Publication printing color correction method and system based on deep learning

Abstract

The invention relates to the technical field of color correction, and discloses a publication printing color correction method and system based on deep learning, and the method achieves the precise correction and compensation of the color of a publication printed matter through a series of steps of cosine distance calculation, color-environment matching model, color deviation calculation, operation parameter adjustment and the like. The core of the method lies in that the accuracy of the printing color is effectively improved and the color quality of the printed matter is ensured to meet the expected standard through careful analysis and processing of the printing color by using the deep learning technology. The correction method provided by the invention has obvious advantages, not only can realize automatic color correction, reduce human intervention and improve the stability and consistency of correction, but also can optimize the operation parameters of the printing equipment, so that the printing equipment can keep a good working state in different printing environments, and the adaptability of the printing process is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of color correction, and in particular to a method and system for publication printing color correction based on deep learning. Background Art

[0002] The importance of publication printing color correction lies in its ability to ensure the color accuracy and consistency of printed materials. In the publishing process, color is one of the key factors in conveying information and attracting readers' attention. If the color of the printed material is inaccurate or deviated, it may lead to errors in information transmission and even affect the reader's reading experience.

[0003] Color correction can improve the quality of printed products. During the printing process, due to the influence of various factors, such as printing equipment, paper, ink, etc., the color of the printed product may deviate. Color correction can make the color of the printed product more accurate, bright, and full by adjusting the printing parameters and color management software, thereby improving the quality of the printed product and enhancing the reader's reading experience. In addition, color correction can also help reduce printing costs. If the color of the printed product is inaccurate, it may take multiple adjustments and corrections to achieve the desired effect, which will not only increase printing time and cost, but also waste resources and materials. Through color correction, the color can be pre-processed and optimized before printing, reducing the number of adjustments and correction costs during the printing process, thereby reducing printing costs and improving printing efficiency.

[0004] The Chinese patent application (publication number CN103152511A) discloses a method for color correction of digital image printing in the field of printing and publishing. The detailed steps of the method are as follows: First, the image processing workstation is responsible for reading the digital image captured by the aerial photogrammetry system and transmitting it to the Eizo-ColorEdge display for intuitive display; secondly, to ensure the color accuracy of the display, the Eizo-ColorEdge display is professionally calibrated using the X-Rite display color calibrator; then, the WINDOWS operating system in the image processing workstation is used for key color gamut control development; finally, in the PHOTOSHOP software environment, the digital image is color corrected, and soft proofing technology under printing color gamut control is developed.

[0005] From the above examples, we can easily find that the limitations of the current color correction methods for printed products are mainly reflected in their high reliance on manual operation and specific software environments, and their limited adaptability to different printing equipment and materials. In addition, given the many complex factors involved in the printing process, such as ink volume, printing speed, printing pressure, etc., they all have a significant impact on the color of the printed product. Therefore, relying solely on software correction may not meet the stringent requirements of high-precision printing.

[0006] Therefore, inventing a method and system for color correction of publication printing is of great significance for ensuring the color accuracy and consistency of printed products. Summary of the invention

[0007] In view of this, the present invention proposes a publication printing color correction method and system based on deep learning, aiming to improve the color accuracy and consistency of printed products.

[0008] The present invention proposes a method for color correction of publication printing based on deep learning, comprising: Printing is performed under a number of printing environment parameters according to the selected standard color parameters to obtain a number of versions of the publication printouts, and image data of the number of versions of the publication printouts are obtained at the same time; Comparing the standard color parameters with each image data, and obtaining a number of comprehensive evaluation values ​​based on cosine distance; Selecting a publication printout corresponding to the smallest comprehensive evaluation value among all comprehensive evaluation values ​​as a standard reference version, and using the image data of the standard reference version as a standard reference parameter; Establishing a color-environment matching model based on the image data and printing environment parameters of all versions of the printed publication; Inputting the target environment parameter into the color-environment matching model, predicting and obtaining the color output parameter corresponding to the target environment parameter, calculating the difference between the predicted color output parameter and the target color parameter, and calculating the color deviation value; It is determined whether the operating parameters of the printing device need to be adjusted according to the color deviation value, and when the determination result is that the operating parameters of the printing device need to be adjusted, the operating parameters of the printing device are adjusted.

[0009] Preferably, the step of comparing the standard color parameter with each image data and obtaining a plurality of comprehensive evaluation values ​​based on the cosine distance comprises the following steps: Preprocessing the acquired image data of the printed publication, including image denoising and color space conversion steps, and recording the preprocessed image data as usable image data; Extracting color characteristic parameters from the available image data; Comparing the extracted color characteristic parameters with the standard color parameters; According to the comparison result, the cosine distance between each image data and the standard color parameter is calculated, and a comprehensive evaluation value between each image data and the standard color parameter is obtained according to the cosine distance; Among all the calculated comprehensive evaluation values, the printed publication corresponding to the smallest comprehensive evaluation value is selected as the standard reference version.

[0010] Preferably, all colors in the image data are extracted and numbered, and the numbers are recorded as i, i=1, 2, ..., n; The color characteristic parameters include brightness, saturation and hue of the color, and the brightness, saturation and hue are expressed as a color vector; ; Among them, Ci is the color vector, Li is the brightness, Si is the saturation, and Hi is the hue; The cosine distance is calculated using the color vector, and the calculation formula includes: ; ; Among them, Ti is the cosine similarity, Di is the cosine distance, and C0 is the color vector of the standard color parameters.

[0011] Preferably, when obtaining a plurality of comprehensive evaluation values ​​based on the cosine distance, it includes: Comprehensively evaluate the cosine distances of all colors of all printed publications of each version to obtain a number of comprehensive evaluation values; The calculation formula of the comprehensive evaluation value is: ; Among them, Z is the comprehensive evaluation value, Wi is the weight of the i-th color, and Di is the cosine distance of the i-th color; Sort several comprehensive evaluation values ​​in descending order, and select the publication printout with the smallest comprehensive evaluation value as the standard reference version.

[0012] Preferably, when the color-environment matching model is established based on the image data of all versions of the printed publications and the printing environment parameters, it includes: Establishing a database based on image data and printing environment parameters of all publication prints, and dividing the database into a test set database and a training set database; Performing model training on a training set database using a deep learning algorithm to establish the color-environment matching model; After the color-environment matching model training is completed, the model is tested and verified using the test set database; The goodness of fit R 2 Based on this, the output effect of the color-environment matching model is evaluated; when R 2 When the value is greater than 0.7, the output effect of the color-environment matching model is good; The color-environment matching model is used to predict the color output parameters of the publication print product by inputting the printing environment parameters.

[0013] Preferably, during the model training process, the model parameters are optimized by a back propagation algorithm so that the color-environment matching model can accurately predict the color output parameters under different printing environment parameters.

[0014] Preferably, calculating the difference between the predicted color output parameter and the target color parameter to calculate the color deviation value includes: Calculating the Euclidean distance between the predicted color output parameters and the target color parameters as the color deviation value; wherein the color output parameters include predicted brightness, predicted saturation and predicted hue, and the target color parameters include target brightness, target saturation and target hue; The calculation formula of the color deviation value is: ; Among them, ΔC is the color deviation value, Lo is the predicted brightness, So is the predicted saturation, Ho is the predicted hue, Lt is the target brightness, St is the target saturation, and Ht is the target hue.

[0015] Preferably, judging whether it is necessary to adjust the operating parameters of the printing device according to the color deviation value includes: Preset a color deviation threshold Cx, and compare the color deviation value ΔC with the color deviation threshold Cx; When the color deviation value is less than or equal to the color deviation threshold, the judgment result is that the operating parameters of the printing device do not need to be adjusted; When the color deviation value is greater than the color deviation threshold, the judgment result is that the operating parameters of the printing device need to be adjusted.

[0016] Preferably, when the judgment result is that the operating parameters of the printing device need to be adjusted, the operating parameters of the printing device are adjusted, including: When the color deviation value is greater than the color deviation threshold, obtaining the difference between the color output parameter and the target color parameter, and adjusting the operating parameters of the printing device according to the difference; The operating parameters include the ink volume, printing speed, printing pressure and printing color sequence of the printing equipment.

[0017] The present invention also provides a publication printing color correction system based on deep learning, comprising: A color feature extraction module is used to extract color feature parameters from image data of printed publications; Standard color parameter library, storing preset standard color parameters; A comparison module, used for comparing the extracted color feature parameters with the standard color parameters in the standard color parameter library; A cosine distance calculation module, used for calculating the cosine distance between each image data and the standard color parameter according to the comparison result; A comprehensive evaluation module, used for obtaining a comprehensive evaluation value between each image data and a standard color parameter according to a cosine distance; A standard reference version selection module is used to select a publication printout corresponding to the smallest comprehensive evaluation value among all calculated comprehensive evaluation values ​​as a standard reference version; A color-environment matching model building module is used to build a color-environment matching model based on image data and printing environment parameters of all versions of printed publications; A color deviation calculation module is used to calculate the difference between the predicted color output parameter and the target color parameter to obtain a color deviation value; An operating parameter adjustment module is used to determine whether the operating parameters of the printing device need to be adjusted according to the color deviation value, and adjust the operating parameters of the printing device according to the color deviation value when necessary; Among them, the color-environment matching model establishment module uses a deep learning algorithm to perform model training on the training set database, establishes a color-environment matching model, and tests and verifies the model through the test set database; the color deviation calculation module calculates the Euclidean distance between the predicted color output parameters and the target color parameters as the color deviation value; the operating parameter adjustment module compares the color deviation value with the preset color deviation threshold. When the color deviation value is greater than the color deviation threshold, the difference between the color output parameter and the target color parameter is obtained, and the operating parameters of the printing equipment are adjusted according to the difference, where the operating parameters include the ink volume, printing speed, printing pressure and printing color sequence of the printing equipment.

[0018] Compared with the prior art, the present invention has the following beneficial effects: 1. Through cosine distance calculation and comprehensive evaluation, the present invention can accurately evaluate the similarity between the color of the publication print and the standard color parameters, thereby selecting the publication print closest to the standard color parameters as the standard reference version, thereby improving the accuracy and reliability of color correction.

[0019] 2. By establishing a color-environment matching model, the present invention can predict the color output parameters of publication printouts based on printing environment parameters, thereby achieving effective compensation for the printing environment and further improving the accuracy and adaptability of color correction.

[0020] 3. The present invention adopts deep learning algorithm for model training and testing, which can automatically optimize model parameters, improve the prediction ability and generalization performance of the color-environment matching model, and make color correction more stable and reliable.

[0021] 4. The present invention can accurately determine whether it is necessary to adjust the operating parameters of the printing equipment by calculating the color deviation value, and adjust the parameters according to the color deviation value, thereby achieving precise control of the printing equipment and further improving the color quality and consistency of publications and printed products.

[0022] 5. The deep learning-based publication printing color correction system provided by the present invention realizes the automatic extraction, comparison, evaluation and correction of the color characteristics of publication prints, greatly improves the automation and intelligence level of printing production, reduces labor costs and improves production efficiency.

[0023] In summary, the present invention provides a method and system for publication printing color correction based on deep learning. Through the steps of cosine distance calculation, color-environment matching model, color deviation calculation and operation parameter adjustment, it realizes the precise correction and compensation of the color of publication prints, improves the color quality and consistency of prints, and has important practical application value and broad market prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings: Figure 1 A schematic diagram of a process flow of a method for color correction of publication printing based on deep learning provided by an embodiment of the present invention; Figure 2 A structural block diagram of a publication printing color correction system based on deep learning provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0025] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art. It should be noted that, in the absence of conflict, the embodiments of the present invention and the features described in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0026] See also Figure 1 As shown, this embodiment provides a method for color correction of publication printing based on deep learning, including: S1. Printing is performed under a number of printing environment parameters according to the selected standard color parameters to obtain a number of versions of the publication printouts, and image data of the number of versions of the publication printouts are obtained at the same time; S2, comparing the standard color parameter with each image data, and obtaining a number of comprehensive evaluation values ​​based on cosine distance; S3, selecting a publication printout corresponding to the smallest comprehensive evaluation value among all comprehensive evaluation values ​​as a standard reference version, and using the image data of the standard reference version as a standard reference parameter; S4, establishing a color-environment matching model based on image data and printing environment parameters of all versions of publication prints; S5, inputting the target environment parameters into the color-environment matching model, predicting and obtaining color output parameters corresponding to the target environment parameters, calculating the difference between the predicted color output parameters and the target color parameters, and calculating the color deviation value; S6. Determine whether it is necessary to adjust the operating parameters of the printing device according to the color deviation value. When the determination result is that it is necessary to adjust the operating parameters of the printing device, adjust the operating parameters of the printing device.

[0027] In some embodiments of the present application, when comparing the standard color parameter with each image data and obtaining a plurality of comprehensive evaluation values ​​based on the cosine distance, the following steps are included: Preprocessing the acquired image data of the printed publication, including image denoising and color space conversion steps, and recording the preprocessed image data as usable image data; Extracting color characteristic parameters from the available image data; Compare the extracted color characteristic parameters with the standard color parameters; According to the comparison result, the cosine distance between each image data and the standard color parameter is calculated, and the comprehensive evaluation value between each image data and the standard color parameter is obtained according to the cosine distance; Among all the calculated comprehensive evaluation values, the printed publication corresponding to the smallest comprehensive evaluation value is selected as the standard reference version.

[0028] It is understandable that this embodiment describes in detail a method for color correction of publication printing based on deep learning. The method ensures the color accuracy of the printed product through multiple steps, thereby improving the quality of the printed product.

[0029] First, printing is performed under multiple different printing environment parameters according to the selected standard color parameters to obtain a series of versions of publication prints. These printing environment parameters include temperature, humidity, printing materials, printing speed, etc. At the same time, the image data of these prints is obtained to provide data support for subsequent color correction.

[0030] Next, the standard color parameters are compared with each image data. Here, the cosine distance is used to measure the similarity between the standard color parameters and the image data. Cosine distance is a distance measurement method commonly used to measure the cosine value of the angle between two vectors. It can effectively capture the subtle differences between colors. By calculating the cosine distance, several comprehensive evaluation values ​​can be obtained, which reflect the degree of match between each version of the print and the standard color parameters.

[0031] Then, the smallest of all the comprehensive evaluation values ​​is selected, and the corresponding publication print is used as the standard reference version. The image data of this version is used as a reference for subsequent color correction. By selecting the print corresponding to the smallest comprehensive evaluation value as the standard reference version, it can be ensured that the selected version is closest to the standard color parameters, thereby improving the accuracy of color correction.

[0032] Subsequently, the color-environment matching model is established using the image data and printing environment parameters of all versions of the publication. This model can help predict the color output that may be produced under different printing environments. By inputting the target environment parameters, the corresponding predicted color output parameters can be obtained.

[0033] Next, the difference between the predicted color output parameters and the target color parameters is calculated to obtain the color deviation value. This deviation value reflects the gap between the predicted color and the actual demand. If the color deviation value exceeds the preset threshold, the operating parameters of the printing equipment need to be adjusted to reduce the deviation.

[0034] Finally, based on the size of the color deviation value, it can be determined whether the operating parameters of the printing equipment need to be adjusted. If the result of the determination is that adjustment is required, the operating parameters of the printing equipment can be fine-tuned to reduce the color deviation value. These adjustments include adjusting the ink volume, printing speed, printing pressure, and printing color sequence.

[0035] It is worth mentioning that when comparing standard color parameters with image data, a preprocessing step is used to improve the accuracy of the comparison. Preprocessing includes two steps: image denoising and color space conversion. Image denoising can remove noise interference in the image and improve the accuracy of color recognition. Color space conversion converts image data from one color space to another to adapt to subsequent color feature extraction and comparison.

[0036] In addition, the present invention also extracts color characteristic parameters from the available image data. These color characteristic parameters include brightness, saturation and hue of the color. The extracted color characteristic parameters are compared with standard color parameters to calculate cosine distance and comprehensive evaluation value.

[0037] In summary, this embodiment provides a method for color correction of publication printing based on deep learning. Through the synergy of multiple steps, the method can ensure the color accuracy of printed products and improve the quality of printed products. At the same time, the method also has strong flexibility and adjustability, and can be fine-tuned according to actual conditions to adapt to different printing environments and needs.

[0038] In some embodiments of the present application, all colors in the image data are extracted and numbered, and the numbers are recorded as i, i=1, 2, ..., n; The color characteristic parameters include brightness, saturation and hue of the color, and the brightness, saturation and hue are expressed as a color vector; ; Among them, Ci is the color vector, Li is the brightness, Si is the saturation, and Hi is the hue; The cosine distance is calculated using the color vector, and the calculation formula includes: ; ; Among them, Ti is the cosine similarity, Di is the cosine distance, and C0 is the color vector of the standard color parameters.

[0039] It is understandable that in many application scenarios of image processing and analysis, the extraction and comparison of color features is a crucial link. This application proposes a cosine distance calculation method based on color vectors, which provides an effective means for color feature matching and recognition of images. The specific implementation steps and advantages of this method will be described in detail below.

[0040] First, extract all the colors in the image data and number them, denoted as i, where i ranges from 1 to n, and n represents the total number of colors in the image. The purpose of this step is to uniquely identify each color in the image, laying the foundation for the subsequent calculation of color feature parameters.

[0041] Next, define color feature parameters, including color brightness, saturation, and hue. These parameters together constitute the basic properties of color, which are essential for describing and distinguishing different colors. This application represents brightness, saturation, and hue as color vectors, denoted as Ci, where Li represents brightness, Si represents saturation, and Hi represents hue. The introduction of color vectors can more intuitively describe and compare the similarities and differences between colors.

[0042] In order to measure the similarity between color vectors, cosine distance is used as the calculation basis. Cosine distance is a measure of the angle between vectors. The smaller the value, the smaller the angle between the vectors, that is, the higher the similarity. By calculating the cosine similarity Ti, we can get the similarity between color vectors, and then quantify the difference between colors by calculating the cosine distance Di.

[0043] This method has broad application prospects in image processing and analysis. For example, in image retrieval, the cosine distance can be used to sort the images in the image library, so as to quickly find images with color features similar to the query image. In addition, in tasks such as target tracking and scene recognition, this method can also provide effective color feature information for the algorithm, improving the accuracy and robustness of the algorithm.

[0044] In summary, this application proposes a cosine distance calculation method based on color vectors, which provides an effective means for color feature matching and recognition of images. By extracting all colors in an image and constructing color vectors, the similarities and differences between colors can be intuitively described and compared. At the same time, using cosine distance as a metric, the differences between colors can be accurately quantified, providing strong support for image processing and analysis tasks.

[0045] In some embodiments of the present application, when obtaining a plurality of comprehensive evaluation values ​​based on the cosine distance, it includes: Comprehensively evaluate the cosine distances of all colors of all printed publications of each version to obtain a number of comprehensive evaluation values; The calculation formula of the comprehensive evaluation value is: ; Among them, Z is the comprehensive evaluation value, Wi is the weight of the i-th color, and Di is the cosine distance of the i-th color; Sort several comprehensive evaluation values ​​in descending order, and select the publication printout with the smallest comprehensive evaluation value as the standard reference version.

[0046] It can be understood that this embodiment deeply explores the comprehensive evaluation method based on cosine distance.

[0047] First, cosine distance is an indicator that measures the angle between two vectors, which reflects the similarity between vectors. In the evaluation of publication prints, the colors of all publication prints of each version can be regarded as a multidimensional vector, and cosine distance can help us measure the similarity between these color vectors.

[0048] Next, a comprehensive evaluation is performed based on the cosine distance of all colors of all printed publications of each version to obtain several comprehensive evaluation values. Specifically, in the formula, Wi is the weight of the i-th color, and Di is the cosine distance of the i-th color. Here, Wi can be set according to actual needs, for example, it can be adjusted according to factors such as the importance of color and the audience's sensitivity to color. Di is obtained by calculating the cosine distance of the corresponding color vectors between versions.

[0049] After obtaining several comprehensive evaluation values, they are sorted in descending order, and the printed publication with the smallest comprehensive evaluation value is selected as the standard reference version. This means that among multiple versions, the version that is closest to other versions in color and has the highest similarity will be selected as the standard reference version.

[0050] To better understand this method, we can give a specific application example. Suppose two publishing houses publish two versions of the same book, and we need to select one of the two versions as the standard reference version. By collecting the color data of the printed materials of each publication in the two versions and calculating the cosine distance between them, we can get the comprehensive evaluation value of each version. Based on these values, we can find that the comprehensive evaluation value of version A is smaller, which means that version A is closer to the other versions in color, so we can select version A as the standard reference version.

[0051] In summary, the comprehensive evaluation method based on cosine distance provides an effective means to select a standard reference version from multiple printed versions of publications. This method fully considers the influence of color similarity between versions, making the selected standard reference version more representative and comparable.

[0052] In some embodiments of the present application, when a color-environment matching model is established based on image data of all versions of printed publications and printing environment parameters, it includes: Establishing a database based on image data and printing environment parameters of all publication prints, and dividing the database into a test set database and a training set database; Use deep learning algorithms to train models on the training set database and establish a color-environment matching model; After the color-environment matching model training is completed, the model is tested and verified using the test set database; The goodness of fit R 2 Based on this, the output effect of the color-environment matching model is evaluated; when R 2 When the value is greater than 0.7, the output effect of the color-environment matching model is good; The color-environment matching model is used to predict the color output parameters of the publication print product by inputting the printing environment parameters.

[0053] It is understandable that this embodiment proposes a color-environment matching model based on deep learning. First, the image data of all publications and printed materials and the corresponding printing environment parameters are collected. These data include but are not limited to the image features such as color, brightness, saturation, and the model of the printing equipment, the type of ink, the material of the paper, and other environmental parameters. By integrating these data together, a huge database is established to provide sufficient data support for the subsequent color-environment matching model.

[0054] Next, this huge database is divided into a test set database and a training set database. The test set database is mainly used for subsequent model testing and verification, while the training set database is used to train the color-environment matching model. This division helps to more comprehensively evaluate the performance of the model and ensure the reliability of the model in practical applications.

[0055] In the model training phase, a deep learning algorithm is used. The deep learning algorithm has powerful feature extraction and classification capabilities and can automatically learn useful information from the data. Through continuous learning and adjustment, a color-environment matching model that can accurately predict color output is established.

[0056] After the color-environment matching model is trained, it is necessary to use the test set database to test and verify the model. This stage is very important because it can help us find possible problems and defects in the model and make corresponding optimizations and improvements. In the testing and verification process, the goodness of fit R 2 As an evaluation indicator. 2 The closer the value is to 1, the better the prediction effect of the model. 2 When the value is greater than 0.7, the output effect of the color-environment matching model is considered good and can be applied to actual production. 2 When the value is less than or equal to 0.7, the model is further optimized and improved to improve its prediction accuracy and reliability.

[0057] Finally, the color-environment matching model can predict the color output parameters of publication prints by inputting printing environment parameters. This model has high practical value in practical applications and can help printing companies more accurately predict and control the color output of printed products, thereby improving production efficiency and product quality.

[0058] In summary, this embodiment provides a method for establishing a color-environment matching model based on deep learning. By collecting and analyzing the image data and printing environment parameters of publications and printed materials, a model with strong predictive capabilities can be established to provide strong support for color management in the printing industry.

[0059] In some embodiments of the present application, during the model training process, the model parameters are optimized by a back propagation algorithm so that the color-environment matching model can accurately predict the color output parameters under different printing environment parameters.

[0060] It is understandable that the training process of the color-environment matching model is studied in depth in the embodiment. The model aims to accurately predict the color output parameters under different printing environment parameters. To achieve this goal, we use a back propagation algorithm to optimize the parameters of the model.

[0061] First of all, we need to clarify the importance of the color-environment matching model. In the printing industry, color accuracy is crucial. Different printing environment parameters, such as printer model, ink type, paper texture, etc., will affect the final color output. Therefore, developing a model that can accurately predict color output is of great significance to improving the quality of printed products and production efficiency.

[0062] During the model training process, the back propagation algorithm is used. This is an optimization algorithm based on gradient descent, which minimizes the loss function by continuously adjusting the model parameters. Specifically, an initial parameter value is set first, and then the gradient of the loss function is iteratively calculated, and the parameter value is updated according to the gradient. In this way, the model can gradually learn how to accurately predict the color output parameters.

[0063] In order to make model training more efficient, a large amount of training data is used. This data comes from different printing environments and color outputs, covering a variety of possible scenarios. By training with this data, the model can gradually learn the influence of different environmental parameters on color output, thereby improving the accuracy of prediction.

[0064] In addition to the training data, we also added some additional features to the model to improve its predictive power. For example, we introduced nonlinear activation functions to enable the model to handle more complex input relationships. In addition, we also used regularization techniques to prevent the model from overfitting and thus improve its generalization ability on unknown data.

[0065] In summary, by using the back-propagation algorithm and a large amount of training data, we have successfully trained a color-environment matching model. This model can accurately predict the color output parameters under different printing environment parameters, providing strong support for the printing industry.

[0066] In some embodiments of the present application, calculating the difference between the predicted color output parameter and the target color parameter, and calculating the color deviation value, includes: Calculating the Euclidean distance between the predicted color output parameters and the target color parameters as the color deviation value; wherein the color output parameters include predicted brightness, predicted saturation and predicted hue, and the target color parameters include target brightness, target saturation and target hue; The calculation formula of the color deviation value is: ; Among them, ΔC is the color deviation value, Lo is the predicted brightness, So is the predicted saturation, Ho is the predicted hue, Lt is the target brightness, St is the target saturation, and Ht is the target hue.

[0067] It is understandable that in the process of in-depth exploration of color management and image processing, the calculation of color deviation is particularly important. Color deviation is the difference between the predicted color output parameters and the actual target color parameters, which reflects the accuracy and control ability of the image processing system over color. This embodiment describes in detail how to calculate this color deviation value.

[0068] First, color output parameters usually include predicted brightness, predicted saturation and predicted hue, which are quantitative descriptions of the output color of the image processing system, while target color parameters, such as target brightness, target saturation and target hue, represent the desired ideal color.

[0069] When calculating the color deviation value, the method used is to calculate the Euclidean distance between the predicted color output parameters and the target color parameters. Euclidean distance is a method of measuring the distance between two points in a multidimensional space. Its calculation formula is based on the square root of the sum of the squares of the differences in each dimension. In this application, we regard this "multidimensional space" as a color space composed of three dimensions: brightness, saturation, and hue.

[0070] By calculating the difference between the predicted color parameters and the target color parameters in each dimension and obtaining the square root of the sum of their squares, the color deviation value ΔC can be obtained. The size of the color deviation value ΔC directly reflects the degree of deviation between the predicted color output parameters and the target color parameters. The smaller the ΔC value, the closer the predicted color is to the target color, and the higher the accuracy and control ability of color management. Conversely, the larger the ΔC value, the greater the deviation between the predicted color and the target color, and corresponding adjustments and optimizations are required.

[0071] In practical applications, the color deviation value ΔC can be reduced by continuously adjusting the parameters of the image processing system, such as brightness, saturation, and hue, thereby improving the accuracy and control ability of the image processing system over color. In addition, the color deviation problem can be further improved by introducing more advanced color management algorithms and technologies, such as color correction and color mapping, to improve the overall performance of the image processing system.

[0072] In short, calculating the color deviation between the predicted color output parameters and the target color parameters is an important part of color management. By deeply understanding the calculation method and significance of color deviation, we can better evaluate and improve the color performance of the image processing system and provide users with a more accurate and vivid image experience.

[0073] In some embodiments of the present application, judging whether it is necessary to adjust the operating parameters of the printing device according to the color deviation value includes: Preset a color deviation threshold Cx, and compare the color deviation value ΔC with the color deviation threshold Cx; When the color deviation value is less than or equal to the color deviation threshold, the judgment result is that the operating parameters of the printing device do not need to be adjusted; When the color deviation value is greater than the color deviation threshold, the judgment result is that the operating parameters of the printing device need to be adjusted.

[0074] It is understandable that this embodiment proposes a method for determining whether it is necessary to adjust the operating parameters of a printing device based on a color deviation value. The core of this method is to set a color deviation threshold and compare it with the actual color deviation value to draw a conclusion on whether it is necessary to adjust the parameters of the printing device.

[0075] First, the color deviation value is expressed as ΔC, which refers to the degree of color difference between the printed product and the standard color sample. This difference may be caused by many factors, such as the accuracy of the printing equipment, the degree of ink mixing, the quality of the paper, etc. Therefore, the color deviation value is one of the important indicators to measure the quality of printed products.

[0076] Next, set a color deviation threshold Cx. This threshold is determined based on our quality requirements, equipment performance, industry standards, etc. Setting a reasonable color deviation threshold can help us more accurately determine whether the operating parameters of the printing equipment need to be adjusted.

[0077] In actual operation, the color deviation value ΔC is compared with the color deviation threshold Cx. If ΔC is less than or equal to Cx, it means that the color deviation of the printed product is within an acceptable range, and there is no need to adjust the operating parameters of the printing equipment. In this case, it can be considered that the printing equipment is operating stably and can meet the quality requirements.

[0078] However, if ΔC is greater than Cx, then the operating parameters of the printing equipment need to be adjusted. This may be due to equipment failure, uneven ink mixing or other reasons that cause excessive color deviation. In this case, it is necessary to adjust the parameters of the printing equipment according to the specific deviation situation to reduce color deviation and improve the quality of printed products.

[0079] In addition, the color deviation value and threshold can be dynamically adjusted based on historical data and statistical analysis. For example, the distribution of color deviation values ​​over a period of time can be analyzed and the threshold can be adjusted based on the actual situation. This can not only improve the accuracy of judgment, but also better adapt to different production environments and quality requirements.

[0080] In short, by setting the color deviation threshold and comparing it with the actual color deviation value, it is possible to effectively determine whether the operating parameters of the printing equipment need to be adjusted. This method is simple and practical, and can ensure the stable and reliable quality of printed products. At the same time, the color deviation value and threshold can be dynamically adjusted in combination with historical data and statistical analysis to further improve the quality and production efficiency of printed products.

[0081] In some embodiments of the present application, when the judgment result is that the operating parameters of the printing device need to be adjusted, the operating parameters of the printing device are adjusted, including: When the color deviation value is greater than the color deviation threshold, obtaining the difference between the color output parameter and the target color parameter, and adjusting the operating parameters of the printing device according to the difference; The operating parameters include the ink volume, printing speed, printing pressure and printing color sequence of the printing equipment.

[0082] When the color deviation value is less than or equal to the color deviation threshold, the operating parameters of the current printing device are kept unchanged, and the change of the color deviation value continues to be monitored.

[0083] It is understandable that in the printing industry, the adjustment of equipment operating parameters is a crucial task. It is related to the quality of printed products, production efficiency and cost control. In this embodiment, when it is detected that the color deviation value exceeds the preset color deviation threshold, the operating parameters of the printing equipment are adjusted to ensure that the color quality of the printed product meets the requirements.

[0084] This adjustment process involves the precise calculation of the difference between the color output parameters and the target color parameters. Based on this difference, the system will intelligently adjust the operating parameters of the printing equipment, including ink volume, printing speed, printing pressure and printing color sequence. These parameters have a direct impact on the color effect of the printed product.

[0085] First of all, the adjustment of ink volume is the key to ensure color vividness and saturation. Too much or too little ink will cause color distortion. The system will accurately adjust the ink volume according to the color deviation value to make the printed product present the best color effect.

[0086] Secondly, the adjustment of printing speed is crucial to production efficiency. Under the premise of ensuring color quality, appropriate printing speed can improve production efficiency and reduce costs. The system will reasonably adjust the printing speed according to actual production needs to achieve a balance between color and efficiency.

[0087] In addition, the adjustment of printing pressure has a significant impact on the clarity of printed products. Appropriate printing pressure can ensure the uniform distribution of ink on the paper and improve the clarity of printed products. The system will accurately adjust the printing pressure according to the paper type and printing requirements to obtain the best printing effect.

[0088] Finally, the adjustment of printing color sequence is also very important. Different color sequence combinations will have different effects on the color effects of printed products. The system will intelligently adjust the color sequence combination according to the color deviation value and target color parameters to make the printed product present the best color effect.

[0089] In summary, this embodiment ensures the color quality of printed products by adjusting the operating parameters of the printing equipment.

[0090] See also Figure 2 As shown, the present invention also provides a publication printing color correction system based on deep learning, comprising: A color feature extraction module is used to extract color feature parameters from image data of printed publications; Standard color parameter library, storing preset standard color parameters; A comparison module, used for comparing the extracted color feature parameters with the standard color parameters in the standard color parameter library; A cosine distance calculation module, used for calculating the cosine distance between each image data and the standard color parameter according to the comparison result; A comprehensive evaluation module, used for obtaining a comprehensive evaluation value between each image data and a standard color parameter according to a cosine distance; A standard reference version selection module is used to select a publication printout corresponding to the smallest comprehensive evaluation value among all calculated comprehensive evaluation values ​​as a standard reference version; A color-environment matching model building module is used to build a color-environment matching model based on image data and printing environment parameters of all versions of printed publications; A color deviation calculation module is used to calculate the difference between the predicted color output parameter and the target color parameter to obtain a color deviation value; An operating parameter adjustment module is used to determine whether the operating parameters of the printing device need to be adjusted according to the color deviation value, and adjust the operating parameters of the printing device according to the color deviation value when necessary; Among them, the color-environment matching model establishment module uses a deep learning algorithm to perform model training on the training set database, establishes a color-environment matching model, and tests and verifies the model through the test set database; the color deviation calculation module calculates the Euclidean distance between the predicted color output parameters and the target color parameters as the color deviation value; the operating parameter adjustment module compares the color deviation value with the preset color deviation threshold. When the color deviation value is greater than the color deviation threshold, the difference between the color output parameter and the target color parameter is obtained, and the operating parameters of the printing equipment are adjusted according to the difference, where the operating parameters include the ink volume, printing speed, printing pressure and printing color sequence of the printing equipment.

[0091] It is understandable that the present invention also proposes a publication printing color correction system based on deep learning, which achieves high-precision correction of the color of publication prints through the collaborative work of multiple modules.

[0092] First, the color feature extraction module is responsible for extracting color feature parameters from the image data of the publication print. These parameters include key information such as color hue, saturation, brightness, etc., which provide basic data for subsequent color correction.

[0093] Secondly, the standard color parameter library stores preset standard color parameters. These standard color parameters are obtained by statistics and analysis of a large number of samples and are highly representative and accurate. By comparing with the extracted color feature parameters, it can be preliminarily determined whether the color of the publication print meets the requirements.

[0094] In the comparison module, the system compares the extracted color feature parameters with the standard color parameters in the standard color parameter library. By comparing the difference between the two, it can be preliminarily determined whether there is any deviation in the color of the publication print.

[0095] The cosine distance calculation module calculates the cosine distance between each image data and the standard color parameters based on the comparison results. Cosine distance is a method to measure the similarity between two vectors. The smaller the value, the more similar the two vectors are. By calculating the cosine distance, the difference between the color of the publication print and the standard color can be quantified.

[0096] The comprehensive evaluation module obtains the comprehensive evaluation value between each image data and the standard color parameters based on the cosine distance. The comprehensive evaluation value comprehensively considers multiple aspects of color deviation, such as hue, saturation, brightness, etc., and provides a basis for the subsequent selection of standard reference versions.

[0097] In the standard reference version selection module, the system selects the publication print corresponding to the smallest comprehensive evaluation value among all calculated comprehensive evaluation values ​​as the standard reference version. The standard reference version is the benchmark for color correction, and other versions will be adjusted based on it.

[0098] The color-environment matching model building module builds a color-environment matching model based on the image data and printing environment parameters of all versions of the publication printouts. The model uses a deep learning algorithm to train the model on the training set database, and tests and verifies the model through the test set database. By building a color-environment matching model, the system can more accurately predict the changes in color output under different printing environments.

[0099] The color deviation calculation module calculates the difference between the predicted color output parameters and the target color parameters to obtain the color deviation value. The color deviation value reflects the difference between the predicted color and the actual color, and is an important indicator for evaluating the color correction effect.

[0100] Finally, the operating parameter adjustment module determines whether the operating parameters of the printing device need to be adjusted based on the color deviation value. When the color deviation value is greater than the color deviation threshold, the system obtains the difference between the color output parameter and the target color parameter, and adjusts the operating parameters of the printing device based on the difference. These operating parameters include the ink volume, printing speed, printing pressure, and printing color sequence of the printing device. By adjusting these parameters, the quality of color output can be effectively improved.

[0101] In summary, the publication printing color correction system based on deep learning provided by the present invention realizes high-precision correction of the color of publication prints through the collaborative work of multiple modules. The system not only improves the color quality of prints, but also reduces printing costs and time costs, and has important application value.

[0102] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0103] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0104] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0105] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0106] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for color correction of publication printing based on deep learning, characterized in that: include: Printing is performed under a number of printing environment parameters according to the selected standard color parameters to obtain a number of versions of the publication printouts, and image data of the number of versions of the publication printouts are obtained at the same time; Comparing the standard color parameters with each image data, and obtaining a number of comprehensive evaluation values ​​based on cosine distance; Selecting a publication printout corresponding to the smallest comprehensive evaluation value among all comprehensive evaluation values ​​as a standard reference version, and using the image data of the standard reference version as a standard reference parameter; Establishing a color-environment matching model based on the image data and printing environment parameters of all versions of the printed publication; Inputting the target environment parameter into the color-environment matching model, predicting and obtaining the color output parameter corresponding to the target environment parameter, calculating the difference between the predicted color output parameter and the target color parameter, and calculating the color deviation value; Determining whether it is necessary to adjust the operating parameters of the printing device according to the color deviation value, and adjusting the operating parameters of the printing device when the determination result is that it is necessary to adjust the operating parameters of the printing device; The step of comparing the standard color parameters with each image data and obtaining a plurality of comprehensive evaluation values ​​based on the cosine distance comprises the following steps: Preprocessing the acquired image data of the printed publication, including image denoising and color space conversion steps, and recording the preprocessed image data as usable image data; Extracting color characteristic parameters from the available image data; Comparing the extracted color characteristic parameters with the standard color parameters; According to the comparison result, the cosine distance between each image data and the standard color parameter is calculated, and a comprehensive evaluation value between each image data and the standard color parameter is obtained according to the cosine distance; Among all the calculated comprehensive evaluation values, the printed publication corresponding to the smallest comprehensive evaluation value is selected as the standard reference version.

2. The method for color correction of publication printing based on deep learning according to claim 1, characterized in that: Extract all colors in the image data and number them, and record the numbers as i, i=1, 2, ..., n; The color characteristic parameters include brightness, saturation and hue of the color, and the brightness, saturation and hue are expressed as a color vector; ； Among them, Ci is the color vector, Li is the brightness, Si is the saturation, and Hi is the hue; The cosine distance is calculated using the color vector, and the calculation formula includes: ； ； Wherein, Ti is the cosine similarity, Di is the cosine distance, and C0 is the color vector of the standard color parameter.

3. The method for color correction of publication printing based on deep learning according to claim 2, characterized in that: When obtaining several comprehensive evaluation values ​​based on cosine distance, including: Comprehensively evaluate the cosine distances of all colors of all printed publications of each version to obtain a number of comprehensive evaluation values; The calculation formula of the comprehensive evaluation value is: ； Among them, Z is the comprehensive evaluation value, Wi is the weight of the i-th color, and Di is the cosine distance of the i-th color; Several comprehensive evaluation values ​​are sorted in descending order, and the publication printout with the smallest comprehensive evaluation value is selected as the standard reference version.

4. The method for color correction of publication printing based on deep learning according to claim 2, characterized in that: The method of establishing a color-environment matching model based on the image data of all versions of the printed publications and the printing environment parameters includes: Establishing a database based on image data and printing environment parameters of all publication prints, and dividing the database into a test set database and a training set database; Performing model training on a training set database using a deep learning algorithm to establish the color-environment matching model; After the color-environment matching model training is completed, the model is tested and verified using the test set database; The goodness of fit R 2 Based on this, the output effect of the color-environment matching model is evaluated; when R 2 When the value is greater than 0.7, the output effect of the color-environment matching model is good; The color-environment matching model is used to predict the color output parameters of the publication print product by inputting the printing environment parameters.

5. The method for color correction of publication printing based on deep learning according to claim 4, characterized in that: During the model training process, the model parameters are optimized by a back propagation algorithm so that the color-environment matching model can accurately predict the color output parameters under different printing environment parameters.

6. The method for color correction of publication printing based on deep learning according to claim 4, characterized in that: Calculate the difference between the predicted color output parameters and the target color parameters. Calculate the color deviation value, including: Calculating the Euclidean distance between the predicted color output parameters and the target color parameters as the color deviation value; wherein the color output parameters include predicted brightness, predicted saturation and predicted hue, and the target color parameters include target brightness, target saturation and target hue; The calculation formula of the color deviation value is: ； Among them, ΔC is the color deviation value, Lo is the predicted brightness, So is the predicted saturation, Ho is the predicted hue, Lt is the target brightness, St is the target saturation, and Ht is the target hue.

7. The method for color correction of publication printing based on deep learning according to claim 6, characterized in that: When judging whether it is necessary to adjust the operating parameters of the printing device according to the color deviation value, it includes: Preset a color deviation threshold Cx, and compare the color deviation value ΔC with the color deviation threshold Cx; When the color deviation value is less than or equal to the color deviation threshold, the judgment result is that the operating parameters of the printing device do not need to be adjusted; When the color deviation value is greater than the color deviation threshold, the judgment result is that the operating parameters of the printing device need to be adjusted.

8. The method for color correction of publication printing based on deep learning according to claim 7, characterized in that: When the judgment result is that the operating parameters of the printing device need to be adjusted, the operating parameters of the printing device are adjusted, including: When the color deviation value is greater than the color deviation threshold, obtaining the difference between the color output parameter and the target color parameter, and adjusting the operating parameters of the printing device according to the difference; The operating parameters include the ink volume, printing speed, printing pressure and printing color sequence of the printing equipment.

9. A publication printing color correction system based on deep learning, characterized in that: include: A color feature extraction module is used to extract color feature parameters from image data of printed publications; Standard color parameter library, storing preset standard color parameters; A comparison module, used for comparing the extracted color feature parameters with the standard color parameters in the standard color parameter library; A cosine distance calculation module, used for calculating the cosine distance between each image data and the standard color parameter according to the comparison result; A comprehensive evaluation module, used for obtaining a comprehensive evaluation value between each image data and a standard color parameter according to a cosine distance; A standard reference version selection module is used to select a publication printout corresponding to the smallest comprehensive evaluation value among all calculated comprehensive evaluation values ​​as a standard reference version; A color-environment matching model building module is used to build a color-environment matching model based on image data and printing environment parameters of all versions of printed publications; A color deviation calculation module is used to calculate the difference between the predicted color output parameter and the target color parameter to obtain a color deviation value; An operating parameter adjustment module is used to determine whether the operating parameters of the printing device need to be adjusted according to the color deviation value, and adjust the operating parameters of the printing device according to the color deviation value when necessary; Among them, the color-environment matching model establishment module uses a deep learning algorithm to perform model training on the training set database, establishes a color-environment matching model, and tests and verifies the model through the test set database; the color deviation calculation module calculates the Euclidean distance between the predicted color output parameters and the target color parameters as the color deviation value; the operating parameter adjustment module compares the color deviation value with the preset color deviation threshold. When the color deviation value is greater than the color deviation threshold, the difference between the color output parameter and the target color parameter is obtained, and the operating parameters of the printing equipment are adjusted according to the difference, where the operating parameters include the ink volume, printing speed, printing pressure and printing color sequence of the printing equipment.

Images