CN118396987B - Image evaluation method and system for printed publications - Google Patents
Image evaluation method and system for printed publications Download PDFInfo
- Publication number
- CN118396987B CN118396987B CN202410824178.1A CN202410824178A CN118396987B CN 118396987 B CN118396987 B CN 118396987B CN 202410824178 A CN202410824178 A CN 202410824178A CN 118396987 B CN118396987 B CN 118396987B
- Authority
- CN
- China
- Prior art keywords
- data
- image
- color
- light sensation
- balance
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000011156 evaluation Methods 0.000 title claims abstract description 92
- 230000035807 sensation Effects 0.000 claims abstract description 335
- 230000009467 reduction Effects 0.000 claims abstract description 142
- 238000004458 analytical method Methods 0.000 claims abstract description 133
- 238000013210 evaluation model Methods 0.000 claims abstract description 75
- 238000007639 printing Methods 0.000 claims abstract description 45
- 238000000034 method Methods 0.000 claims abstract description 44
- 238000004364 calculation method Methods 0.000 claims abstract description 41
- 230000008447 perception Effects 0.000 claims abstract description 39
- 230000008859 change Effects 0.000 claims description 136
- 238000004422 calculation algorithm Methods 0.000 claims description 60
- 238000005286 illumination Methods 0.000 claims description 48
- 238000012549 training Methods 0.000 claims description 44
- 238000012360 testing method Methods 0.000 claims description 43
- 238000012545 processing Methods 0.000 claims description 42
- 230000006978 adaptation Effects 0.000 claims description 36
- 238000007637 random forest analysis Methods 0.000 claims description 26
- 238000005070 sampling Methods 0.000 claims description 22
- 238000012937 correction Methods 0.000 claims description 21
- 238000010276 construction Methods 0.000 claims description 16
- 206010061274 Malocclusion Diseases 0.000 claims description 14
- 238000003706 image smoothing Methods 0.000 claims description 14
- 230000004069 differentiation Effects 0.000 claims description 12
- 238000000547 structure data Methods 0.000 claims description 12
- 238000012795 verification Methods 0.000 claims description 10
- 230000011218 segmentation Effects 0.000 claims description 9
- 238000000611 regression analysis Methods 0.000 claims description 7
- 238000012614 Monte-Carlo sampling Methods 0.000 claims description 5
- 238000005516 engineering process Methods 0.000 abstract description 9
- 230000000694 effects Effects 0.000 description 18
- 230000008569 process Effects 0.000 description 15
- 230000007704 transition Effects 0.000 description 8
- 238000005457 optimization Methods 0.000 description 7
- 230000000007 visual effect Effects 0.000 description 7
- 239000003086 colorant Substances 0.000 description 6
- 238000013441 quality evaluation Methods 0.000 description 4
- 230000001105 regulatory effect Effects 0.000 description 4
- 230000002349 favourable effect Effects 0.000 description 3
- 238000009499 grossing Methods 0.000 description 3
- 238000004445 quantitative analysis Methods 0.000 description 3
- 238000007619 statistical method Methods 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 230000001186 cumulative effect Effects 0.000 description 2
- 238000013136 deep learning model Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 230000035945 sensitivity Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 230000001276 controlling effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 230000008451 emotion Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000004451 qualitative analysis Methods 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
- 238000011179 visual inspection Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/09—Supervised learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30144—Printing quality
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30168—Image quality inspection
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of image evaluation, in particular to an image evaluation method and an image evaluation system for printed publications. The method comprises the following steps: collecting image data of the printed publication and analyzing a color mode to obtain publication image color mode data; performing image content environment perception analysis and content environment light sensation analysis according to the publication image color mode data to obtain image content light sensation data; carrying out quantile regression mean square error calculation according to the image content light sensation data to obtain balanced quantile error data; constructing an image light sensation color evaluation model according to the balance bit error data to obtain an image light sensation color evaluation model; and performing image printing reduction difficulty evaluation according to the image light sensation color evaluation model to obtain image printing reduction difficulty evaluation data. The invention enables the image evaluation technology to be more accurate through optimizing the image evaluation technology.
Description
Technical Field
The invention relates to the technical field of image evaluation, in particular to an image evaluation method and an image evaluation system for printed publications.
Background
With the development of printing technology, the requirements of people on printing quality are increasing, and the previous visual inspection cannot meet the requirements of accurate evaluation of image quality. Therefore, various image evaluation methods have been developed. These methods include, but are not limited to, color measurement, image analysis, image processing, etc., and objective evaluation of print quality is achieved by performing quantitative analysis on aspects of color, sharpness, contrast, etc. of the image. The evaluation methods not only improve the consistency and stability of the printing quality, but also provide technical support for the development and innovation of the printing industry. However, the image evaluation method of the conventional printed publication cannot accurately analyze the problem that the image is affected by the ambient light sensation, which results in great difficulty in restoration of the printed image, and low evaluation accuracy.
Disclosure of Invention
Based on this, there is a need to provide a method and a system for evaluating images of printed publications to solve at least one of the above-mentioned technical problems.
To achieve the above object, a method for evaluating an image of a printed publication, the method comprising the steps of:
step S1: acquiring image data of the printed publication to obtain an image dataset of the printed publication; performing color pattern analysis on the printed publication image dataset to obtain publication image color pattern data; performing image content environment perception analysis according to the publication image color mode data to obtain image content environment perception data;
Step S2: performing content environment light sensation analysis on the image content environment perception data to obtain image content light sensation data; performing image color granularity reduction degree analysis according to the image content light sensation data to obtain image light sensation color reduction degree data; performing light sensation color gradient balance analysis according to the image light sensation color reduction degree data to obtain light sensation color gradient balance data; performing fractional regression mean square error calculation on the light sensation color gradient balance data to obtain balance fractional error data;
Step S3: performing error compensation on the light sensation color gradual change balance data according to the balance bit error data to obtain light sensation color gradual change balance compensation data; carrying out image light sensation color evaluation model construction on the light sensation color gradient balance compensation data based on a random forest algorithm to obtain an image light sensation color evaluation model;
Step S4: performing image printing reduction difficulty evaluation according to the image light sensation color evaluation model to obtain image printing reduction difficulty evaluation data; and feeding back the image printing reduction difficulty evaluation data to the terminal.
The present invention provides for image data acquisition of printed publications by using a digital camera or scanner to acquire a high quality image dataset, color pattern analysis of the image dataset to determine the color patterns used in the image, e.g., RGB, CMYK, and context aware analysis of the image content based on the color pattern data, including consideration of the scene, light conditions, etc. in the image to better understand the image content; the image content environment perception data are analyzed to know the light sensation condition in the image, the visual effect of the image is evaluated, the light sensation in the image refers to the content environment light sensation data obtained after machine vision analysis, the granularity reduction degree of the image color is analyzed according to the light sensation data, namely the detail degree and the definition of the color in the image, the gradient balance of the light sensation color is further analyzed to determine the smoothness and the naturalness of the color transition in the image, finally, the balance score error data are obtained through fractional regression mean square error calculation on the light sensation color gradient balance data, the balance and the accuracy of the image color gradient balance data can be quantized, the balance score error data are used for carrying out error compensation, the balance problem existing in the image can be automatically corrected, the color gradient balance of the image is improved, the image light sensation color evaluation model can be more accurately evaluated based on a random forest algorithm, a more accurate basis is provided for subsequent evaluation, and the image light sensation color reduction model is evaluated based on the image light sensation printing evaluation. The system can evaluate the difficulty of the image during printing according to the model, such as problems in color reduction, light sensation balance and the like, and feed back image printing reduction difficulty evaluation data to the terminal, so that a printing person or a producer can adjust printing parameters or an image processing strategy according to an evaluation result to improve the reduction quality and effect of the image during printing. Therefore, the invention is the optimization processing of the traditional image evaluation method of the printed publication, solves the problems that the traditional image evaluation method of the printed publication cannot accurately analyze the difficulty of restoring the printed image caused by the influence of the ambient light sensation and the evaluation accuracy is low, and reduces the problems that the difficulty of restoring the printed image caused by the influence of the ambient light sensation and the evaluation accuracy is low.
Preferably, step S1 comprises the steps of:
Step S11: acquiring image data of the printed publication to obtain an image dataset of the printed publication;
Step S12: performing image smoothing processing on the printed publication image data set to obtain a publication image smoothing data set;
step S13: performing color pattern analysis on the printed publication image dataset according to the publication image smoothing dataset to obtain publication image color pattern data;
step S14: and performing image content environment perception analysis according to the publication image smoothing data set and the publication image color mode data to obtain image content environment perception data.
According to the invention, through image data acquisition of the printed publications, the system can acquire the original data to be analyzed, necessary materials are provided for subsequent processing, the acquired image data set is required to have comprehensive and representativeness so as to ensure the accuracy and reliability of the subsequent analysis, the image smoothing processing is beneficial to reducing noise and interference in images, improving the image quality, enabling the subsequent analysis to be more accurate and reliable, the smoothing processing can enable the images to be softer and clearer, improving the machine vision effect, enabling the subsequent analysis to be easier to understand and operate, the color mode analysis can accurately identify the color mode adopted in the images, which is vital for the subsequent color processing and analysis, the color mode data provides a basis for the subsequent image content environment perception, the system can better understand the color characteristics of the image content, and the image content environment perception analysis is carried out through the image smoothing data set and the color mode data, so that the system can comprehensively consider the environmental factors such as scenes and light in the images, the image content environment perception data provides richer information for the subsequent analysis, the accuracy and reliability are improved, and the system can better correspond to the complex conditions.
Preferably, step S2 comprises the steps of:
step S21: performing content environment light sensation analysis on the image content environment perception data to obtain image content light sensation data;
step S22: performing image color granularity reduction degree analysis on the publication image color mode data according to the image content light sensation data to obtain image light sensation color reduction degree data;
step S23: performing light sensation color gradual change fluency assessment according to the image light sensation color restoration degree data to obtain light sensation color gradual change fluency data;
step S24: performing light sensation color gradient balance analysis according to the image light sensation color reduction degree data and the light sensation color gradient smoothness data to obtain light sensation color gradient balance data;
Step S25: and carrying out fractional regression mean square error calculation on the light sensation color gradient balance data to obtain balance fractional error data.
According to the invention, the system can extract the light sensation characteristics in the image by carrying out light sensation analysis on the image content environment perception data, the system comprises information such as illumination intensity, light ray direction and the like, the image content light sensation data provides a basis for subsequent analysis, so that the system can better understand the light sensation condition in the image, and further evaluate the color characteristics of the image more accurately, the system can evaluate the color granularity reduction degree of the image, namely the definition and reduction degree of color details in the image, the image light sensation color reduction degree data can be used as one of important indexes for measuring the image quality, so that reference is provided for subsequent processing and printing quality of the image, the system evaluates the color gradient smoothness of the image, namely the smoothness and naturalness of color transition according to the light sensation color reduction degree data, the light sensation color gradient smoothness data can help to improve the visual experience of the image, so that the color transition is more natural and comfortable, the color gradient balance of the image is analyzed according to the light sensation color reduction degree data and the gradual change balance degree, the color transition balance evaluation of the image can be quantized, and the subsequent adjustment and optimization basis is provided; the method comprises the steps of carrying out fractional regression mean square error calculation on photosensitive color gradient balance data, thereby quantitatively evaluating the balance and accuracy of image color gradient, solving the problem of sensitivity to abnormal values in the traditional mean square error calculation, and reducing the sensitivity of mean square error to the abnormal values by introducing quantiles, so as to obtain more accurate balance quantile error data.
Preferably, step S22 comprises the steps of:
step S221: analyzing the illumination background condition in the image content to obtain illumination background condition data; performing illumination intensity analysis according to the illumination background condition data to obtain illumination background intensity data;
Step S222: carrying out non-uniformity distribution analysis on the illumination intensity of the image content according to the illumination background intensity data to obtain non-uniformity distribution data of the illumination intensity; carrying out image shadow distribution area identification on the illumination intensity non-uniformity distribution data by using a preset image shadow identification model to obtain image shadow distribution area data;
Step S223: performing image color distortion structure analysis on the publication image color mode data according to the image shadow distribution area data and the illumination intensity non-uniformity distribution data to obtain image color distortion structure data; performing image color deformation evaluation according to the image color distortion structure data to obtain image color deformation data;
Step S224: performing image color dislocation analysis according to the image color distortion structure data and the image color deformation data to obtain image color dislocation data; performing adjacent image color distortion rate calculation according to the image color dislocation data to obtain adjacent image color distortion rate data;
Step S225: performing distortion rate gradient blockiness value interval fitting on adjacent image color distortion rate data to obtain distortion rate blockiness value fitting data;
step S226: and carrying out image color granularity reduction degree analysis on the publication image color mode data according to the distortion rate blockiness value fitting data and the image content light sensation data to obtain image light sensation color reduction degree data.
The invention analyzes the light sensing data of the image content, the system can extract the light background condition data in the image, including the information of the intensity, direction and uniformity of the light, according to the light background condition data, the system can analyze the intensity of the light, understand the light condition of different areas in the image, provide basis for the subsequent processing, according to the light background intensity data, the system analyzes the non-uniformity distribution of the light in the light sensing data of the image content, namely the difference condition of the light intensity of different areas, the system can identify the shadow distribution area in the image by utilizing the preset image shadow identification model, thereby further understanding the light condition and the characteristics of the image, according to the data of the shadow distribution area and the non-uniformity distribution data of the light intensity, the system analyzes the color distortion structure of the image, namely the color change and distortion condition under the light condition, based on the image color distortion structure data, the system evaluates the color deformation of the image, knows the color change condition of the image under different illumination conditions, including color deviation, saturation change and the like, can extract the color dislocation condition of the image, namely the deviation or dislocation condition of the color position in the image by analyzing the color distortion structure data and the color deformation data of the image, calculates the distortion rate of the adjacent image color, namely the discontinuity degree of the color transition of the adjacent area according to the image color dislocation data, can help evaluate the overall color quality of the image, carries out gradient block effect value interval fitting on the adjacent image color distortion rate data, can extract fitting data of distortion block effect values, namely the gradient block effect condition of each distortion block effect interval, and fits the distortion rate data, the system can better understand the change condition of the distortion rate in different intervals, thereby providing more accurate basis for subsequent analysis, fitting data and image content light sensation data according to the distortion rate blockiness value, carrying out light sensation color rendition analysis on the color mode data of the publication image by the system, and the image light sensation color rendition data can help to evaluate the color rendition of the image under different conditions, namely the color rendition degree and quality of the image.
Preferably, the performing distortion rate gradient blockiness value interval fitting on the adjacent image color distortion rate data comprises the following steps:
Calculating the adjacent image color distortion rate data to obtain adjacent image color distortion rate difference data; drawing an image color distortion rate gradient change curve according to the adjacent image color distortion rate difference data to obtain a color distortion rate gradient change curve;
Extracting extreme points according to the color distortion rate gradient change curve to obtain a gradient change extreme point data set; calculating the turning angles of the curves among different polar points according to the gradient change extreme point data set to obtain gradient change curve turning angle data;
Carrying out turning angle classification processing on gradient change curve turning angle data by using a preset turning angle classification judgment threshold value to obtain turning angle classification data; carrying out staged feature segmentation analysis on the color distortion rate gradient change curve according to the turning angle classification data and the gradient change curve turning angle data to obtain distortion rate staged feature segmentation data;
Calculating distortion rate stage change variance according to the distortion rate stage characteristic segment data to obtain distortion rate stage change variance data; performing distortion rate normal distribution analysis on the distortion rate stage characteristic segmented data according to the distortion rate stage variation variance data to obtain distortion rate stage normal distribution data;
Carrying out Monte Carlo sampling treatment on the normal distribution data of the distortion rate stage to obtain normal uniform sampling data of the distortion rate;
performing uniform Lagrange interpolation processing on the distortion rate distribution according to the distortion rate normal uniform sampling data to obtain distortion rate distribution fitting data;
And performing distortion rate gradient blockiness value interval fitting on the adjacent image color distortion rate data according to the distortion rate distribution fitting data and the distortion rate stage variation variance data to obtain distortion rate blockiness value fitting data.
According to the invention, by calculating the color distortion rate difference value between adjacent images, the system can accurately capture the change trend of the color distortion rate, is favorable for analyzing the color change condition between the images, drawing the color distortion rate gradient change curve can intuitively show the change condition of the image color distortion rate, helping a user understand the distribution and change trend of the color distortion rate in the images, extracting extreme points in the gradient change curve, calculating curve turning angles among different extreme points, is favorable for determining key nodes of the color distortion rate change in the images, providing a basis for subsequent analysis, classifying the turning angles of the gradient change curve according to a preset turning angle classification threshold, then carrying out stepwise feature sectional analysis on the color distortion rate gradient change curve, can more clearly understand the change feature and trend of the color distortion rate in the images, calculate the variance of the change of the distortion rate stage, then carry out the normal distribution analysis of the distortion rate stepwise feature sectional data, is favorable for determining the stability and distribution condition of the color distortion rate change in the images, the distortion rate is uniformly obtained through the sampling process of the Monte-Carlo, the distribution of the distortion rate is uniformly distributed and the distortion rate is more uniform, and the image distortion rate is more accurate, and the image distortion rate is more completely fitted to the distribution and the image distortion rate is more accurate, and the image distortion rate is more completely fitted, and the image distortion rate is more accurate, and the image distortion rate is more completely fitted.
Preferably, step S25 comprises the steps of:
Step S251: carrying out balance numerical variance calculation on the light sensation color gradient balance data to obtain balance numerical variance data; carrying out multiple comparison analysis on the balance value variance data to obtain balance value difference multiple comparison data;
Step S252: performing balance value Bonferroni correction processing on the light-sensitive color gradient balance data according to the balance value difference multiple comparison data to obtain light-sensitive color gradient balance correction data;
Step S253: performing fractional regression analysis on the light sensation color gradient balance data according to the light sensation color gradient balance correction data to obtain color gradient balance quantile data;
Step S254: and carrying out quantile regression mean square error calculation according to the color gradient balance quantile data to obtain balance quantile error data.
The system can evaluate the balance degree of the color gradation by calculating the numerical variance of the light-sensitive color gradation balance data. The multiple comparison analysis can help to determine the difference between different balance data, provides a basis for subsequent analysis, and can effectively control the statistical significance level, reduce the possibility of false positive discovery, ensure the reliability and accuracy of data analysis, and use the light-sensitive color gradient balance correction data to carry out quantile regression analysis, thereby being beneficial to determining the distribution rule and trend of color gradient. The method can help the user to know the distribution condition of the color gradient in the image and understand the distribution condition more deeply, and the fitting degree of the regression model can be estimated by calculating the regression mean square error of the color gradient balance dividing data. This helps determine the accuracy and reliability of the quantile regression model.
Preferably, step S3 comprises the steps of:
Step S31: performing error compensation on the light sensation color gradual change balance data according to the balance bit error data to obtain light sensation color gradual change balance compensation data;
step S32: carrying out required weight proportion adaptation calculation on the light sensation color gradient balance compensation data and the image light sensation color reduction degree data by using a weight factor proportion division algorithm to obtain a weight adaptation proportion value; wherein the weight adaptation ratio value comprises a color gradient balance weight ratio value and a color rendition weight ratio value;
Step S33: and carrying out image light sensation color evaluation model construction on the light sensation color gradient balance compensation data and the image light sensation color reduction degree data based on a random forest algorithm and a weight adaptation proportion value to obtain an image light sensation color evaluation model.
According to the invention, the balance of the light sensation color gradual change balance data is compensated according to the balance bit error data, the color gradual change in the image can be regulated, the balance error is reduced, the visual quality and the color accuracy of the image are improved, the weight factor proportion division algorithm is utilized to carry out weight proportion adaptation calculation on the light sensation color gradual change balance compensation data and the image light sensation color reduction degree data, the weight between the color gradual change balance and the color reduction degree can be regulated according to specific conditions, so that an evaluation model is more fit with actual requirements, and the construction of the image light sensation color evaluation model is carried out on the light sensation color gradual change balance compensation data and the image light sensation color reduction degree data based on a random forest algorithm and weight adaptation proportion value. The evaluation model can comprehensively consider the color gradient balance and the color reproducibility of the image, and provides a more comprehensive and accurate index for image quality evaluation.
Preferably, the weight factor scaling algorithm in step S32 is as follows:
;
Wherein P represents a required weight proportion adaptation result value, t represents a calculation required time pre-estimated value, b represents a complexity coefficient of light sensation color gradient balance compensation data, c represents an image color reduction difficulty coefficient in image light sensation color reduction degree data, w represents a color type quantity value in an image, a represents a color temperature range coefficient in the image, e represents a natural constant, gamma represents color reduction estimation accuracy in the image light sensation color reduction degree data, d represents differentiation, dt represents differentiation of t, namely, a variation quantity of t, epsilon represents an error correction value of a weight factor proportion division algorithm.
The invention constructs a weight factor proportion dividing algorithm, which can adjust the weight between the color gradual change balance and the color reduction degree according to specific conditions, so that the evaluation model is more fit with the actual demand; the algorithm fully considers the calculated required time predicted value t, the parameter represents the time required by the algorithm to execute, and the larger time predicted value is added with the weight proportion adaptive result value P, because the longer the algorithm is executed, the more resources and accuracy are needed to determine the result; the complexity factor b of the light-sensitive color gradation balance compensation data, which represents the complexity of the light-sensitive color gradation balance compensation data. A larger complexity factor b will increase the value of P, as more complex data requires more processing and compensation to achieve the adaptation of the weight ratio; the image color reproduction difficulty coefficient c in the image light-sensitive color reproduction degree data represents the difficulty of image color reproduction. A larger difficulty factor c will increase the value of P because more difficult color reproduction requires more effort and resources to achieve the adaptation of the weight ratio; a color class number value w in the image, the parameter representing the number of color classes in the image. A larger number of color categories will increase the value of P, as different color categories require more weight to adapt; a color temperature range coefficient a in the image, which parameter represents the color temperature range in the image. A larger color temperature range coefficient a will increase the value of P, as a wider color temperature range requires more weight to adapt; a natural constant e, the approximation of which is 2.71828, wherein in the weight factor scaling algorithm, e is used for performing mathematical operation to influence the calculation of the weight ratio adaptation result value P; the color reproduction estimation accuracy gamma in the image light-sensitive color reproduction degree data increases, and when gamma increases, that is, the color reproduction accuracy increases, the corresponding weight ratio adaptation result value P also increases. This is because higher accuracy requirements typically require more computational and processing resources to implement, and thus in the weight factor scaling algorithm, more accurate color reproduction tasks will get more weight ratios. In this way, the accuracy of color reproduction is ensured to be valued and proper weight distribution is obtained in the whole image processing process; d represents differentiation, dt represents differentiation on t, and the differentiation dt can capture the slight change of time t, so that a formula can more accurately simulate the change rule of a t estimated value along with time under the actual condition, and the dt is incorporated into integral operation, so that the cumulative effect of time change on a final result P can be better reflected. This can allow the algorithm to more accurately predict the required weight ratio adaptation result; the error correction value epsilon of the weight factor proportion dividing algorithm is used for correcting the error of the algorithm so as to improve the accuracy of the result.
Preferably, step S33 includes the steps of:
Step S331: based on the color gradient balance weight proportion value and the color rendition weight proportion value, respectively carrying out data weight proportion division on the light sensation color gradient balance compensation data and the image light sensation color rendition weight data to obtain light sensation color gradient balance weight data and image light sensation color rendition weight data;
Step S332: dividing a test set and a training set respectively for light sensation color gradient balance weight data and image light sensation color reduction weight data to obtain a gradient balance weight test set and a gradient balance weight training set respectively, and obtaining a color reduction weight test set and a color reduction weight training set respectively;
Step S333: training set characteristic association is carried out on the gradual change balance weight training set and the color reduction weight training set, and a balance-reduction association training set is obtained; carrying out random sampling treatment on the test set of the gradual change balance weight test set and the color reduction weight test set to obtain a balance-reduction random sampling test set;
Step S334: constructing an initial image light sensation color evaluation model on the balance-reduction degree association training set based on a random forest algorithm to obtain an initial image light sensation color evaluation model;
Step S335: and (3) inputting the balance-reduction degree random sampling test set into an initial image light sensation color evaluation model for model verification test to obtain the image light sensation color evaluation model.
The invention carries out weight proportion division on the data based on the color gradient balance weight proportion value and the color reduction weight proportion value, and can ensure that the influence of the two key factors is fully considered when an evaluation model is constructed. This helps to more accurately evaluate the color gradation balance and color rendition of the image, separating the data set into a test set and a training set, and can provide a reliable data basis during model training and verification. Through respectively obtaining the gradual change balance weight test set, the gradual change balance weight training set, the color rendition weight test set and the color rendition weight training set, the generalization capability and the accuracy of the evaluation model can be ensured, and the correlation between balance and rendition can be effectively captured by carrying out characteristic association on the training set, so that a more reliable evaluation model is established. Meanwhile, the bias can be reduced by carrying out random sampling treatment on the test set, the credibility of the evaluation model is improved, and the initial image light sensation color evaluation model is constructed by utilizing the balance-reduction degree association training set based on a random forest algorithm. The random forest algorithm has the advantages of high efficiency, accuracy, overfitting resistance and the like, can effectively process a large-scale data set, generate a high-quality evaluation model, input a balance-reduction degree random sampling test set into an initial image light sensation color evaluation model for verification test, and can evaluate the performance and accuracy of the model. Such a verification process can help to confirm the reliability of the evaluation model and make necessary adjustments and optimizations to the model to achieve better evaluation results.
Preferably, the present invention provides an image evaluation system of a printed publication for performing the image evaluation method of a printed publication as described above, the image evaluation system of a printed publication comprising:
the image content environment sensing module is used for acquiring image data of the printed publication to obtain an image dataset of the printed publication; performing color pattern analysis on the printed publication image dataset to obtain publication image color pattern data; performing image content environment perception analysis according to the publication image color mode data to obtain image content environment perception data;
The image light sense color analysis module is used for carrying out content environment light sense analysis on the image content environment sensing data to obtain image content light sense data; performing image color granularity reduction degree analysis according to the image content light sensation data to obtain image light sensation color reduction degree data; performing light sensation color gradient balance analysis according to the image light sensation color reduction degree data to obtain light sensation color gradient balance data; performing fractional regression mean square error calculation on the light sensation color gradient balance data to obtain balance fractional error data;
the image light sensation color evaluation model construction module is used for carrying out error compensation on light sensation color gradual change balance data according to balance bit error data to obtain light sensation color gradual change balance compensation data; carrying out image light sensation color evaluation model construction on the light sensation color gradient balance compensation data based on a random forest algorithm to obtain an image light sensation color evaluation model;
the image printing reduction difficulty evaluation module is used for evaluating the image printing reduction difficulty according to the image light sensation color evaluation model to obtain image printing reduction difficulty evaluation data; and feeding back the image printing reduction difficulty evaluation data to the terminal.
The invention has the advantages that the image data acquisition is carried out on the printed publications, which involves using a digital camera or a scanner to acquire a high-quality image data set, carrying out color pattern analysis on the image data set to determine the color patterns used in the image, such as RGB and CMYK, carrying out environment perception analysis on the image content according to the color pattern data, and considering the scene, the light condition and other factors in the image to better understand the image content; the image content environment perception data are analyzed to know the light sensation condition in the image, the visual effect of the image is evaluated, the light sensation in the image refers to the content environment light sensation data obtained after machine vision analysis, the granularity reduction degree of the image color is analyzed according to the light sensation data, namely the detail degree and the definition of the color in the image, the gradient balance of the light sensation color is further analyzed to determine the smoothness and the naturalness of the color transition in the image, finally, the balance score error data are obtained through fractional regression mean square error calculation on the light sensation color gradient balance data, the balance and the accuracy of the image color gradient balance data can be quantized, the balance score error data are used for carrying out error compensation, the balance problem existing in the image can be automatically corrected, the color gradient balance of the image is improved, the image light sensation color evaluation model can be more accurately evaluated based on a random forest algorithm, a more accurate basis is provided for subsequent evaluation, and the image light sensation color reduction model is evaluated based on the image light sensation printing evaluation. The system can evaluate the difficulty of the image during printing according to the model, such as problems in color reduction, light sensation balance and the like, and feed back image printing reduction difficulty evaluation data to the terminal, so that a printing person or a producer can adjust printing parameters or an image processing strategy according to an evaluation result to improve the reduction quality and effect of the image during printing. Therefore, the invention is the optimization processing of the traditional image evaluation method of the printed publication, solves the problems that the traditional image evaluation method of the printed publication cannot accurately analyze the difficulty of restoring the printed image caused by the influence of the ambient light sensation and the evaluation accuracy is low, and reduces the problems that the difficulty of restoring the printed image caused by the influence of the ambient light sensation and the evaluation accuracy is low.
Drawings
FIG. 1 is a flow chart of steps of a method for evaluating images of printed publications;
FIG. 2 is a flowchart illustrating the detailed implementation of step S2 in FIG. 1;
fig. 3 is a detailed implementation step flow diagram of step S3 in fig. 1.
The achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.
Detailed Description
The following is a clear and complete description of the technical method of the present patent in conjunction with the accompanying drawings, and it is evident that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, are intended to fall within the scope of the present invention.
Furthermore, the drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus a repetitive description thereof will be omitted. Some of the block diagrams shown in the figures are functional entities and do not necessarily correspond to physically or logically separate entities. The functional entities may be implemented in software or in one or more hardware modules or integrated circuits or in different networks and/or processor methods and/or microcontroller methods.
To achieve the above object, referring to fig. 1 to 3, a method for evaluating an image of a printed publication, the method comprising the steps of:
step S1: acquiring image data of the printed publication to obtain an image dataset of the printed publication; performing color pattern analysis on the printed publication image dataset to obtain publication image color pattern data; performing image content environment perception analysis according to the publication image color mode data to obtain image content environment perception data;
Step S2: performing content environment light sensation analysis on the image content environment perception data to obtain image content light sensation data; performing image color granularity reduction degree analysis according to the image content light sensation data to obtain image light sensation color reduction degree data; performing light sensation color gradient balance analysis according to the image light sensation color reduction degree data to obtain light sensation color gradient balance data; performing fractional regression mean square error calculation on the light sensation color gradient balance data to obtain balance fractional error data;
Step S3: performing error compensation on the light sensation color gradual change balance data according to the balance bit error data to obtain light sensation color gradual change balance compensation data; carrying out image light sensation color evaluation model construction on the light sensation color gradient balance compensation data based on a random forest algorithm to obtain an image light sensation color evaluation model;
Step S4: performing image printing reduction difficulty evaluation according to the image light sensation color evaluation model to obtain image printing reduction difficulty evaluation data; and feeding back the image printing reduction difficulty evaluation data to the terminal.
In the embodiment of the present invention, as described with reference to fig. 1, a schematic flow chart of steps of an image evaluation method of a printed publication according to the present invention is provided, and in this example, the image evaluation method of the printed publication includes the following steps:
step S1: acquiring image data of the printed publication to obtain an image dataset of the printed publication; performing color pattern analysis on the printed publication image dataset to obtain publication image color pattern data; performing image content environment perception analysis according to the publication image color mode data to obtain image content environment perception data;
In embodiments of the present invention, a printed publication is photographed or scanned using a suitable image capturing device (e.g., a camera or scanner) to obtain high quality image data. Ensuring that the captured images cover various types and styles of printed publications, performing color pattern analysis on the captured image data to determine the color pattern employed by each image, such as RGB, CMYK, or other color space. This may be accomplished by image processing software or libraries in programming languages, such as the OpenCV library in Python, to extract color pattern data of the publication image from the color pattern analysis results, including information on distribution of color channels, color depth, etc. These data will be used for subsequent image content context awareness analysis, using image processing and computer vision techniques, to conduct context awareness analysis on the content of each image. This includes identifying objects, scenes, people, etc. in the image and understanding their location, relationship, and context in the image. This can be achieved by deep learning models such as Convolutional Neural Networks (CNNs) or Recurrent Neural Networks (RNNs), or by combining with conventional image processing methods to extract key data from the results of the image content context awareness analysis, such as the main objects contained in the image, scene descriptions, emotion color information.
Step S2: performing content environment light sensation analysis on the image content environment perception data to obtain image content light sensation data; performing image color granularity reduction degree analysis according to the image content light sensation data to obtain image light sensation color reduction degree data; performing light sensation color gradient balance analysis according to the image light sensation color reduction degree data to obtain light sensation color gradient balance data; performing fractional regression mean square error calculation on the light sensation color gradient balance data to obtain balance fractional error data;
In the embodiment of the invention, firstly, content environment light sensation analysis is performed on image content environment perception data. This includes quantitative and qualitative assessment of light intensity, light source direction, illumination uniformity, etc. in the image. The light sensation condition of the image is analyzed by image processing technologies such as histogram equalization, illumination models and machine vision technologies, and then the color granularity reduction degree analysis of the image is performed according to the light sensation data of the image content. This involves evaluating subtle changes in color and granularity in the image to determine the color rendition of the image. The quality and accuracy of color restoration are evaluated by comparing the original image with the processed image, and then, the balance analysis of the gradient of the light sensation color is performed according to the light sensation color restoration degree data of the image. This step aims to evaluate the balance and naturalness of the color gradation in the image. And judging the color gradient balance of the image by analyzing the color transition conditions among different areas in the image, and finally, carrying out fractional regression mean square error calculation on the light sensation color gradient balance data to obtain balance bit error data. This step evaluates the quality of the light-sensitive color gradation balance by comparing the error between the model predicted color gradation and the actual observed color gradation.
Step S3: performing error compensation on the light sensation color gradual change balance data according to the balance bit error data to obtain light sensation color gradual change balance compensation data; carrying out image light sensation color evaluation model construction on the light sensation color gradient balance compensation data based on a random forest algorithm to obtain an image light sensation color evaluation model;
In the embodiment of the invention, firstly, error compensation is carried out on the light sensation color gradual change balance data according to the balance bit error data. This involves adjusting and correcting the color gradation balance of the image to reduce the error based on the previously calculated balance split error data; then, based on a random forest algorithm or other suitable machine learning algorithm, the image light sensation color evaluation model is constructed on the light sensation color gradual change balance data subjected to error compensation. This step aims at establishing a model capable of automatically evaluating the light-sensitive color quality of an image, and requires preprocessing of data before constructing the model, including the steps of data cleaning, feature selection, feature engineering and the like. This helps to improve the performance and generalization ability of the model, training the preprocessed data using a random forest algorithm to construct an image light sensation color evaluation model. In the training process, the model learns the light sensation color characteristics of the image, establishes the association between the characteristics and the color quality, and evaluates and optimizes the model after the training is completed. The method comprises the steps of performing model performance evaluation by using a verification data set, adjusting model parameters or algorithms according to evaluation results to improve accuracy and stability of a model, and finally, applying a trained image light sensation color evaluation model to actual image data to perform color quality evaluation. The model automatically evaluates the light sensation color of the image, and outputs a corresponding evaluation result, thereby providing reference for subsequent image processing and optimization.
Step S4: performing image printing reduction difficulty evaluation according to the image light sensation color evaluation model to obtain image printing reduction difficulty evaluation data; and feeding back the image printing reduction difficulty evaluation data to the terminal.
In the embodiment of the invention, firstly, the image light sensation color evaluation model constructed in the step S3 is utilized to evaluate the light sensation color quality of the image to be evaluated. The step can use the model to predict the light sensation color quality of the image, output a corresponding evaluation result and evaluate the printing reduction difficulty of the image according to the evaluation result of the image light sensation color evaluation model. This involves quantifying and analyzing the print reduction difficulty of the image based on the evaluation data output by the model. For example, the print reduction difficulty of the image may be classified or scored according to the extent and scope of the evaluation data, and the image print reduction difficulty evaluation data may be fed back to the end user or related stakeholders. This may be done by reporting, visual presentation, or data interface, among others. Feedback of the evaluation data can help a user to know the print reduction difficulty of the image and guide subsequent printing or processing operations.
Therefore, the invention is the optimization processing of the traditional image evaluation method of the printed publication, solves the problems that the traditional image evaluation method of the printed publication cannot accurately analyze the difficulty of restoring the printed image caused by the influence of the ambient light sensation and the evaluation accuracy is low, and reduces the problems that the difficulty of restoring the printed image caused by the influence of the ambient light sensation and the evaluation accuracy is low.
Preferably, step S1 comprises the steps of:
Step S11: acquiring image data of the printed publication to obtain an image dataset of the printed publication;
Step S12: performing image smoothing processing on the printed publication image data set to obtain a publication image smoothing data set;
step S13: performing color pattern analysis on the printed publication image dataset according to the publication image smoothing dataset to obtain publication image color pattern data;
step S14: and performing image content environment perception analysis according to the publication image smoothing data set and the publication image color mode data to obtain image content environment perception data.
In the embodiment of the invention, professional image acquisition equipment is used for acquiring high-quality image data of the printed publications. The collected images are ensured to cover various printed publications with different types and styles, the definition and the accuracy of the images are ensured, after the image data are collected, the images are subjected to smoothing treatment so as to reduce noise points and details, and the overall quality of the images is enhanced. This may be accomplished by applying smoothing filters or other image processing techniques to ensure stability and accuracy of subsequent analysis, and performing color pattern analysis on the smoothed image dataset to determine the color patterns employed by each image, e.g., RGB, CMYK, etc. This can be achieved using libraries in image processing software or programming language to ensure accurate acquisition of color information, and image content context awareness analysis using image processing and computer vision techniques in combination with smoothed image data and color pattern data. This includes identifying objects, scenes, and people in the image and understanding their location, relationship, and context in the image. Analysis is realized through a deep learning model or a traditional image processing method, so that accurate environment perception data is ensured to be acquired.
Preferably, step S2 comprises the steps of:
step S21: performing content environment light sensation analysis on the image content environment perception data to obtain image content light sensation data;
step S22: performing image color granularity reduction degree analysis on the publication image color mode data according to the image content light sensation data to obtain image light sensation color reduction degree data;
step S23: performing light sensation color gradual change fluency assessment according to the image light sensation color restoration degree data to obtain light sensation color gradual change fluency data;
step S24: performing light sensation color gradient balance analysis according to the image light sensation color reduction degree data and the light sensation color gradient smoothness data to obtain light sensation color gradient balance data;
Step S25: and carrying out fractional regression mean square error calculation on the light sensation color gradient balance data to obtain balance fractional error data.
As an example of the present invention, referring to fig. 2, the step S2 in this example includes:
step S21: performing content environment light sensation analysis on the image content environment perception data to obtain image content light sensation data;
in the embodiment of the invention, the image processing technology, such as histogram equalization, gray level conversion, machine vision and other methods, are used for carrying out light sensation analysis on the image content environment perception data, and the factors of light intensity, light source direction, illumination uniformity and the like in the image are evaluated to obtain the image content light sensation data, and the light sensation condition of the image is depicted by carrying out quantitative and qualitative analysis on the characteristics of brightness, contrast and the like of the image.
Step S22: performing image color granularity reduction degree analysis on the publication image color mode data according to the image content light sensation data to obtain image light sensation color reduction degree data;
In the embodiment of the invention, the image content light sensation data is utilized to analyze the image color granularity reduction degree of the color mode data of the publication image, the subtle change and granularity of the color in the image are evaluated by comparing the original image with the processed image so as to determine the light sensation color reduction degree of the image, and the color of the image is restored and optimized by using a color reduction algorithm or a color space conversion technology so as to improve the reduction degree of the light sensation color.
Step S23: performing light sensation color gradual change fluency assessment according to the image light sensation color restoration degree data to obtain light sensation color gradual change fluency data;
In the embodiment of the invention, the gradient fluency of the color in the image is evaluated by utilizing the light-sensitive color rendition data of the image, firstly, the position and the intensity of the color gradient are determined by analyzing the color difference of adjacent pixels in the image, and then, the color gradient in the image is quantitatively evaluated by considering the continuity and the smoothness of the gradient. The method involves calculation and analysis of indexes such as gradient, change rate and the like of a gradual change region, and finally, quantitative scoring of light sensation color gradual change smoothness is given according to the frequency and degree of gradual change so as to describe the smoothness and naturalness of color change in an image.
Step S24: performing light sensation color gradient balance analysis according to the image light sensation color reduction degree data and the light sensation color gradient smoothness data to obtain light sensation color gradient balance data;
in the embodiment of the invention, the light-sensitive color reproducibility data and the light-sensitive color gradient smoothness data of the image are combined to carry out deep analysis on the gradient balance of the color in the image, firstly, the gradient transition condition among different colors in the image is considered, including the change of hue, saturation and brightness, secondly, the balance and harmony among various colors in the gradient process and the whole smoothness of the gradient are evaluated, and finally, the evaluation of the gradient balance of the light-sensitive color is given based on the quantized analysis result to reflect the whole quality and the harmony of the gradient of the color in the image.
Step S25: and carrying out fractional regression mean square error calculation on the light sensation color gradient balance data to obtain balance fractional error data.
In the embodiment of the invention, the photosensitive color gradient balance data is further analyzed, a quantile regression mean square error calculation method is used for evaluating the balance of the image, firstly, the quantile needing to be calculated is determined, a series of key quantiles are generally selected to cover the whole data set, then, regression analysis is carried out on each quantile, the corresponding mean square error is calculated, the balance error under the quantile is represented, and finally, the mean square error under each quantile is comprehensively considered to obtain balance quantile error data for comprehensively evaluating the color gradient balance of the image.
Preferably, step S22 comprises the steps of:
step S221: analyzing the illumination background condition in the image content to obtain illumination background condition data; performing illumination intensity analysis according to the illumination background condition data to obtain illumination background intensity data;
Step S222: carrying out non-uniformity distribution analysis on the illumination intensity of the image content according to the illumination background intensity data to obtain non-uniformity distribution data of the illumination intensity; carrying out image shadow distribution area identification on the illumination intensity non-uniformity distribution data by using a preset image shadow identification model to obtain image shadow distribution area data;
Step S223: performing image color distortion structure analysis on the publication image color mode data according to the image shadow distribution area data and the illumination intensity non-uniformity distribution data to obtain image color distortion structure data; performing image color deformation evaluation according to the image color distortion structure data to obtain image color deformation data;
Step S224: performing image color dislocation analysis according to the image color distortion structure data and the image color deformation data to obtain image color dislocation data; performing adjacent image color distortion rate calculation according to the image color dislocation data to obtain adjacent image color distortion rate data;
Step S225: performing distortion rate gradient blockiness value interval fitting on adjacent image color distortion rate data to obtain distortion rate blockiness value fitting data;
step S226: and carrying out image color granularity reduction degree analysis on the publication image color mode data according to the distortion rate blockiness value fitting data and the image content light sensation data to obtain image light sensation color reduction degree data.
In the embodiment of the invention, illumination background conditions are analyzed on the image content light sensing data to know illumination conditions in the image, analysis on the aspects of overall distribution of illumination in the image, positions and intensities of main light sources and the like is involved, quantitative analysis is carried out on the illumination intensities according to the illumination background condition data, illumination intensity data of different areas in the image are obtained through measuring illumination brightness or using devices such as a photometer and the like, the illumination intensity distribution conditions in the image content light sensing data are analyzed based on the illumination background intensity data, whether obvious non-uniformity exists in the illumination intensity in the different areas of the image is detected, a preset image shadow recognition model is utilized, the illumination intensity non-uniformity distribution data is analyzed to identify shadow distribution areas in the image, technologies such as threshold segmentation and area growth based on pixel intensity are involved, the shadow parts in the image are identified and marked, image color distortion structure analysis is carried out on image color distortion pattern data according to the image shadow distribution area data and the illumination intensity non-uniformity distribution data, color distortion effect of the shadow areas on colors in the image is analyzed, color distortion effect of the image is carried out on the color distortion effect of the image, and color distortion effect of color distortion change is carried out on the image color distortion structure is estimated based on the overall distortion structure after the whole image is compared with color distortion is estimated by comparing the color distortion degree in the image with the original color distortion; based on image color distortion structure data and image color deformation data, image color dislocation analysis is carried out, the degree and form of color distortion in an image are determined by analyzing the deviation and dislocation situation among colors in the image, the distortion rate of colors of adjacent images is calculated according to the image color dislocation data, quantization analysis is carried out on color difference among adjacent pixels so as to determine the size and distribution situation of the distortion rate, fitting analysis of gradient block effect value intervals is carried out on the adjacent image color distortion rate data, gradient block effect values of the distortion rate in different intervals are fitted through a statistical method or a mathematical modeling technology so as to describe the distribution characteristics of the distortion rate, image color granularity reduction analysis is carried out on the color mode data of a publication, and the granularity and distribution situation of the colors in the image are analyzed so as to evaluate the color reduction degree and quality of the image.
Preferably, the performing distortion rate gradient blockiness value interval fitting on the adjacent image color distortion rate data comprises the following steps:
Calculating the adjacent image color distortion rate data to obtain adjacent image color distortion rate difference data; drawing an image color distortion rate gradient change curve according to the adjacent image color distortion rate difference data to obtain a color distortion rate gradient change curve;
Extracting extreme points according to the color distortion rate gradient change curve to obtain a gradient change extreme point data set; calculating the turning angles of the curves among different polar points according to the gradient change extreme point data set to obtain gradient change curve turning angle data;
Carrying out turning angle classification processing on gradient change curve turning angle data by using a preset turning angle classification judgment threshold value to obtain turning angle classification data; carrying out staged feature segmentation analysis on the color distortion rate gradient change curve according to the turning angle classification data and the gradient change curve turning angle data to obtain distortion rate staged feature segmentation data;
Calculating distortion rate stage change variance according to the distortion rate stage characteristic segment data to obtain distortion rate stage change variance data; performing distortion rate normal distribution analysis on the distortion rate stage characteristic segmented data according to the distortion rate stage variation variance data to obtain distortion rate stage normal distribution data;
Carrying out Monte Carlo sampling treatment on the normal distribution data of the distortion rate stage to obtain normal uniform sampling data of the distortion rate;
performing uniform Lagrange interpolation processing on the distortion rate distribution according to the distortion rate normal uniform sampling data to obtain distortion rate distribution fitting data;
And performing distortion rate gradient blockiness value interval fitting on the adjacent image color distortion rate data according to the distortion rate distribution fitting data and the distortion rate stage variation variance data to obtain distortion rate blockiness value fitting data.
According to the embodiment of the invention, the difference value between the adjacent image color distortion ratios is calculated according to the adjacent image color distortion ratio data, the difference value can be realized through simple subtraction operation, the adjacent image color distortion ratio difference value data is obtained and is used for describing the color distortion change condition between the adjacent images, the image color distortion ratio gradient change curve is drawn according to the adjacent image color distortion ratio difference value data, the position or index of the adjacent image is represented on the horizontal axis, the difference value of the adjacent image color distortion ratio is represented on the vertical axis, the curve of the color distortion ratio changing along with the image position is drawn, the extremum points are extracted on the color distortion ratio gradient change curve, the extremum points represent the local maximum value or the local minimum value of the curve, the critical point of the color distortion ratio change is reflected, the curve turning angle between different extremum points is calculated according to the gradient change point data set, the turning angle of the curve at the point is obtained, the turning angle of the curve is classified and is judged by utilizing the preset turning angle classification threshold value, the turning angle is classified into different categories according to whether the turning angle exceeds the threshold value, so that the turning angle is further analyzed and processed, the color distortion ratio is different in the different stages, and the characteristic of the gradient change curve is analyzed and the different in the gradient change characteristic of the gradient change phase, and the characteristic of the curve is analyzed at different stages; according to the distortion rate stage characteristic segmentation data, calculating the variance of distortion rate stage change, wherein the variance measures the degree of difference between each data point in a data set and the mean value of the data points, the variance is used for describing the change amplitude of distortion rate in different stages, normal distribution analysis is carried out on the distortion rate stage characteristic segmentation data according to the distortion rate stage change variance data, normal distribution analysis can be used for describing the distribution characteristics of the distortion rate in different stages, the mean value and standard deviation are included, monte Carlo sampling processing is carried out on the distortion rate stage normal distribution data, monte Carlo sampling is a random sampling method, a random phenomenon is simulated by extracting samples from the probability distribution, even Lagrange interpolation processing is carried out according to the distortion rate normal even sampling data, lagrange interpolation is an interpolation method and is used for deducing the values of other points according to the relation between known data points, the distortion rate gradient block effect value interval fitting is carried out on the adjacent image color distortion rate data by using the distortion rate distribution fitting data and the distortion rate stage change variance data, and the purpose of analyzing and the distortion rate gradient block effect value interval is more careful to provide a distortion rate analysis result.
Preferably, step S25 comprises the steps of:
Step S251: carrying out balance numerical variance calculation on the light sensation color gradient balance data to obtain balance numerical variance data; carrying out multiple comparison analysis on the balance value variance data to obtain balance value difference multiple comparison data;
Step S252: performing balance value Bonferroni correction processing on the light-sensitive color gradient balance data according to the balance value difference multiple comparison data to obtain light-sensitive color gradient balance correction data;
Step S253: performing fractional regression analysis on the light sensation color gradient balance data according to the light sensation color gradient balance correction data to obtain color gradient balance quantile data;
Step S254: and carrying out quantile regression mean square error calculation according to the color gradient balance quantile data to obtain balance quantile error data.
In the embodiment of the invention, firstly, light sensation color gradient balance data to be analyzed are collected. These data are obtained experimentally, measured, or otherwise, and represent characteristics of the light-sensitive color gradient, and the collected data are averaged. This average represents the center position of the dataset, which is the center of balance of all data points, and the difference between each data point and the average is calculated. The mean value of the differences can be obtained by subtracting the mean value from each data point and then squaring to obtain the square value of the differences, adding the square values of all the differences, and dividing the square values by the total number of data points to obtain the mean value of the differences, wherein the mean value of the differences is the balance numerical variance. Variance represents the average amount of the discrete degree of the data distribution, and is one of the statistical characteristics of the balance value, wherein the balance value refers to the balance degree of the gradual change in the analysis image, namely whether the gradual change of color, illumination or other characteristics in the image is smooth, natural and uniform, the time value required in the gradual change process, the balance value variance of each region or characteristic in the image is comprehensively considered, whether the visual effect of the whole image is balanced and harmonious is evaluated, and in another embodiment, the balance value variance calculation involves evaluating the color gradual change balance of each sample in the data set. The "balance numerical variance" is a statistical index for measuring the degree of change in light-sensitive color gradient balance of each sample in the data set, and if the variance is smaller, it means that the sample in the data set has little change in light-sensitive color gradient balance, i.e. has higher balance. Conversely, if the variance is large, the samples in the representation dataset are more behaving differently in this respect, with lower balance; according to the balance value variance data, multiple comparison analysis is carried out, multiple comparison analysis can be used for comparing differences among a plurality of groups, so that the degree of difference among balance values can be determined, bonferroni correction processing is carried out on the balance value gradient balance data based on the balance value difference multiple comparison data, bonferroni correction is a multiple comparison correction method and is used for controlling error rate in multiple comparisons, reliability and accuracy of comparison results are improved, quantile regression analysis is carried out based on the balance correction data of the light sense color gradient, quantile regression analysis is a statistical method and is used for researching data regression relations at different quantiles, balance characteristics of light sense color gradient can be comprehensively evaluated, mean square error of quantile regression is calculated according to the balance quantile data of color gradient, the mean square error is an index used for measuring fitting degree of quantile regression model, and adaptability and accuracy of the balance value of the light sense color gradient in regression model can be reflected.
Preferably, step S3 comprises the steps of:
Step S31: performing error compensation on the light sensation color gradual change balance data according to the balance bit error data to obtain light sensation color gradual change balance compensation data;
step S32: carrying out required weight proportion adaptation calculation on the light sensation color gradient balance compensation data and the image light sensation color reduction degree data by using a weight factor proportion division algorithm to obtain a weight adaptation proportion value; wherein the weight adaptation ratio value comprises a color gradient balance weight ratio value and a color rendition weight ratio value;
Step S33: and carrying out image light sensation color evaluation model construction on the light sensation color gradient balance compensation data and the image light sensation color reduction degree data based on a random forest algorithm and a weight adaptation proportion value to obtain an image light sensation color evaluation model.
As an example of the present invention, referring to fig. 3, the step S3 in this example includes:
Step S31: performing error compensation on the light sensation color gradual change balance data according to the balance bit error data to obtain light sensation color gradual change balance compensation data;
In the embodiment of the present invention, first, raw data for balanced quantile error data is collected and prepared. This involves acquiring data from sensors, cameras or other data sources. These data are then pre-processed, including denoising, correction and normalization, to ensure the quality and comparability of the data; and calculating the bit-dividing error of the light-sensitive color gradient balance by using the prepared original data. This typically involves ordering the data by a certain quantile (e.g., fifty percent, ninety percent) and determining an error value at each quantile, and building an error compensation model based on the quantile error data. This involves the use of statistical methods, machine learning techniques, or other mathematical modeling methods. The model is selected by taking the characteristics of the data, the error distribution and the system requirement into consideration, and each light sensation color gradient data point is compensated by using the established error compensation model so as to generate light sensation color gradient balance compensation data. This involves applying some sort of correction or adjustment to each data point to eliminate or reduce the bias introduced by the quantile error, and verifying and adjusting it after the compensation data is generated. This includes comparing with the actual scenario, evaluating the accuracy and feasibility of the compensation effect. According to the verification result, the compensation model or parameters need to be adjusted to further improve the compensation effect, and finally, the generated light sensation color gradient balance compensation data is output to the system and applied to the corresponding light sensation color gradient processing process. Ensuring the correct compensation data.
Step S32: carrying out required weight proportion adaptation calculation on the light sensation color gradient balance compensation data and the image light sensation color reduction degree data by using a weight factor proportion division algorithm to obtain a weight adaptation proportion value; wherein the weight adaptation ratio value comprises a color gradient balance weight ratio value and a color rendition weight ratio value;
In the embodiment of the invention, the weight factor proportion dividing algorithm is utilized to carry out weight proportion adaptation calculation on the light sensation color gradient balance compensation data and the image light sensation color reduction degree data to obtain a color gradient balance weight proportion value and a color reduction degree weight proportion value. Firstly, for color gradient balance compensation data, the step can quantitatively analyze gradient balance effects of different color channels, and determine contribution degree of each channel to overall gradient balance. This involves analyzing the brightness, saturation, etc. of the color channels and giving weight ratio values accordingly. Similar analysis is also performed for the image light-sensitive color rendition data, but focusing on the rendition of the image color, i.e., the degree of color difference between the image and the original scene. Through the two processes, the weight adaptation ratio value for the gradation balance and the color reproducibility can be obtained.
Step S33: and carrying out image light sensation color evaluation model construction on the light sensation color gradient balance compensation data and the image light sensation color reduction degree data based on a random forest algorithm and a weight adaptation proportion value to obtain an image light sensation color evaluation model.
In the embodiment of the invention, an image light sensation color evaluation model is constructed based on a random forest algorithm, and the model is suitable for the scale value by using the weight obtained in the previous step. Specifically, the step can utilize a random forest algorithm, an integrated learning method, to model the light sensation color gradient balance compensation data and the image light sensation color reduction degree data. In the process, the weight adaptation proportion value is used as an important parameter for model training, so that the model can accurately reflect the influence of gradual change balance and color reduction degree on the light sensation color of the image. Finally, the image light sensation color evaluation model obtained through the step can comprehensively consider the gradient balance and the color reduction degree, and provides a reliable basis for evaluating the image color quality.
According to the invention, the balance of the light sensation color gradual change balance data is compensated according to the balance bit error data, the color gradual change in the image can be regulated, the balance error is reduced, the visual quality and the color accuracy of the image are improved, the weight factor proportion division algorithm is utilized to carry out weight proportion adaptation calculation on the light sensation color gradual change balance compensation data and the image light sensation color reduction degree data, the weight between the color gradual change balance and the color reduction degree can be regulated according to specific conditions, so that an evaluation model is more fit with actual requirements, and the construction of the image light sensation color evaluation model is carried out on the light sensation color gradual change balance compensation data and the image light sensation color reduction degree data based on a random forest algorithm and weight adaptation proportion value. The evaluation model can comprehensively consider the color gradient balance and the color reproducibility of the image, and provides a more comprehensive and accurate index for image quality evaluation.
Preferably, the weight factor scaling algorithm in step S32 is as follows:
;
Wherein P represents a required weight proportion adaptation result value, t represents a calculation required time pre-estimated value, b represents a complexity coefficient of light sensation color gradient balance compensation data, c represents an image color reduction difficulty coefficient in image light sensation color reduction degree data, w represents a color type quantity value in an image, a represents a color temperature range coefficient in the image, e represents a natural constant, gamma represents color reduction estimation accuracy in the image light sensation color reduction degree data, d represents differentiation, dt represents differentiation of t, namely, a variation quantity of t, epsilon represents an error correction value of a weight factor proportion division algorithm.
The invention constructs a weight factor proportion dividing algorithm, which can adjust the weight between the color gradual change balance and the color reduction degree according to specific conditions, so that the evaluation model is more fit with the actual demand; the algorithm fully considers the calculated required time predicted value t, the parameter represents the time required by the algorithm to execute, and the larger time predicted value is added with the weight proportion adaptive result value P, because the longer the algorithm is executed, the more resources and accuracy are needed to determine the result; the complexity factor b of the light-sensitive color gradation balance compensation data, which represents the complexity of the light-sensitive color gradation balance compensation data. A larger complexity factor b will increase the value of P, as more complex data requires more processing and compensation to achieve the adaptation of the weight ratio; the image color reproduction difficulty coefficient c in the image light-sensitive color reproduction degree data represents the difficulty of image color reproduction. A larger difficulty factor c will increase the value of P because more difficult color reproduction requires more effort and resources to achieve the adaptation of the weight ratio; a color class number value w in the image, the parameter representing the number of color classes in the image. A larger number of color categories will increase the value of P, as different color categories require more weight to adapt; a color temperature range coefficient a in the image, which parameter represents the color temperature range in the image. A larger color temperature range coefficient a will increase the value of P, as a wider color temperature range requires more weight to adapt; a natural constant e, the approximation of which is 2.71828, wherein in the weight factor scaling algorithm, e is used for performing mathematical operation to influence the calculation of the weight ratio adaptation result value P; the color reproduction estimation accuracy gamma in the image light-sensitive color reproduction degree data increases, and when gamma increases, that is, the color reproduction accuracy increases, the corresponding weight ratio adaptation result value P also increases. This is because higher accuracy requirements typically require more computational and processing resources to implement, and thus in the weight factor scaling algorithm, more accurate color reproduction tasks will get more weight ratios. In this way, the accuracy of color reproduction is ensured to be valued and proper weight distribution is obtained in the whole image processing process; d represents differentiation, dt represents differentiation on t, and the differentiation dt can capture the slight change of time t, so that a formula can more accurately simulate the change rule of a t estimated value along with time under the actual condition, and the dt is incorporated into integral operation, so that the cumulative effect of time change on a final result P can be better reflected. This can allow the algorithm to more accurately predict the required weight ratio adaptation result; the error correction value epsilon of the weight factor proportion dividing algorithm is used for correcting the error of the algorithm so as to improve the accuracy of the result.
Preferably, step S33 includes the steps of:
Step S331: based on the color gradient balance weight proportion value and the color rendition weight proportion value, respectively carrying out data weight proportion division on the light sensation color gradient balance compensation data and the image light sensation color rendition weight data to obtain light sensation color gradient balance weight data and image light sensation color rendition weight data;
Step S332: dividing a test set and a training set respectively for light sensation color gradient balance weight data and image light sensation color reduction weight data to obtain a gradient balance weight test set and a gradient balance weight training set respectively, and obtaining a color reduction weight test set and a color reduction weight training set respectively;
Step S333: training set characteristic association is carried out on the gradual change balance weight training set and the color reduction weight training set, and a balance-reduction association training set is obtained; carrying out random sampling treatment on the test set of the gradual change balance weight test set and the color reduction weight test set to obtain a balance-reduction random sampling test set;
Step S334: constructing an initial image light sensation color evaluation model on the balance-reduction degree association training set based on a random forest algorithm to obtain an initial image light sensation color evaluation model;
Step S335: and (3) inputting the balance-reduction degree random sampling test set into an initial image light sensation color evaluation model for model verification test to obtain the image light sensation color evaluation model.
In the embodiment of the invention, in this step, the data weight ratio division is performed on the photosensitive color gradation balance compensation data and the image photosensitive color reproduction data based on the color gradation balance weight ratio value and the color reproduction weight ratio value. This process aims to ensure trade-off and balance between light-sensitive color gradation balance and image color rendition. First, weight ratio values are set for the color gradation balance and the color reproduction degree, respectively, reflecting the relative importance of weight allocation between the light-sensitive color gradation balance and the image light-sensitive color reproduction degree. Then, these weight ratio values are applied to the raw data to determine weight data of the light sensation color gradation balance and the image light sensation color reproducibility; in this step, the light sensation color gradation balance weight data and the image light sensation color reproducibility weight data are respectively divided into a test set and a training set. The purpose of this step is to enable reliable data training and verification when building the model. By dividing the data set into a training set and a test set, the training set can be used to adjust model parameters when training the model, and the performance of the model can be verified on the test set; in this step, a training set feature association is performed on the graded balance weight training set and the color rendition weight training set. This process involves correlating features in the training set with the desired output (i.e., balance and degree of restoration) so that the model can learn the relationship between the features and the output. Then, the test set random sampling process is carried out on the gradual balance weight test set and the color reproducibility weight test set. The purpose of this step is to ensure the representativeness and diversity of the test set in order to effectively evaluate the generalization ability of the model; in this step, a random forest algorithm is used to construct an initial image light sensation color evaluation model for the balance-reduction degree association training set. Random forests are an integrated learning algorithm that improves the performance and stability of the model by building and combining multiple decision trees. In the step, the model is trained by utilizing the association relation between the balance and the reduction degree so as to accurately evaluate the light sensation color of the image, and finally, in the step, a random sampling test set of the balance-reduction degree is input into an initial image light sensation color evaluation model for model verification test, so that the image light sensation color evaluation model is obtained. By inputting the test set into the model and verifying according to the output result of the model, the performance and accuracy of the model can be evaluated. This step is the last step of the whole process, and aims to confirm that the constructed image light sensation color evaluation model can effectively evaluate the light sensation color of the image and plays a role in practical application.
Preferably, the present invention provides an image evaluation system of a printed publication for performing the image evaluation method of a printed publication as described above, the image evaluation system of a printed publication comprising:
the image content environment sensing module is used for acquiring image data of the printed publication to obtain an image dataset of the printed publication; performing color pattern analysis on the printed publication image dataset to obtain publication image color pattern data; performing image content environment perception analysis according to the publication image color mode data to obtain image content environment perception data;
The image light sense color analysis module is used for carrying out content environment light sense analysis on the image content environment sensing data to obtain image content light sense data; performing image color granularity reduction degree analysis according to the image content light sensation data to obtain image light sensation color reduction degree data; performing light sensation color gradient balance analysis according to the image light sensation color reduction degree data to obtain light sensation color gradient balance data; performing fractional regression mean square error calculation on the light sensation color gradient balance data to obtain balance fractional error data;
the image light sensation color evaluation model construction module is used for carrying out error compensation on light sensation color gradual change balance data according to balance bit error data to obtain light sensation color gradual change balance compensation data; carrying out image light sensation color evaluation model construction on the light sensation color gradient balance compensation data based on a random forest algorithm to obtain an image light sensation color evaluation model;
the image printing reduction difficulty evaluation module is used for evaluating the image printing reduction difficulty according to the image light sensation color evaluation model to obtain image printing reduction difficulty evaluation data; and feeding back the image printing reduction difficulty evaluation data to the terminal.
Therefore, the invention is the optimization processing of the traditional image evaluation method of the printed publication, solves the problems that the traditional image evaluation method of the printed publication cannot accurately analyze the difficulty of restoring the printed image caused by the influence of the ambient light sensation and the evaluation accuracy is low, and reduces the problems that the difficulty of restoring the printed image caused by the influence of the ambient light sensation and the evaluation accuracy is low.
The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.
The foregoing is only a specific embodiment of the invention to enable those skilled in the art to understand or practice the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (7)
1. A method of evaluating an image of a printed publication, comprising the steps of:
step S1: acquiring image data of the printed publication to obtain an image dataset of the printed publication; performing color pattern analysis on the printed publication image dataset to obtain publication image color pattern data; performing image content environment perception analysis according to the publication image color mode data to obtain image content environment perception data;
Step S2: performing content environment light sensation analysis on the image content environment perception data to obtain image content light sensation data; performing image color granularity reduction degree analysis according to the image content light sensation data to obtain image light sensation color reduction degree data; performing light sensation color gradient balance analysis according to the image light sensation color reduction degree data to obtain light sensation color gradient balance data; performing fractional regression mean square error calculation on the light sensation color gradient balance data to obtain balance fractional error data;
Step S3: performing error compensation on the light sensation color gradual change balance data according to the balance bit error data to obtain light sensation color gradual change balance compensation data; carrying out image light sensation color evaluation model construction on the light sensation color gradient balance compensation data based on a random forest algorithm to obtain an image light sensation color evaluation model; wherein, step S3 includes:
Step S31: performing error compensation on the light sensation color gradual change balance data according to the balance bit error data to obtain light sensation color gradual change balance compensation data;
Step S32: carrying out required weight proportion adaptation calculation on the light sensation color gradient balance compensation data and the image light sensation color reduction degree data by using a weight factor proportion division algorithm to obtain a weight adaptation proportion value; wherein the weight adaptation ratio value comprises a color gradient balance weight ratio value and a color rendition weight ratio value; the weight factor proportion dividing algorithm is as follows:
，
Wherein P represents a required weight proportion adaptation result value, t represents a calculation required time pre-estimated value, b represents a complexity coefficient of light sensation color gradual change balance compensation data, c represents an image color reduction difficulty coefficient in image light sensation color reduction degree data, w represents a color type quantity value in an image, a represents a color temperature range coefficient in the image, e represents a natural constant, gamma represents color reduction estimation accuracy in the image light sensation color reduction degree data, d represents differentiation, dt represents differentiation on t, namely a variation quantity of t, epsilon represents an error correction value of a weight factor proportion division algorithm;
step S33: carrying out image light sensation color evaluation model construction on the light sensation color gradient balance compensation data and the image light sensation color reduction degree data based on a random forest algorithm and a weight adaptation proportion value to obtain an image light sensation color evaluation model; wherein, step S33 includes the following steps:
Step S331: based on the color gradient balance weight proportion value and the color rendition weight proportion value, respectively carrying out data weight proportion division on the light sensation color gradient balance compensation data and the image light sensation color rendition weight data to obtain light sensation color gradient balance weight data and image light sensation color rendition weight data;
Step S332: dividing a test set and a training set respectively for light sensation color gradient balance weight data and image light sensation color reduction weight data to obtain a gradient balance weight test set and a gradient balance weight training set respectively, and obtaining a color reduction weight test set and a color reduction weight training set respectively;
Step S333: training set characteristic association is carried out on the gradual change balance weight training set and the color reduction weight training set, and a balance-reduction association training set is obtained; carrying out random sampling treatment on the test set of the gradual change balance weight test set and the color reduction weight test set to obtain a balance-reduction random sampling test set;
Step S334: constructing an initial image light sensation color evaluation model on the balance-reduction degree association training set based on a random forest algorithm to obtain an initial image light sensation color evaluation model;
Step S335: inputting the balance-reduction degree random sampling test set into an initial image light sensation color evaluation model for model verification test to obtain an image light sensation color evaluation model;
Step S4: performing image printing reduction difficulty evaluation according to the image light sensation color evaluation model to obtain image printing reduction difficulty evaluation data; and feeding back the image printing reduction difficulty evaluation data to the terminal.
2. The image evaluation method of a printed publication according to claim 1, wherein step S1 comprises the steps of:
Step S11: acquiring image data of the printed publication to obtain an image dataset of the printed publication;
Step S12: performing image smoothing processing on the printed publication image data set to obtain a publication image smoothing data set;
step S13: performing color pattern analysis on the printed publication image dataset according to the publication image smoothing dataset to obtain publication image color pattern data;
step S14: and performing image content environment perception analysis according to the publication image smoothing data set and the publication image color mode data to obtain image content environment perception data.
3. The image evaluation method of printed publication according to claim 2, wherein step S2 comprises the steps of:
step S21: performing content environment light sensation analysis on the image content environment perception data to obtain image content light sensation data;
step S22: performing image color granularity reduction degree analysis on the publication image color mode data according to the image content light sensation data to obtain image light sensation color reduction degree data;
step S23: performing light sensation color gradual change fluency assessment according to the image light sensation color restoration degree data to obtain light sensation color gradual change fluency data;
step S24: performing light sensation color gradient balance analysis according to the image light sensation color reduction degree data and the light sensation color gradient smoothness data to obtain light sensation color gradient balance data;
Step S25: and carrying out fractional regression mean square error calculation on the light sensation color gradient balance data to obtain balance fractional error data.
4. A method of evaluating an image of a printed publication according to claim 3, wherein step S22 comprises the steps of:
step S221: analyzing the illumination background condition in the image content to obtain illumination background condition data; performing illumination intensity analysis according to the illumination background condition data to obtain illumination background intensity data;
Step S222: carrying out non-uniformity distribution analysis on the illumination intensity of the image content according to the illumination background intensity data to obtain non-uniformity distribution data of the illumination intensity; carrying out image shadow distribution area identification on the illumination intensity non-uniformity distribution data by using a preset image shadow identification model to obtain image shadow distribution area data;
Step S223: performing image color distortion structure analysis on the publication image color mode data according to the image shadow distribution area data and the illumination intensity non-uniformity distribution data to obtain image color distortion structure data; performing image color deformation evaluation according to the image color distortion structure data to obtain image color deformation data;
Step S224: performing image color dislocation analysis according to the image color distortion structure data and the image color deformation data to obtain image color dislocation data; performing adjacent image color distortion rate calculation according to the image color dislocation data to obtain adjacent image color distortion rate data;
Step S225: performing distortion rate gradient blockiness value interval fitting on adjacent image color distortion rate data to obtain distortion rate blockiness value fitting data;
step S226: and carrying out image color granularity reduction degree analysis on the publication image color mode data according to the distortion rate blockiness value fitting data and the image content light sensation data to obtain image light sensation color reduction degree data.
5. The method of image evaluation of printed publications of claim 4, wherein performing a distortion rate gradient blockiness value interval fit to adjacent image color distortion rate data comprises the steps of:
Calculating the adjacent image color distortion rate data to obtain adjacent image color distortion rate difference data; drawing an image color distortion rate gradient change curve according to the adjacent image color distortion rate difference data to obtain a color distortion rate gradient change curve;
Extracting extreme points according to the color distortion rate gradient change curve to obtain a gradient change extreme point data set; calculating the turning angles of the curves among different polar points according to the gradient change extreme point data set to obtain gradient change curve turning angle data;
Carrying out turning angle classification processing on gradient change curve turning angle data by using a preset turning angle classification judgment threshold value to obtain turning angle classification data; carrying out staged feature segmentation analysis on the color distortion rate gradient change curve according to the turning angle classification data and the gradient change curve turning angle data to obtain distortion rate staged feature segmentation data;
Calculating distortion rate stage change variance according to the distortion rate stage characteristic segment data to obtain distortion rate stage change variance data; performing distortion rate normal distribution analysis on the distortion rate stage characteristic segmented data according to the distortion rate stage variation variance data to obtain distortion rate stage normal distribution data;
Carrying out Monte Carlo sampling treatment on the normal distribution data of the distortion rate stage to obtain normal uniform sampling data of the distortion rate;
performing uniform Lagrange interpolation processing on the distortion rate distribution according to the distortion rate normal uniform sampling data to obtain distortion rate distribution fitting data;
And performing distortion rate gradient blockiness value interval fitting on the adjacent image color distortion rate data according to the distortion rate distribution fitting data and the distortion rate stage variation variance data to obtain distortion rate blockiness value fitting data.
6. A method of evaluating an image of a printed publication according to claim 3, wherein step S25 comprises the steps of:
Step S251: carrying out balance numerical variance calculation on the light sensation color gradient balance data to obtain balance numerical variance data; carrying out multiple comparison analysis on the balance value variance data to obtain balance value difference multiple comparison data;
Step S252: performing balance value Bonferroni correction processing on the light-sensitive color gradient balance data according to the balance value difference multiple comparison data to obtain light-sensitive color gradient balance correction data;
Step S253: performing fractional regression analysis on the light sensation color gradient balance data according to the light sensation color gradient balance correction data to obtain color gradient balance quantile data;
Step S254: and carrying out quantile regression mean square error calculation according to the color gradient balance quantile data to obtain balance quantile error data.
7. An image evaluation system of a printed publication, characterized in that it is used to perform the image evaluation method of a printed publication as claimed in claim 1, the image evaluation system of a printed publication comprising:
the image content environment sensing module is used for acquiring image data of the printed publication to obtain an image dataset of the printed publication; performing color pattern analysis on the printed publication image dataset to obtain publication image color pattern data; performing image content environment perception analysis according to the publication image color mode data to obtain image content environment perception data;
The image light sense color analysis module is used for carrying out content environment light sense analysis on the image content environment sensing data to obtain image content light sense data; performing image color granularity reduction degree analysis according to the image content light sensation data to obtain image light sensation color reduction degree data; performing light sensation color gradient balance analysis according to the image light sensation color reduction degree data to obtain light sensation color gradient balance data; performing fractional regression mean square error calculation on the light sensation color gradient balance data to obtain balance fractional error data;
the image light sensation color evaluation model construction module is used for carrying out error compensation on light sensation color gradual change balance data according to balance bit error data to obtain light sensation color gradual change balance compensation data; carrying out image light sensation color evaluation model construction on the light sensation color gradient balance compensation data based on a random forest algorithm to obtain an image light sensation color evaluation model;
the image printing reduction difficulty evaluation module is used for evaluating the image printing reduction difficulty according to the image light sensation color evaluation model to obtain image printing reduction difficulty evaluation data; and feeding back the image printing reduction difficulty evaluation data to the terminal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410824178.1A CN118396987B (en) | 2024-06-25 | 2024-06-25 | Image evaluation method and system for printed publications |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410824178.1A CN118396987B (en) | 2024-06-25 | 2024-06-25 | Image evaluation method and system for printed publications |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118396987A CN118396987A (en) | 2024-07-26 |
| CN118396987B true CN118396987B (en) | 2024-09-20 |
Family
ID=91992972
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410824178.1A Active CN118396987B (en) | 2024-06-25 | 2024-06-25 | Image evaluation method and system for printed publications |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118396987B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2012083349A1 (en) * | 2010-12-19 | 2012-06-28 | Darling Matthew Ross | System for integrated wound analysis |
| CN112200807A (en) * | 2020-12-07 | 2021-01-08 | 光谷技术有限公司 | Video quality diagnosis method and system |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP4290100B2 (en) * | 2003-09-29 | 2009-07-01 | キヤノン株式会社 | Imaging apparatus and control method thereof |
| US8229215B2 (en) * | 2007-12-03 | 2012-07-24 | Omnivision Technologies, Inc. | Image sensor apparatus and method for scene illuminant estimation |
| JP5814555B2 (en) * | 2011-02-02 | 2015-11-17 | キヤノン株式会社 | Image processing apparatus and image processing apparatus control method |
| CN111953955B (en) * | 2020-08-26 | 2022-01-04 | 维沃移动通信有限公司 | White balance compensation method and device and electronic equipment |
-
2024
- 2024-06-25 CN CN202410824178.1A patent/CN118396987B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2012083349A1 (en) * | 2010-12-19 | 2012-06-28 | Darling Matthew Ross | System for integrated wound analysis |
| CN112200807A (en) * | 2020-12-07 | 2021-01-08 | 光谷技术有限公司 | Video quality diagnosis method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118396987A (en) | 2024-07-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108090902B (en) | Non-reference image quality objective evaluation method based on multi-scale generation countermeasure network | |
| Korhonen | Two-level approach for no-reference consumer video quality assessment | |
| JP7006567B2 (en) | Shooting method and shooting equipment | |
| JP7078175B2 (en) | Inspection equipment and method | |
| CN113554629A (en) | Strip steel red rust defect detection method based on artificial intelligence | |
| CN112785560A (en) | Air tightness detection water body updating method and system based on artificial intelligence | |
| CN111507426A (en) | No-reference image quality grading evaluation method and device based on visual fusion characteristics | |
| CN118196439B (en) | Certificate photo color auditing method based on visual language model and multiple agents | |
| CN117314826A (en) | Performance detection method of display screen | |
| CN115601747A (en) | Method and system for calculating confluency of adherent cells | |
| CN108257117B (en) | Image exposure evaluation method and device | |
| CN117372435B (en) | Connector pin detection method based on image characteristics | |
| CN118329910A (en) | Printing paper defect detection system and method based on machine vision | |
| CN118396987B (en) | Image evaluation method and system for printed publications | |
| CN117495827A (en) | Image evaluation method based on skin detector | |
| CN116523912B (en) | Cleanliness detection system and method based on image recognition | |
| CN117392461A (en) | Silkworm cocoon picture classification method and system based on ResNet18 and attention mechanism | |
| CN113808079B (en) | Industrial product surface defect self-adaptive detection method based on deep learning model AGLNet | |
| CN115482227A (en) | Machine vision self-adaptive imaging environment adjusting method | |
| CN118015355B (en) | Intelligent classification method and system for construction waste | |
| CN116754467B (en) | Evaluation method for ageing performance of natural rubber | |
| CN118570208B (en) | Lens assembly defect identification method and system based on machine vision | |
| KR102554824B1 (en) | Plating thickness measurement system and method | |
| CN118411932B (en) | LED display screen brightness correction method and correction system thereof | |
| CN117577033B (en) | Screen defect detection method, system and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |