is corrected to 0; if the adjusted value is greater than 255, it is corrected to 255. This processing method aims to prevent color overflow or undersaturation, ensuring that the corrected image color information remains within a displayable and processable valid range.

[0107] This application's solution effectively solves the pixel value overflow or undersaturation problems that may occur during illumination compensation by introducing a judgment and correction mechanism for the effective range of adjusted pixel values. It is precisely because the adjusted RGB values ​​of each color channel are rigorously checked for range and clamped based on their proximity to the threshold node value of 0 or 255 that the corrected underwear image can maintain accurate color information and visual quality. This processing method avoids color distortion, loss of detail, or image artifacts caused by simply truncating or ignoring values ​​outside the range, thus providing high-quality input data for subsequent image segmentation and feature extraction.

[0108] In some preferred embodiments, a specific example is given below. Assume that the original RGB value of a pixel in an underwear image is (100, 120, 150). After calculating the illumination difference, the pixel gain is 1.2 and the pixel offset is 30.

[0109] First, the original RGB values ​​of the pixel are identified as R=100, G=120, B=150.

[0110] Secondly, adjustments are made to each color channel based on pixel gain and pixel offset:

[0111] Adjusted R value = 100 1.2 + 30 = 120 + 30 = 150

[0112] Adjusted G value = 120 1.2 + 30 = 144 + 30 = 174

[0113] The adjusted B value = 150 1.2 + 30 = 180 + 30 = 210

[0114] At this point, the adjusted RGB values ​​are (150, 174, 210).

[0115] Next, determine whether the adjusted RGB values ​​are within the valid range of color values ​​(0-255). In this example, 150, 174, and 210 are all within the range of 0 to 255.

[0116] Therefore, the adjusted RGB values ​​(150, 174, 210) are directly used as the corrected RGB values.

[0117] As another specific implementation, assume that the original RGB value of another pixel is (200, 50, 10). The pixel gain is 1.5 and the pixel offset is 50.

[0118] Adjusted R value = 200 1.5 + 50 = 300 + 50 = 350

[0119] Adjusted G value = 50 1.5 + 50 = 75 + 50 = 125

[0120] The adjusted value of B = 10 1.5 + 50 = 15 + 50 = 65

[0121] At this point, the adjusted RGB values ​​are (350, 125, 65).

[0122] Determine if the adjusted RGB values ​​are within the valid range for color values. The R value of 350 exceeds the upper limit of 255.

[0123] Correction is performed based on the proximity relationship between the adjusted RGB values ​​and the threshold node values:

[0124] The R value of 350 is greater than 255, so it is corrected to the threshold node value of 255.

[0125] The G value of 125 remains constant within the range of 0-255.

[0126] The B value of 65 remains constant within the range of 0-255.

[0127] Ultimately, the corrected RGB values ​​are (255, 125, 65).

[0128] In this way, even when pixel values ​​are adjusted significantly, the final color data is ensured to be valid and usable, avoiding abnormal colors or data errors in the image.

[0129] It should be noted that the image analysis processing of the corrected underwear image in this application to obtain underwear image features includes: segmenting the skin area and underwear area of ​​the person trying on the clothes based on image segmentation technology to obtain segmented underwear images; and extracting underwear image features from the segmented underwear images.

[0130] Specifically, image segmentation technology refers to dividing and labeling different regions in an image using specific algorithms and models, achieving separation at the pixel level. Its purpose is to distinguish the target object (such as underwear) from the background (such as skin) in an image. The skin region of the person trying on underwear refers to the part of their body not covered by underwear, while the underwear region refers to the part of the body actually covered by underwear and itself. Segmentation involves using image segmentation technology to distinguish the set of pixels belonging to the skin from the set of pixels belonging to the underwear in the corrected underwear image. This results in a segmented underwear image, which is an image containing only the pixel data of the underwear region after segmentation, or at least an image where the underwear region is clearly identified. After obtaining the segmented underwear image, further underwear image features can be extracted. Specifically, this step involves extracting visual features related to underwear design, such as texture, color, shape, size, wrinkles, and fit, for the underwear region. For example, the edge contour, color distribution, fabric texture, and degree of fit to the body can be extracted.

[0131] This application's solution introduces image segmentation technology to first accurately distinguish between the wearer's skin area and the underwear area in the corrected underwear image. This distinction allows subsequent image analysis to focus on the underwear itself, avoiding interference from non-target areas such as the skin. By extracting features from the segmented underwear image, it can be ensured that the obtained underwear image features purely and accurately reflect the underwear's visual attributes and wearing effect, thus providing more reliable input data for subsequent adjustments to underwear design parameters.

[0132] The above technical solution effectively eliminates the interference of the wearer's skin area on the extraction of underwear image features, significantly improving the accuracy and relevance of underwear image features. Therefore, the obtained underwear image features can more realistically and accurately reflect the wearing effect and design details of the underwear, providing a high-quality data foundation for subsequent adjustments to underwear design parameters, thereby improving the accuracy and effectiveness of these adjustments.

[0133] In some preferred embodiments, when performing image analysis on the corrected lingerie images, deep learning models, such as convolutional neural network architectures based on U-Net or Mask R-CNN, can be used for semantic segmentation of the images. This model, trained on a large amount of labeled data, can accurately identify and distinguish between skin and lingerie regions in the image. After obtaining the segmented lingerie images, computer vision algorithms can be further utilized, such as edge detection algorithms (e.g., the Canny operator) to extract the lingerie's outline, color histogram analysis to analyze the lingerie's color distribution, texture analysis algorithms (e.g., Gabor filters) to extract fabric texture features, or geometric dimensional information by measuring the pixel dimensions of the lingerie in specific areas. These extracted features are then used to guide adjustments to lingerie design parameters, such as adjusting cup shape, shoulder strap width, underbust size, or fabric selection, to better suit the wearer's body shape and wearing needs.

[0134] To address this, this application further proposes the following steps for segmenting the skin region and underwear region of the person trying on clothes based on the aforementioned image segmentation technology to obtain a segmented underwear image: performing fabric optical property preprocessing on the pixel data in the corrected underwear image; adjusting the image segmentation parameters according to the fabric optical property preprocessing result to obtain the adjusted image segmentation parameters; and segmenting the skin region and underwear region of the person trying on clothes according to the fabric optical property preprocessing result and the adjusted image segmentation parameters to obtain a segmented underwear image.

[0135] Specifically, fabric optical property preprocessing refers to analyzing and adjusting the pixel data related to the fabric in the corrected underwear image before image segmentation. Its purpose is to eliminate or reduce the interference of the fabric's own optical properties (such as translucency, gloss, and color saturation) on image segmentation, enabling the image segmentation algorithm to more accurately identify the actual contours and areas of the underwear. For example, for semi-transparent fabrics, preprocessing can estimate its translucency and calculate the contribution ratio of the skin color and texture beneath the underwear to the apparent color of the underwear fabric based on this translucency. This allows the pure inherent optical properties of the underwear fabric to be inferred from the apparent color, and the pixel data of the underwear area can be adjusted accordingly.

[0136] Adjusting image segmentation parameters can be understood as dynamically modifying or optimizing the parameters used by the image segmentation algorithm based on information obtained from preprocessing the fabric's optical properties. For example, this could involve adjusting color thresholds, texture feature weights, edge detection sensitivity, region growing criteria, or specific layer weights in a deep learning model. The aim is to enable the image segmentation algorithm to better adapt to the optical properties of different fabrics, thereby improving segmentation accuracy and robustness.

[0137] In practical applications, segmenting the skin and underwear areas of a person trying on clothes to obtain a segmented underwear image involves preprocessing the image based on the fabric's optical properties and adjusting the image segmentation parameters. Then, image segmentation techniques (such as threshold-based segmentation, edge detection, region growing, cluster analysis, or deep learning models) are used to process the corrected underwear image, precisely separating the skin and underwear areas. This yields an image containing only the underwear area, or an image with the underwear area clearly marked, providing accurate input for subsequent underwear image feature extraction.

[0138] This application's solution, by introducing preprocessing of fabric optical properties, effectively identifies and quantifies the unique optical attributes of lingerie fabrics, such as their translucency, reflectivity, or texture features. Because these optical properties are considered and adjusted in advance, the subsequent image segmentation process obtains more accurate input data. Based on this, the image segmentation parameters are dynamically adjusted according to the preprocessing results, allowing the segmentation algorithm to better adapt to the actual conditions of different fabrics and avoid segmentation errors caused by the complexity of fabric optical properties. In this way, this application overcomes the limitations that traditional image segmentation methods may encounter when processing diverse lingerie fabrics, ensuring accurate separation of the skin region and the lingerie region.

[0139] In some preferred embodiments, a specific example is given below. Suppose the person trying on underwear is wearing a bra made of semi-transparent lace fabric. In traditional methods, due to the semi-transparent nature of lace, the color and texture of the skin underneath may be visible through the fabric, making it difficult for image segmentation algorithms to accurately distinguish between lace areas and skin areas. This could lead to misidentification of some skin areas as underwear, or misidentification of some lace areas as skin.

[0140] According to the scheme of this application, the pixel data in the corrected underwear image is first preprocessed to reflect the optical properties of the fabric. Specifically, the translucency of the lace fabric can be estimated, and the contribution ratio of the skin color and texture beneath the lace to the apparent color of the lace can be calculated based on this translucency. For example, by analyzing the local features of the skin within the lace-covered area and combining this with the material properties of the lace fabric, the translucency estimation parameters can be adjusted to estimate the local translucency of the lace fabric. Subsequently, the skin contribution ratio is subtracted from the apparent color of the lace fabric to obtain the pure inherent optical properties of the lace fabric. Based on this, the pixel data of the lace area is adjusted to make it closer to its true color and texture, rather than an apparent color influenced by skin color.

[0141] Next, the image segmentation parameters are adjusted based on the preprocessing results of the fabric's optical properties. For example, if the preprocessing results show that the lace fabric has high translucency, the threshold used to distinguish colors or textures in the image segmentation algorithm can be adjusted, or the weight of edge features can be increased to better identify the fine structure of the lace.

[0142] Finally, based on the preprocessing results of the fabric's optical properties and the adjusted image segmentation parameters, the skin area and underwear area of ​​the person trying on the garment are segmented. In this way, even with semi-transparent lace fabric, the boundary between the fabric and the skin can be more accurately identified and segmented, resulting in a precise segmented underwear image. For example, a deep learning-based semantic segmentation model can be used, and preprocessed pixel data and adjusted model parameters can be utilized during the training or inference phase to achieve accurate identification of the lace underwear area.

[0143] In response, this application further proposes a specific method for preprocessing the pixel data of the corrected underwear image to improve the fabric optical properties. The aim is to eliminate or reduce the influence of skin color and texture on the apparent color of the underwear fabric through refined processing, thereby obtaining more accurate inherent optical properties of the underwear fabric and providing a more reliable data basis for subsequent image segmentation.

[0144] The preprocessing of pixel data in the corrected underwear image for fabric optical properties in this application includes: estimating the semi-transparency of the underwear fabric; calculating the contribution ratio of the skin color and texture under the underwear to the apparent color of the underwear fabric based on the semi-transparency; subtracting the contribution ratio from the apparent color of the underwear fabric to invert the pure inherent optical properties of the underwear fabric; and adjusting the pixel data of the underwear area based on the pure inherent optical properties.

[0145] Specifically, estimating the translucency of lingerie fabric involves analyzing the fabric's physical properties, such as material, thickness, and weave structure, combined with lighting conditions and the fabric's visual appearance in an image, to quantify the degree to which light can penetrate. This estimation process can utilize a pre-established fabric optical database for matching, or it can analyze pixel data within the lingerie-covered area using image processing algorithms, such as indirectly inferring through local feature sampling. The aim is to provide key parameters for subsequently accurate calculations of the skin's contribution ratio.

[0146] The calculation of the contribution of skin color and texture beneath the underwear to the apparent color of the underwear fabric, based on the aforementioned translucency, can be understood as quantifying the degree to which skin's transmission through the translucent fabric affects the visual effect of the underwear. This typically involves establishing a physical optics model that considers the scattering and absorption of light within the fabric, as well as the path of light reflected from the skin surface and through the fabric. By combining the translucency information of the underwear fabric with the color and texture data of the skin area beneath the underwear, the superposition effect of the skin on the apparent color and texture of the underwear can be calculated. The aim is to accurately separate the underwear fabric's own color from the color transmitted through the skin to avoid confusion.

[0147] In practical applications, subtracting the aforementioned contribution ratio from the apparent color of the underwear fabric to invert its pure inherent optical properties involves removing the skin's contribution ratio from the apparent color observed in the underwear image after obtaining it. This yields the true optical properties of the underwear fabric, such as color, texture, and luster, unaffected by skin. This inversion process aims to restore the fabric's essential visual characteristics, such as its color performance in a completely opaque state. Its purpose is to obtain the true optical properties of the underwear fabric, avoiding skin color interference, thereby providing an accurate benchmark for subsequent image analysis.

[0148] Therefore, adjusting the pixel data of the underwear area based on the aforementioned pure inherent optical properties refers to correcting the pixel values ​​of the underwear area using the inverted pure inherent optical properties. This adjustment can include operations such as color correction, brightness adjustment, or texture enhancement to ensure that the pixel data of the underwear area accurately reflects the visual characteristics of the fabric itself, rather than the blended effect after being superimposed on the skin. The purpose is to provide standardized and accurate input data for subsequent image segmentation and feature extraction, thereby improving the robustness of the overall processing flow.

[0149] This application's solution introduces an estimation of the translucency of the underwear fabric and, based on this, calculates the contribution ratio of the skin color and texture beneath the underwear to the apparent color of the underwear fabric. This allows for the precise extraction of the skin's influence from the apparent color of the underwear. By subtracting this contribution ratio, the pure inherent optical properties of the underwear fabric can be retrieved—that is, the true optical properties of the fabric unaffected by skin color and texture. Subsequently, these pure inherent optical properties are used to adjust the pixel data of the underwear area, enabling the pixel values ​​in the underwear image to more accurately reflect the fabric's inherent color, texture, and gloss characteristics. This series of processes effectively solves the problem of skin color interfering with the apparent color of the underwear under translucent fabric, ensuring the accuracy of subsequent image segmentation and avoiding segmentation errors caused by color confusion.

[0150] In some preferred embodiments, a specific example is given below. Suppose the person trying on underwear is wearing a bra made of semi-transparent lace fabric, beneath which the skin is darker. In traditional image processing methods, due to the semi-transparency of the lace, the dark skin will show through the lace, making the underwear appear darker in the image than its actual color, and the texture details may be blurred by the skin texture, making it difficult to accurately identify the boundaries of the underwear during image segmentation.

[0151] According to the scheme of this application, firstly, the semi-transparency of the lace fabric is estimated by analyzing the material properties of the lace fabric and the lighting conditions in the image. For example, a deep learning model can be used to analyze the image, or a pre-set fabric database can be queried to obtain a semi-transparency of 0.4. Next, based on the estimated semi-transparency, combined with the color and texture information of the skin area of ​​the person trying on the garment, the contribution ratio of skin color and texture to the apparent color of the lace underwear is calculated. For example, if the skin color is RGB(50, 30, 20) and the lace semi-transparency is 0.4, the contribution value of the skin to the apparent color of the lace can be calculated. Subsequently, this contribution ratio is subtracted from the apparent color of the lace underwear, thereby retrieving the pure inherent optical properties of the lace fabric in the case of complete opacity, such as its true pink or white. Finally, based on the retrieved pure inherent optical properties, the pixel data of the underwear area is adjusted so that the color and texture of the lace underwear in the image can accurately reflect its true fabric properties, rather than a mixed effect influenced by skin color. In this way, subsequent image segmentation algorithms can more accurately identify the boundaries and details of lace lingerie, avoiding confusion with skin areas, thus providing more precise image features for adjusting lingerie design parameters.

[0152] Specifically, estimating the translucency of the underwear fabric includes: sampling local features of the skin within the area covered by the underwear to obtain local skin features; adjusting the translucency estimation parameters based on the local skin features and the material properties of the underwear fabric; and estimating the local translucency of the underwear fabric based on the adjusted translucency estimation parameters.

[0153] Local feature sampling of the skin within the area covered by the underwear refers to extracting visual information such as skin color, texture, and brightness in the area where the underwear contacts or covers the skin using image processing techniques. For example, image segmentation algorithms can be used to identify the edges of the underwear, and then pixel-level sampling can be performed on the skin area inside the underwear edges to obtain the average RGB value, texture features (such as Gabor filter response and Local Binary Pattern (LBP) features), and microstructure information of the skin in that area. These local skin features reflect individual differences in the wearer's skin and are important factors affecting the translucency of the underwear fabric.

[0154] Furthermore, adjusting the semi-transparency estimation parameters based on the aforementioned local skin characteristics and the material properties of the underwear fabric refers to combining the obtained local skin characteristics with the inherent material properties of the underwear fabric (such as fiber type, weave density, thickness, color, and luster) to dynamically adjust the parameters in the mathematical model or lookup table used for semi-transparency estimation. For example, for individuals with darker skin tones, their skin may absorb light more readily, thus requiring adjustment of the corresponding absorption coefficient when estimating semi-transparency. Similarly, different fabric materials exhibit significant differences in their optical properties such as light transmittance and scattering, necessitating the selection or adjustment of corresponding semi-transparency model parameters based on their material properties (e.g., cotton, silk, lace).

[0155] Therefore, estimating the local semi-transparency of the lingerie fabric based on the adjusted semi-transparency estimation parameters means using the adjusted parameters and a preset semi-transparency estimation model or algorithm to calculate the semi-transparency value of the lingerie fabric in a specific local area. This local semi-transparency value can more accurately reflect the actual light transmission performance of the lingerie fabric on the skin of the person currently trying on the garment under specific lighting conditions.

[0156] This application's solution obtains the true optical properties of the wearer's skin by sampling local features within the area covered by the underwear, avoiding the biases that may arise from using a generic skin model. Simultaneously, by incorporating the material properties of the underwear fabric, the translucency estimation model fully considers the fabric's physical characteristics. It is precisely because these two key factors—individual differences in the wearer's skin and the material properties of the underwear fabric—are included in the adjustment process of the translucency estimation parameters that the estimated local translucency of the underwear fabric more closely reflects reality. This personalized and refined estimation method effectively solves the problem of insufficient accuracy in translucency estimation in traditional methods, laying the foundation for accurately calculating the contribution ratio of the skin color and texture beneath the underwear to the apparent color of the underwear fabric.

[0157] In some preferred embodiments, a specific example is given below. Suppose a person trying on lace lingerie has a darker skin tone and some fine texture on the skin surface.

[0158] First, the system performs local feature sampling on the skin within the area covered by the lingerie. For example, through image analysis, it identifies the skin beneath the openwork or semi-transparent areas of the lace lingerie and extracts the average RGB value, brightness, contrast, and texture features (such as skin pores and fine lines) of that area.

[0159] Secondly, based on these local skin characteristics and the material properties of the lace fabric (e.g., the size of the lace holes, the thickness of the threads, the weave density, and the material composition such as nylon or cotton), the system adjusts the translucency estimation parameters. For example, if the skin color is darker, the system increases the weight of the skin's light absorption; if the lace holes are larger, the system adjusts the light transmittance coefficient to reflect higher light transmittance.

[0160] Finally, using these adjusted translucency estimation parameters, the system estimates the local translucency of the lace lingerie on specific skin areas of the wearer. For example, the estimated translucency may be slightly lower for areas with darker skin tones than it would be for lighter skin, as darker skin absorbs more light. In this way, the system obtains a highly personalized and accurate translucency value, providing a reliable basis for subsequent preprocessing of the fabric's optical properties.

[0161] In some embodiments described above in this application, in order to more accurately preprocess the pixel data of the corrected underwear image for fabric optical properties, it is necessary to calculate the contribution ratio of the skin color and texture beneath the underwear to the apparent color of the underwear fabric based on the translucency of the underwear fabric. To this end, this application further proposes a specific method for calculating this contribution ratio to achieve accurate inversion of the optical properties of the underwear fabric.

[0162] Specifically, the above-mentioned calculation of the contribution ratio of the skin color and texture under the underwear to the apparent color of the underwear fabric based on translucency includes the following steps: obtaining transmission spectrum data of the underwear fabric at different incident angles; obtaining the translucency response curve of the underwear fabric at different wavelengths; calculating the directional translucency of the underwear fabric based on the local geometric orientation and incident angle of the light, combined with the transmission spectrum data; calculating the wavelength-dependent translucency of the underwear fabric based on the local color information of the underwear fabric, combined with the translucency response curve; obtaining the local comprehensive translucency of the underwear fabric based on the directional translucency and the wavelength-dependent translucency; and calculating the contribution ratio of the skin color and texture under the underwear to the apparent color of the underwear fabric based on the local comprehensive translucency and the skin color and texture information under the underwear.

[0163] Obtaining the transmission spectrum data of the underwear fabric at different incident angles refers to obtaining the intensity and spectral distribution of transmitted light from the underwear fabric at different incident angles through experimental measurements, physical modeling, or consulting material databases. This data reflects the fabric's absorption, scattering, and transmission characteristics of light, and is fundamental to understanding the fabric's optical behavior. Obtaining the translucency response curve of the underwear fabric at different wavelengths refers to determining the degree of translucency exhibited by the underwear fabric at different light wavelengths. This curve can reveal the fabric's selective transmission capability for specific wavelengths of light; for example, some fabrics may have higher transmittance for red light and lower transmittance for blue light.

[0164] Furthermore, the directional translucency of the lingerie fabric is calculated based on its local geometric orientation and the angle of light incidence, combined with the transmission spectral data. Specifically, the local geometric orientation of the fabric (e.g., the normal direction of the fabric surface) and the angle of light incidence affect the effective path length of light penetrating the fabric and the internal scattering effect. By combining this geometric and lighting information with pre-acquired transmission spectral data, the directional translucency exhibited by the fabric under specific lighting and observation conditions can be accurately calculated. Simultaneously, the wavelength-dependent translucency of the lingerie fabric is calculated based on its local color information and the translucency response curve. The local color information of the fabric reflects its absorption and reflection characteristics of light at different wavelengths. By combining this color information with the translucency response curve, it is possible to deduce how the translucency of the fabric changes with wavelength at different wavelengths, thus providing a more refined description of the fabric's light transmission characteristics.

[0165] Therefore, the local comprehensive translucency of the underwear fabric is obtained based on the directional translucency and the wavelength-dependent translucency. Directional translucency considers the influence of lighting and viewing angle on light transmission, while wavelength-dependent translucency considers the selective transmission of different colors of light by the fabric. Combining these two factors yields a more comprehensive and accurate local comprehensive translucency, which reflects the true light transmission performance of the fabric in a specific local area, under specific lighting conditions, and with specific colors. Finally, based on the local comprehensive translucency and the skin color and texture information beneath the underwear, the contribution ratio of the skin color and texture to the apparent color of the underwear fabric is calculated. This means that, given the overall light transmission capability of the fabric and the inherent color and texture of the skin beneath it, the degree of influence of skin transmission through the fabric on the final observed color and texture of the underwear can be quantified.

[0166] This application's solution meticulously decomposes the factors influencing translucency, considering multiple dimensions such as the incident angle of light, the geometric orientation of the fabric, the fabric's response to different wavelengths, and the fabric's local color information. First, by acquiring transmission spectrum data and translucency response curves, it provides the foundational data for subsequent calculations. Second, it calculates directional translucency and wavelength-dependent translucency separately, making the translucency assessment more comprehensive and accurate, reflecting the true light transmission characteristics of the fabric under different lighting and color conditions. It is precisely because these factors are comprehensively considered that a highly accurate local overall translucency can be obtained, thus laying a solid foundation for subsequent accurate calculations of the contribution ratio of skin color and texture to the apparent color of underwear.

[0167] Specifically, Figure 3 A schematic diagram of an image analysis-based underwear design parameter adjustment system according to an embodiment of the present invention is shown. The system includes:

[0168] The reference object setting module is used to construct optical data of underground reference objects in the fitting room under standard lighting conditions;

[0169] The image acquisition module is used to acquire images of the underwear worn by the person trying on clothes in the fitting area, and simultaneously acquire partial reference images of the reference object;

[0170] An illumination difference calculation module is used to identify the pixel value difference between the local reference image and the optical data;

[0171] A pixel compensation calculation module is used to calculate the pixel gain and pixel offset required for the underwear image based on the pixel value difference.

[0172] A pixel correction module is used to correct each pixel in the underwear image based on the pixel gain and pixel offset to obtain a corrected underwear image;

[0173] The image processing module is used to perform image analysis processing on the corrected underwear image to obtain underwear image features;

[0174] The parameter adjustment module is used to adjust the underwear design parameters based on the underwear image features.

[0175] The image analysis-based lingerie design parameter adjustment system proposed in this application effectively addresses the impact of varying lighting conditions in fitting rooms on the accuracy of image analysis through the collaborative operation of its internal functional modules. Specifically, the reference object setting module establishes a standard optical benchmark, the image acquisition module simultaneously acquires images of the lingerie and the reference object, the lighting difference calculation module quantifies lighting deviations, and the pixel compensation calculation module and pixel correction module perform precise image correction. Subsequently, the image processing module extracts key features, and finally, the parameter adjustment module reliably adjusts the lingerie design parameters. By introducing a reference object and utilizing it for real-time lighting condition monitoring and image correction, this system ensures the accuracy of subsequent image analysis, thereby overcoming the image data distortion problem caused by light attenuation in existing technologies and significantly improving the automation efficiency and user satisfaction of lingerie design parameter adjustment.

[0176] The reference object setting module can be configured as a hardware unit, such as a standalone device containing specific sensors and storage units, or as a software module running on a general-purpose computing device. Its function is to collect data from reference objects placed in the fitting area under standard lighting conditions through a preset calibration process, and store the optical data as a reference. For example, this module can guide the user to photograph a standard color chart using a calibrated camera, and automatically extract the average pixel value of each standard color block and grayscale gradient area on the color chart, saving it as optical data.

[0177] The image acquisition module can be implemented as a camera system integrated within the fitting area. This system can simultaneously capture images of the wearer's underwear and partial reference images of a reference object. The module can be configured with a high-resolution image sensor and autofocus to ensure image quality. For example, once the wearer enters the fitting area and assumes a pose, the image acquisition module can trigger multiple cameras to capture images simultaneously. One camera specifically captures the localized area containing the reference object, while the other cameras capture a full-body image of the wearer in their underwear.

[0178] The illumination difference calculation module can be implemented as a software program running on the processing unit. It receives the local reference image output by the image acquisition module and the optical data provided by the reference object setting module. This module quantifies the deviation between the pixel values ​​of the reference object in the local reference image and the standard optical data by executing image analysis algorithms, such as pixel value comparison algorithms. For example, the module can calculate the average RGB value of a specific area in the local reference image and perform a channel-by-channel subtraction operation with the preset standard RGB value to obtain the pixel value difference in the R, G, and B channels.

[0179] The pixel compensation calculation module can be implemented as a software algorithm, taking the pixel value difference output by the illumination difference calculation module as input. This module maps the pixel value difference to corresponding pixel gain and pixel offset based on a preset compensation model or lookup table. For example, if the pixel values ​​in the red channel are generally low, this module can calculate a positive red channel pixel gain to increase the brightness of the red component in the lingerie image, while simultaneously calculating a corresponding offset to correct color cast.

[0180] The pixel correction module can be implemented as a hardware accelerator or a high-efficiency software routine within the image processing unit. This module receives the original lingerie image from the image acquisition module and the pixel gain and pixel offset output from the pixel compensation calculation module. Its function is to adjust the original RGB values ​​of each pixel in the lingerie image channel-by-channel, for example, using the linear transformation formula: P_corrected = P_original. Gain + Offset. As a result, the color and brightness distribution of the original image is corrected to make it closer to the performance under standard lighting conditions.

[0181] The image processing module can be implemented as a library of image analysis algorithms, running on a high-performance processor. This module receives the corrected underwear image from the pixel correction module and performs a series of image analysis processes on it. For example, this module can use image segmentation techniques to separate the skin area and underwear area of ​​the person trying on the underwear, and then extract the underwear's outline, color histogram, texture features, and fit information to the body to form underwear image features.

[0182] The parameter adjustment module can be implemented as a decision support system or expert system, taking as input the lingerie image features provided by the image processing module. Based on preset design rules, user preferences, or machine learning models, this module provides automatic or semi-automatic adjustment suggestions for lingerie design parameters. For example, if image features indicate gaps or indentations in the bra cups, the parameter adjustment module can suggest adjustments to parameters such as cup size, shape, or strap length to optimize the bra's fit and comfort.

[0183] Compared with existing technologies, the core innovation of the image analysis-based lingerie design parameter adjustment system proposed in this application lies in the introduction of a dedicated reference object setting module and an illumination difference calculation module, which enables real-time monitoring and precise quantification of the lighting conditions in the fitting room. Traditional systems often rely on fixed lighting environments or simple white balance algorithms, making it difficult to effectively cope with complex color shifts and uneven brightness caused by long-term light source decay. The system in this application constructs standard optical data through the reference object setting module, and the image acquisition module simultaneously acquires local images of the reference object. Then, the illumination difference calculation module accurately identifies the pixel value differences between the current lighting conditions and the standard conditions. Based on this, the pixel compensation calculation module and pixel correction module can perform fine-grained correction of the lingerie image, ensuring the authenticity and reliability of the lingerie image features extracted by the subsequent image processing module. For example, in a yellowish lighting environment, traditional systems may misjudge the color of the lingerie or skin areas due to image distortion. However, the system in this application, through its modular correction mechanism, can restore the image to the standard color, thereby avoiding misjudgment and significantly improving the automation efficiency and user satisfaction of lingerie design parameter adjustment, providing more reliable technical support for personalized lingerie customization.

[0184] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for adjusting underwear design parameters based on image analysis, characterized in that, Includes the following steps: Construct optical data of underground reference objects in the fitting room under standard lighting conditions; The system captures images of the underwear worn by the person trying on clothes in the fitting area, and simultaneously captures partial reference images of the reference object. Identify the pixel value differences between the local reference image and the optical data; The required pixel gain and pixel offset for the lingerie image are calculated based on the pixel value differences; Each pixel in the underwear image is corrected based on the pixel gain and pixel offset to obtain the corrected underwear image; The corrected underwear image is subjected to image analysis processing to obtain underwear image features; Adjust underwear design parameters based on the underwear image features; The image analysis processing performed on the corrected underwear image to obtain underwear image features includes: The skin area and underwear area of ​​the person trying on clothes are segmented based on image segmentation technology to obtain the segmented underwear image; Extract the underwear image features from the segmented underwear image; The image segmentation technique is used to segment the skin area and underwear area of ​​the person trying on clothes, obtaining a segmented underwear image including: The pixel data in the corrected underwear image are preprocessed to reflect the optical properties of the fabric. The image segmentation parameters are adjusted based on the preprocessing results of the fabric's optical properties to obtain the adjusted image segmentation parameters. Based on the preprocessing results of the fabric optical properties and the adjusted image segmentation parameters, the skin area and underwear area of ​​the person trying on the clothes are segmented to obtain the segmented underwear image; The preprocessing of the pixel data in the corrected underwear image for fabric optical properties includes: Estimate the translucency of underwear fabric; The contribution ratio of the skin color and texture beneath the underwear to the apparent color of the underwear fabric is calculated based on the translucency. The pure inherent optical properties of the underwear fabric are inverted by subtracting the contribution ratio from the apparent color of the underwear fabric. The pixel data of the underwear area is adjusted according to the inherent optical properties of the purity.

2. The method for adjusting underwear design parameters based on image analysis according to claim 1, characterized in that, The process of identifying pixel value differences between the partial reference image and the optical data includes: Identify edge markers on the reference object, and extract the reference image region on the local reference image based on the edge markers; The average pixel value of each standard color block and grayscale gradient region is obtained based on the reference image area; Calculate the pixel value difference between the average pixel value and the optical data.

3. The method for adjusting underwear design parameters based on image analysis according to claim 1, characterized in that, The calculation of the required pixel gain and pixel offset for the underwear image based on the pixel value difference includes: Obtain the current pixel value of each color channel in the underwear image; Based on the pixel value difference and the current pixel value, configure pixel gain and pixel offset for each color channel.

4. The method for adjusting underwear design parameters based on image analysis according to claim 1, characterized in that, The step of correcting each pixel in the underwear image based on the pixel gain and pixel offset to obtain the corrected underwear image includes: Identify the original RGB value of each pixel in the underwear image; The original RGB value of each pixel is adjusted based on the pixel gain and pixel offset to obtain the adjusted RGB value; Determine whether the adjusted RGB value is within the valid range of color values; If it is determined that the adjusted RGB value is within the valid range of color values, then the adjusted RGB value is used as the corrected RGB value; If it is determined that the adjusted RGB value is not within the valid range of color values, then the original RGB value corresponding to the adjusted RGB value is corrected to the threshold node value according to the proximity relationship between the adjusted RGB value and the threshold node value. The threshold node value is 0 or 255.

5. The method for adjusting underwear design parameters based on image analysis according to claim 1, characterized in that, The estimation of the translucency of the underwear fabric includes: Local skin features are obtained by sampling the skin within the area covered by the underwear. The semi-transparency estimation parameters are adjusted based on the local skin characteristics and the material properties of the underwear fabric. The local semi-transparency of the underwear fabric is estimated based on the adjusted semi-transparency estimation parameters.

6. The method for adjusting underwear design parameters based on image analysis according to claim 1, characterized in that, The calculation of the contribution ratio of the skin color and texture beneath the underwear to the apparent color of the underwear fabric based on the translucency includes: Obtain transmission spectral data of underwear fabric at different incident angles; Obtain the translucency response curves of the underwear fabric at different wavelengths; The directional translucency of the underwear fabric is calculated based on the local geometric orientation and incident angle of the light, combined with the transmission spectral data. The wavelength-dependent translucency of the underwear fabric is calculated based on the local color information of the underwear fabric and the translucency response curve. The local overall translucency of the underwear fabric is obtained based on the directional translucency and the wavelength-dependent translucency. Based on the local integrated translucency and the skin color and texture information under the underwear, the contribution ratio of the skin color and texture under the underwear to the apparent color of the underwear fabric is calculated.

7. A system for adjusting underwear design parameters based on image analysis, characterized in that, The system includes: The reference object setting module is used to construct optical data of underground reference objects in the fitting room under standard lighting conditions; The image acquisition module is used to acquire images of the underwear worn by the person trying on clothes in the fitting area, and simultaneously acquire partial reference images of the reference object; An illumination difference calculation module is used to identify the pixel value difference between the local reference image and the optical data; A pixel compensation calculation module is used to calculate the pixel gain and pixel offset required for the underwear image based on the pixel value difference. A pixel correction module is used to correct each pixel in the underwear image based on the pixel gain and pixel offset to obtain a corrected underwear image; The image processing module is used to perform image analysis processing on the corrected underwear image to obtain underwear image features; The parameter adjustment module is used to adjust the underwear design parameters based on the underwear image features; The image processing module is also used for: The skin area and underwear area of ​​the person trying on clothes are segmented based on image segmentation technology to obtain the segmented underwear image; Extract the underwear image features from the segmented underwear image; The image segmentation technique is used to segment the skin area and underwear area of ​​the person trying on clothes, obtaining a segmented underwear image including: The pixel data in the corrected underwear image are preprocessed to reflect the optical properties of the fabric. The image segmentation parameters are adjusted based on the preprocessing results of the fabric's optical properties to obtain the adjusted image segmentation parameters. Based on the preprocessing results of the fabric optical properties and the adjusted image segmentation parameters, the skin area and underwear area of ​​the person trying on the clothes are segmented to obtain the segmented underwear image; The preprocessing of the pixel data in the corrected underwear image for fabric optical properties includes: Estimate the translucency of underwear fabric; The contribution ratio of the skin color and texture beneath the underwear to the apparent color of the underwear fabric is calculated based on the translucency. The pure inherent optical properties of the underwear fabric are inverted by subtracting the contribution ratio from the apparent color of the underwear fabric. The pixel data of the underwear area is adjusted according to the inherent optical properties of the purity.

Images