Graphic design optimization system based on AI

Abstract

The invention relates to the technical field of image processing, in particular to an AI-based graphic design optimization system, which comprises an information entropy distribution evaluation module, an information entropy distribution evaluation module, an information entropy evaluation module, an information entropy evaluation module, a feature evaluation module, a feature evaluation module, a feature evaluation module and a feature evaluation module, wherein the information entropy distribution evaluation module acquires information of each element in a design graph, calculates information entropy of a feature item of each element to acquire an information entropy value of the feature and judges visual importance difference among the elements; and sorting the elements according to information importance, and generating an importance sequence of the design elements. According to the method, the feature information of each element in the design graph is analyzed, the visual importance of the elements is measured by the information entropy, and the size, the position, the color and the like of the element information are quantified and sorted, so that the design element layout is ensured to have a clear visual level. Therefore, the priority of each design element is clearer, and the user can pay attention to the important elements first in the design work. Through harmony of color features such as hue, brightness, saturation and the like, optimized color matching is calculated, and the visual harmony of the design is improved.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to an AI-based graphic design optimization system. Background Art

[0002] The field of image processing technology analyzes, processes, transforms and enhances static or dynamic images to improve their quality or extract useful information. It includes the complete process from image acquisition, preprocessing, feature extraction, image segmentation to target detection and recognition, and can achieve multiple functions such as image enhancement, restoration, compression, segmentation and object detection.

[0003] Among them, the graphic design optimization system is a system used to assist and improve the efficiency and quality of art design. By optimizing various aspects of graphic design, such as color adjustment, layout optimization, coordination of visual elements, etc., it helps designers complete complex creative design tasks more quickly and improves the visual effects and artistic expression of design works, thereby improving the overall level of graphic design.

[0004] In the optimization of graphic design, the existing technology mainly improves the design efficiency and effect through color adjustment and layout optimization. However, when dealing with the importance level of elements, it usually only relies on the basic visual characteristics of the elements, which makes it difficult to accurately distinguish the importance level of each element in the design, and it is difficult for users to quickly focus on the key content when observing the design. In addition, the color coordination of the existing technology is mainly through simple hue and saturation adjustment, which makes the color scheme prone to color conflicts or poor coordination in complex designs, affecting the visual appeal of the design. At the same time, the existing technology's processing of edge blur and burrs is difficult to achieve pixel-level precision, which affects the smoothness and clarity of the graphics. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings in the prior art and to propose an AI-based graphic design optimization system.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: The AI-based graphic design optimization system includes:

[0007] The information entropy distribution evaluation module obtains the information of each element in the design graphics, calculates the information entropy of each element feature item to obtain the information entropy value of the feature, determines the visual importance difference between elements, sorts the elements according to the information importance, and generates the importance sequence of the design elements;

[0008] The color harmony optimization module analyzes the current color feature data of the element information in the importance sequence of the design elements, establishes a color relationship model, extracts the color coordination data in the model, adjusts the color matching of the elements using the color coordination data, and generates an optimized graphic color matching result;

[0009] The graphic boundary optimization module marks the boundaries in the design graphic based on the optimized graphic color matching result, integrates the associated element feature data into a boundary pixel feature matrix, constructs a relationship diagram representing the dependency of the graphic boundary conditions according to the association strength, and obtains the graphic boundary refinement result by moving pixel points;

[0010] The global composition adjustment module integrates boundary information, marks key nodes, analyzes the change rate of boundary curves and performs smoothing processing according to the graphic boundary refinement result, refines boundary details, and obtains a global composition optimization result.

[0011] As a further solution of the present invention, the step of obtaining the information entropy value of the feature is specifically:

[0012] Based on the element feature information in the graphic design, the text, icons and background elements of the graphic are identified, and the size, position, color and contrast are verified. By detecting the boundaries and contours of the elements, analyzing the color composition and brightness of the elements, and sorting and classifying them, an element feature data set is generated;

[0013] According to the element characteristic data set, the formula is adopted:

[0014] H(x)=-∑(P i log P i +αD i )

[0015] Calculate the information entropy H(x) to obtain the information entropy value of each element feature;

[0016] Among them, P i represents the probability of the i-th element, log is used to measure the discreteness of the element distribution, α is an adjustment parameter used to control the effect of visual distance on information entropy, D i Indicates the visual distance from the ith element to the center of the design graphic.

[0017] As a further solution of the present invention, the step of obtaining the order of elements according to the importance of information is specifically as follows:

[0018] According to the information entropy value information of each element feature, the information entropy value of the element in the current design is compared with the information entropy value of the similar element in the previous design image, the information amount and visual importance of the element in the current design are determined, and the visual importance difference analysis result is generated;

[0019] Based on the visual importance difference analysis results, the elements are sorted according to the size of the information entropy value, and the information entropy value of each element is compared and arranged in order from high to low to generate an importance sequence of the design elements.

[0020] As a further solution of the present invention, the step of acquiring the color relationship model is specifically as follows:

[0021] Based on the color feature data extracted from the importance sequence of the design elements, including hue, brightness and saturation, the color feature of each design element is analyzed, and the hue, brightness and saturation values ​​of each element are extracted and normalized to generate a color feature data set;

[0022] Based on the color feature data set, the formula is adopted:

[0023]

[0024] Calculate color coordination C d , get color coordination data;

[0025] Among them, n represents the number of element pairs that need to be compared. and Represent the normalized hue values ​​of the i-th and j-th elements respectively, and Represent the normalized brightness values ​​of the i-th and j-th elements respectively, and Represent the normalized saturation values ​​of the i-th and j-th elements, respectively, w t 、w l and w s Represent the weights of hue, brightness and saturation respectively;

[0026] Based on the color coordination data, the relationships between all color features are sorted out, and the color coordination information between the elements in the graphic is presented in a visualized manner through a matrix. By analyzing the differences in hue, brightness and saturation of each pair of elements in the matrix, combined with the degree of color coordination, the color associations between the elements are identified and classified, and a color relationship model is established.

[0027] As a further solution of the present invention, the step of obtaining the color matching of the adjustment element is specifically as follows:

[0028] Based on the data in the color relationship model, extract the color coordination information of each design element, including hue, brightness and saturation data, identify the color-inharmonious parts in the graphics, adjust the hue, brightness and saturation, and generate a color matching result;

[0029] According to the color matching result, the adjusted color combination is applied to the corresponding elements of the design graphic, and the hue, brightness and saturation of each element are updated to obtain an optimized graphic color matching result.

[0030] As a further solution of the present invention, the step of obtaining the relationship graph representing the dependency of the graphic boundary conditions is specifically as follows:

[0031] Based on the optimized graphic color matching result, the boundaries in the design graphic are marked, the initial positions of the boundaries of the elements in the graphic are identified, and the color, brightness and texture features of each pixel are extracted by scanning the pixel information of each design element point by point, and a boundary pixel feature matrix is ​​established;

[0032] According to the information of the boundary pixel feature matrix, the formula is adopted:

[0033]

[0034] Calculate the correlation strength R(i′,j′) between pixel points i′ and j′ to obtain the correlation information of the pixel pair;

[0035] Among them, A i′ and A j′ Represent the color values ​​of pixel i′ and j′ respectively, B i′ and B j′ Represents the brightness value F of pixel i′ and j′ respectively i′ and F j′ Represent the texture features of pixel i′ and j′ respectively, and N represents the total number of neighboring pixels;

[0036] According to the association information of the pixel pairs, each pixel point is taken as a node in turn, and the pixel pairs with significant association strength are represented by edge connections, and the weights of the edges are assigned according to the strength differences to construct a relationship diagram of the graphic boundary condition dependence.

[0037] As a further solution of the present invention, the step of obtaining the graphic boundary thinning result is specifically as follows:

[0038] According to the data of the relationship diagram of the graphical boundary condition dependency, the formula is adopted:

[0039]

[0040] Calculate the position adjustment amount Δp to obtain the position adjustment amount data of the boundary pixel point;

[0041] Among them, k(i′,j′) is the relative distance between pixel points i′ and j′;

[0042] Based on the position adjustment amount data of the boundary pixel points, the adjustment amount is used to filter the boundary pixel points. By capturing the coordinates of the pixel points and moving the positions unit by unit, the color and brightness of the moved pixel points are detected in real time to be consistent with the features of the adjacent pixels, so as to obtain the graphic boundary refinement result.

[0043] As a further solution of the present invention, the step of obtaining the global composition optimization result is specifically as follows:

[0044] According to the graphic boundary refinement result, the boundary information is integrated, and the boundary intersection points, edge sharp turning points and geometric shape connection points are analyzed and marked by scanning the boundary area pixel by pixel to obtain the key node marking result;

[0045] According to the key node labeling results, the formula is used:

[0046]

[0047] Calculate the change rate Δq of the boundary curve to obtain the change rate data;

[0048] Where M is the total number of boundary nodes, X m is the position of the mth boundary node, X m-1 is the position of the m-1th node;

[0049] Based on the change rate data, the boundary curve is gradually adjusted, and the abrupt changes of the curve are eliminated by optimizing the curvature distribution between key nodes, and the local area is gradually narrowed down for detail smoothing to obtain a global composition optimization result.

[0050] Compared with the prior art, the advantages and positive effects of the present invention are:

[0051] In the present invention, by analyzing the characteristic information of each element in the design graphics, the visual importance of the elements is measured by information entropy, and by quantifying and sorting the size, position, color, etc. of the element information, it is ensured that the design element layout has a clear visual hierarchy. The priority of each design element is made clearer, and users can also focus on important elements first in the design work. By reconciling color features such as hue, brightness, and saturation, the optimized color matching is calculated to enhance the visual harmony of the design. By extracting the associated features and calculating the association strength between pixels, the boundary is optimized to the pixel level, the edge blur and burrs are avoided, the boundary is ensured to be clear and smooth, and the graphics are more holistic and visually beautiful. Further, by smooth adjustment of the overall composition and the marking and position fine-tuning of key nodes, a smooth transition from a large range to local details is achieved, the continuity of the graphic structure is maintained, and the design works are more consistent and professional in global and local visual effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 is a system flow chart of the present invention;

[0053] Figure 2 A flow chart for obtaining information entropy values ​​of features for the present invention;

[0054] Figure 3 A flow chart of sorting elements according to information importance according to the present invention;

[0055] Figure 4 A flow chart for establishing a color interrelationship model for the present invention;

[0056] Figure 5 A flow chart for adjusting the color matching of elements for the present invention;

[0057] Figure 6 A flow chart for constructing a relationship diagram representing the dependency of graphical boundary conditions for the present invention;

[0058] Figure 7 A flow chart of obtaining the result of graphic boundary refinement according to the present invention;

[0059] Figure 8 The present invention is a flow chart for obtaining global composition optimization results. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0061] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0062] See also Figure 1 , the AI-based graphic design optimization system includes:

[0063] The information entropy distribution evaluation module obtains the information of each element in the design graphics, extracts the size, position, color, contrast and other features of text, icons and background elements, calculates the information entropy of each feature item to obtain the information entropy value of the feature, and compares the information entropy values ​​of each element to determine the visual importance difference between elements, sorts the elements according to the information importance, and generates the importance sequence of design elements;

[0064] The color harmony optimization module analyzes the current color feature data of the element information in the importance sequence of the design elements. The color features include hue, brightness, and saturation. The hue, brightness, and saturation values ​​of each element color are obtained respectively. By calculating the coordination between multiple color features, a color relationship model is established, and the color coordination data in the model is extracted. The color matching of the elements is adjusted using the color coordination data to generate an optimized graphic color matching result.

[0065] The graphic boundary optimization module marks the boundaries in the design graphics based on the optimized graphic color matching results, identifies the initial position of the boundaries, and integrates the associated feature data into a boundary pixel feature matrix based on the color, brightness, and texture features extracted from the importance sequence of the design elements. The similarity of the color, brightness, and texture information of the adjacent pixels is counted to calculate the correlation strength between the pixels. A relationship diagram representing the dependence of the graphic boundary conditions is constructed based on the correlation strength. The similarity relationship between the pixel points and the adjacent pixels in the relationship diagram of the graphic boundary condition dependence is referred to to calculate the position adjustment amount of the boundary pixel points, and the pixels with larger position adjustment amounts are moved to ensure smooth and continuous boundaries, eliminate burrs and blurring on the edges, and obtain the graphic boundary refinement results.

[0066] The global composition adjustment module integrates boundary information and marks key nodes based on the result of graphic boundary refinement. Key nodes are boundary intersection points, sharp edge turning points, and geometric shape connection points. By calculating the position change of boundary nodes, the change rate of boundary curves is analyzed and smoothing is performed. By adjusting the boundary curves in a large range to make them smoother as a whole, and then gradually narrowing them to local areas, the boundary details are refined, and the large-scale and smooth local details are integrated to ensure the continuity of the overall boundary, and the global composition optimization results are obtained;

[0067] The importance sequence of design elements includes the sorting results generated according to the differences in the visual importance of the elements, the specific information entropy value corresponding to each element, and the feature weight allocation used to compare the importance of each element. The optimized graphic color matching results include the adjusted hue and brightness combination, the optimized color saturation parameters, and the color scheme that achieves color coordination among different elements. The graphic boundary refinement results include the readjusted boundary pixel position, the color and texture feature matrix of the boundary area, and the smooth edge effect after repairing the boundary blur or burr phenomenon. The global composition optimization results include the smoothed overall boundary curve, the optimized boundary shape changes in a large range, and the optimized local boundary details after integrating key nodes.

[0068] See also Figure 2 , the specific steps for obtaining the information entropy value of the feature are:

[0069] Based on the element feature information in the graphic design, the text, icons and background elements of the graphic are identified, and the size, position, color and contrast are confirmed. By detecting the boundaries and contours of the elements, analyzing the color composition and brightness of the elements, and sorting and classifying them, an element feature data set is generated;

[0070] The convolutional neural network is used to automatically identify and extract elements in the graphics, and determine the element type (such as text, icon, background) and position, size, color and other features. First, the graphics are segmented by CNN to identify the boundaries and outlines of each element, and then the color composition and brightness of the elements are analyzed. After sorting and classification, the element feature data set is generated. In the specific operation, the convolutional neural network first segments and classifies the graphic image, and automatically identifies elements of different categories. The recognition of size and position is achieved through the feature mapping of CNN: the network detects the pixel area of ​​the image, outputs the bounding box of each element, and determines the area parameter based on the pixel ratio, and marks the position attribute through the coordinate system of the image. Color information can be automatically obtained through the color space conversion of the CNN convolution layer, and the color characteristics are represented by the average color value or main hue value of each element. Using the color analysis method and the feature extraction in the convolution kernel, the color feature data of the element can be accurately analyzed. The contrast is obtained by analyzing the color difference of adjacent elements. CNN can automatically detect the brightness change and characterize the contrast by the gray value distribution. At the same time, combined with the edge detection method, CNN is used to automatically extract the contour features of the element, and further generate a feature matrix through the contour point information. After obtaining various characteristic parameters, they are integrated into the element characteristic data set.

[0071] According to the element feature data set, the formula is used:

[0072] H(x)=-∑(P i log P i +αD i )

[0073] Calculate the information entropy H(x) to obtain the information entropy value of each element feature;

[0074] Among them, H(x) is used to measure the information content of each element feature in the design figure. The higher the entropy value, the more uniform the distribution of elements in the figure. i represents the probability of the i-th element. The probability value can be calculated by tools such as ImageJ or GIMP to obtain the area proportion of each element, that is, the probability P of each element. iis the ratio of the pixel area of ​​the element to the total pixel area. The corresponding probability values ​​can be obtained by calculating the areas of elements such as text, icons, background, lines, and colors. Log is a logarithm with a base of 2, which is used to measure the discreteness of element distribution. α is an adjustment parameter used to control the impact of visual distance on information entropy. The optimal value is usually determined through design tests (such as A / B tests). For example, in graphic optimization of art design, A / B tests can be performed by showing two different design versions (version A and version B) to the target user group. The layout, color, size and other design factors of the graphic elements in each version are slightly different. By recording the user's interaction with the two designs (such as click-through rate, dwell time, visual preference survey, etc.), it can be analyzed which version of the design can better attract the user's attention and improve the user experience. Based on these user feedback data, the optimal value of each parameter in the design (such as the visual distance adjustment parameter α) is determined. i Represents the visual distance from the i-th element to the center of the design graphic. The coordinates of each element can be obtained through the OpenCV tool and the Euclidean distance formula is used: Calculate the visual distance, x′ i and i represents the horizontal and vertical coordinates of the i-th element, x′ center and center Represents the horizontal and vertical coordinates of the center point of the figure.

[0075] If the probability of each element (P i ) is: text 0.2, icon 0.25, background 0.15, line 0.3, color 0.1, and the distance between each element and the geometric center of the graphic (D i ) are: text 0.8, icon 0.7, background 0.6, line 0.9, color 0.5 (the distance value is normalized by dividing the maximum distance of all elements in the design graphic), and α is set to 0.5. Substitute the above values ​​into the formula for calculation:

[0076] H(x)=-(0.2log20.2+0.5·0.8)+(0.25log20.25+0.5·0.7)

[0077] +(0.15log20.15+0.5·0.6)

[0078] +(0.3log20.3+0.5·0.9)+(0.1log20.1+0.5·0.5)

[0079] H(x)

[0080] =-

[0081] (-0.464+0.4-0.5+0.35-0.411+0.3-0.521+0.45-0.332+0.25)

[0082] =0.522

[0083] The calculated result is 0.522, which indicates that the amount of information in the design graphics is concentrated on certain elements to a certain extent.

[0084] See also Figure 3 , the steps to obtain the elements according to the importance of information are as follows:

[0085] According to the information entropy value of each element feature, the information entropy value of the element in the current design is compared with the information entropy value of the same element in the previous design image, the information content and visual importance of the element in the current design are judged, and the visual importance difference analysis results are generated;

[0086] First, extract the characteristic information of each element, including area, color, and position, etc. The information entropy value H(x) is 0.522. Extract the characteristic information in the historical version design through a database such as Git, retrieve and match the historical entropy values ​​of similar elements, and if the entropy value range in the historical data is 0.45 to 0.55. Then, use statistical analysis tools (such as pandas or MATLAB in Python) for comparative analysis, and evaluate the difference between the entropy value of the current element and the historical value. The results show that H(x) is in a higher range, indicating that the element in the current design has a higher amount of information and visual importance. The higher entropy value indicates that the element is more visually prominent than similar historical elements.

[0087] Based on the results of the visual importance difference analysis, the elements are sorted according to the size of the information entropy value. By comparing the information entropy value of each element, they are arranged from high to low to generate the importance sequence of the design elements.

[0088] First, input the entropy data of all elements, and use Python's pandas library or MATLAB's sorting algorithm to automatically sort the entropy values ​​from high to low. For example, if the current element H(x) is 0.522, the entropy values ​​of other elements are 0.475, 0.49, and 0.51. The sorting tool will sort the priority of each element in descending order according to the size of the entropy value. Next, use design optimization software (such as AdobeXD or Figma) to adjust the layout of elements in the design drawing, sort the results according to the importance of the elements, and adjust the elements with large amounts of information to the visual core area or a more prominent position. Through methods such as grid layout optimization and automatic hierarchical arrangement, ensure that elements with high entropy values ​​get a better display position, and generate a design element importance sequence, which is used to guide subsequent design layout optimization.

[0089] See also Figure 4 , the acquisition steps for establishing the color relationship model are as follows:

[0090] Based on the color feature data extracted from the importance sequence of the design elements, including hue, brightness and saturation, the color features of each design element are analyzed, and the hue, brightness and saturation values ​​of each element are extracted and normalized to generate a color feature data set;

[0091] First, the current color feature data of each element needs to be analyzed. The extracted color features include hue, brightness, and saturation. The color data of each element is obtained one by one through image processing tools. Hue is expressed as the angle of the color wheel, usually ranging from 0° to 360°, while brightness and saturation are expressed as percentages, ranging from 0% to 100% respectively. Using the color extraction function of these tools, the hue, brightness, and saturation values ​​of each element can be accurately obtained. For example, the hue of an element may be 150°, the brightness is 60%, and the saturation is 80%. These color feature values ​​are saved and finally organized into a color feature data set.

[0092] Based on the color feature data set, the formula is used:

[0093]

[0094] Calculate color coordination C d , get color coordination data;

[0095] Among them, C d The smaller the value, the higher the coordination between colors. n represents the number of pairs of elements that need to be compared, that is, the number of times the elements in the design are compared in pairs. and Represent the normalized hue values ​​of the i-th and j-th elements respectively, and Represent the normalized brightness values ​​of the i-th and j-th elements respectively, and Represent the normalized saturation values ​​of the i-th and j-th elements, respectively, w t 、w l and w s Represents the weights of hue, brightness, and saturation respectively. The weights are adjusted according to design requirements. For example, if the coordination of hue is more important, w t The value of will be set higher and can be determined through design experience or user interface design guidelines.

[0096] If there are two design elements, the hue of the first element is 150°, the brightness is 60%, and the saturation is 80%; the hue of the second element is 180°, the brightness is 50%, and the saturation is 70%. Set the weight w t =0.4, w l =0.3, w s =0.3.

[0097] Hue Normalization:

[0098] Brightness Normalization:

[0099] Saturation Normalization:

[0100] Enter the formula to calculate:

[0101]

[0102] The result is C d =0.09, indicating that the color coordination between the two design elements is high (the value is close to 0).

[0103] Based on the color coordination data, the relationship between all color features is sorted out, and the color coordination information between elements in the graphic is presented in a matrix visualization manner. By analyzing the differences in hue, brightness and saturation of each element pair in the matrix, combined with the degree of color coordination, the color association between elements is identified and classified, and a color relationship model is established;

[0104] Based on the coordination degree C between multiple color features d The color relationship model is established by analyzing the differences in hue, brightness, saturation and other characteristics between elements and combining C d The calculated color coordination establishes the corresponding relationship between different elements. The specific process is as follows: First, the color feature data (such as hue, brightness, saturation) of all design elements are collected, and the color coordination C of each element pair is obtained through the aforementioned calculation process. d These C d The value reflects the coordination of different elements in terms of color characteristics. Then, during the modeling process, the system will d The values ​​establish a corresponding relationship between the colors of each element. For example, the lower C d A value close to 0 indicates a close harmony between the two elements, while a higher C d The values ​​indicate that there is a big difference between the two colors. Next, use these C d Values, through the model to group or classify the relationship between elements. For example, the system will be based on C dThe color values ​​are divided into different levels of color relationships: elements with higher color coordination are assigned to the "high coordination group", while elements with lower coordination are assigned to the "low coordination group". These levels can be divided into design tools (such as Figma, Sketch or AdobeXD). In the specific implementation process, the model will be based on C d The value is the main basis for judgment. By setting a reasonable threshold (such as C below 0.1 d Represents a high coordination relationship, C above 0.2 d represents low coordination relationship), thereby classifying elements with different color characteristics and forming a color relationship model that can be used for design optimization.

[0105] See also Figure 5 , the specific steps for obtaining the color matching of the adjusted elements are:

[0106] Based on the data in the color relationship model, the color coordination information of each design element is extracted, including hue, brightness and saturation data, the color inharmonious parts of the graphics are identified, the hue, brightness and saturation are adjusted, and the color matching results are generated;

[0107] First, extract the color coordination information of each design element. Specifically, use Figma or Adobe Color tools to obtain the coordination data of hue, brightness, and saturation, identify the parts of the design graphics with inconsistent colors, and obtain a list of color elements that need to be adjusted. Then modify the discordant colors and improve color matching by adjusting hue, brightness, and saturation. For example, Adobe Color can adjust the hue through the color wheel, while Figma can directly modify the brightness and saturation parameters in the color style. After the adjustment is completed, further analyze the coordination of the modified colors to ensure that the color matching of all elements is more in line with the overall design requirements and generate color matching results.

[0108] According to the color matching results, the adjusted color combination is applied to the corresponding elements of the design graphics, and the hue, brightness and saturation of each element are updated to obtain the optimized graphic color matching results;

[0109] Use design tools such as AdobeXD or Sketch to apply the optimized color combination to the entire design graphic. First, these tools will apply the adjusted hue, brightness, and saturation values ​​to the relevant elements in the design one by one to ensure that the color features are updated synchronously. In the specific work, AdobeXD applies the adjusted color features in batches through the color matching function, and Sketch uses the style library to update the colors of all elements in the design. Subsequently, the design tools will also evaluate the color transitions and visual contrast in the graphics to ensure that the optimized color matching can achieve a visually harmonious and unified effect. Finally, the optimized graphic color matching results are comprehensively processed to improve the overall color consistency and aesthetics.

[0110] See also Figure 6 , the specific steps for constructing a relationship diagram representing the dependency of the boundary conditions of a graph are:

[0111] Based on the optimized graphic color matching results, the boundaries in the design graphics are marked, the initial positions of the boundaries of the elements in the graphics are identified, and the color, brightness and texture features of each pixel are extracted by scanning the pixel information of each design element point by point, and the boundary pixel feature matrix is ​​established;

[0112] First, mark the boundaries in the design graphics. The system identifies the initial position of the boundaries of each design element in the graphics. When executed, the system extracts the color, brightness and texture features of each element according to the importance sequence of the design elements. The specific method is to scan the entire graphic point by point in turn, and use AdobePhotoshop or AffinityPhoto tools to automatically extract the color (RGB value), brightness (L value, based on HSL or Lab model) and texture (which can be obtained through OpenCV-based texture analysis) of each pixel using its pixel analysis function. These extracted data will be integrated into a boundary pixel feature matrix, and each cell of the matrix represents the color, brightness and texture information of a pixel point. This feature matrix provides an accurate data basis for subsequent analysis and boundary optimization by recording the eigenvalues ​​of each pixel. After the matrix is ​​generated, the system will perform basic verification to ensure the integrity and consistency of pixel feature extraction.

[0113] According to the information of the boundary pixel feature matrix, the formula is used:

[0114]

[0115] Calculate the correlation strength R(i′,j′) between pixel points i′ and j′ to obtain the correlation information of the pixel pair;

[0116] The higher the value of R(i′,j′), the greater the difference in the characteristics of the two pixels. i′ and A j′Respectively represent the color values ​​of pixel i′ and j′, obtained through image processing tools, B i′ and B j′ Represent the brightness values ​​of pixel i′ and j′ respectively, obtained through image processing tools, F i′ and F j′ Respectively represent the texture features of pixel i′ and j′, obtained through image processing tools, |A i′ -A j′ |、|B i′ -B j′ |、|F i′ -F j′ | represents the absolute difference in color, brightness and texture between two pixels, respectively, and is used to quantify the feature differences between pixels. N represents the total number of neighboring pixels, which is determined by the neighborhood definition (such as 3x3 neighborhood).

[0117] If the color difference between pixel i' and j' is 15, the brightness difference is 20, the texture difference is 10, and the total number of neighboring pixels N = 3, substitute the formula and calculate as follows:

[0118]

[0119] The result shows that the correlation strength between pixels i′ and j′ is 15.

[0120] According to the association information of pixel pairs, each pixel point is taken as a node in turn, and the pixel pairs with significant association strength are represented by edge connections. The weight of the edge is assigned according to the strength difference, and a relationship diagram of the dependence of the boundary conditions of the graph is constructed;

[0121] According to the correlation strength R(i′,j′) between the pixels calculated above, a relationship graph representing the dependence of the boundary conditions of the graph is constructed. First, the R(i′,j′) value of each pair of adjacent pixels is compared with the preset benchmark value or historical correlation strength. The system sets a reasonable range of correlation strength based on previous historical data, usually between 10 and 20. For the pixel correlation strength R(i′,j′)=15 calculated above, the system compares it with the benchmark value interval and finds that 15 is within the benchmark range, which is determined to be a high correlation, indicating that the two pixels have similar features and are suitable for connection. When constructing the relationship graph, the system regards all pixels as nodes in the graph and determines the connection method based on the correlation strength. For pixel pairs with high correlation (such as correlation strength between 10 and 20), the system connects them with edges and assigns stronger edge weights, reflecting higher feature consistency; for pixel pairs with correlation strength higher than this interval, the system determines them as low correlation and presents them with weaker edge weights or disconnection. In the process of graph composition, tools such as NetworkX or GraphTool are used to construct a boundary condition dependency graph, and nodes with high correlation in the graph are aggregated into boundary areas in a densely connected manner, while nodes with low correlation form relatively sparse areas. The resulting relationship graph provides graph structure support for boundary smoothing, which helps the system identify and adjust boundary details in the graph to ensure smoothness and consistency of the boundary.

[0122] See also Figure 7 , the specific steps for obtaining the graphics boundary refinement results are:

[0123] According to the data of the relationship diagram of the graphical boundary condition dependence, the formula is adopted:

[0124]

[0125] Calculate the position adjustment amount Δp to obtain the position adjustment amount data of the boundary pixel point;

[0126] Among them, Δp is used to indicate that the position of the boundary pixel points needs to be adjusted to improve the smoothness of the boundary, k(i′,j′) is the relative distance between pixel points i′ and j′, and the distance value can be calculated by NumPy or MATLAB.

[0127] If in the neighborhood of the current pixel, the correlation strengths of the three neighboring pixels with the central pixel are 15, 10, and 12, respectively, and the distances are 2, 3, and 1, respectively. Substituting these data into the formula, the position adjustment is calculated as follows:

[0128] Δp=(15·2)+(10·3)+(12·1)=30+30+12=72

[0129] The results show that the position adjustment amount is 72.

[0130] Based on the position adjustment amount data of the boundary pixel points, the boundary pixel points are screened by the adjustment amount. By capturing the coordinates of the pixel points and moving the positions unit by unit, the color and brightness of the moved pixel points are detected in real time with the feature consistency of the adjacent pixels to obtain the graphic boundary refinement result;

[0131] Based on the calculated position adjustment amount of 72 for the boundary pixel points, the system moves the position of the boundary pixel points with a larger position adjustment amount. In specific operations, the system first scans the adjustment amount data of all boundary pixel points, compares the adjustment amount of each pixel point one by one, and assumes that the reference value is 60. The pixel points with a position adjustment amount of 72 exceed the reference value and are marked as target points that require position optimization. After the system locks these target points, it gradually moves their positions in single-pixel units. During the movement process, the system detects and records the color and brightness values ​​of the pixels after the movement, and compares them with the features of the adjacent pixels to ensure the coordination of the features between adjacent pixels. If the feature comparison shows that the difference between the boundary pixels and the adjacent pixels is reduced, the system continues to fine-tune until the adjustment amount reaches the reference range. By adjusting these boundary pixels one by one, the system eliminates the burrs and blurred areas at the boundaries of the graphics one by one, ensuring a visually smooth and continuous boundary refinement result.

[0132] See also Figure 8 , the steps for obtaining the global composition optimization results are as follows:

[0133] According to the result of the graphic boundary refinement, the boundary information is integrated, and the boundary intersection points, edge sharp turning points and geometric shape connection points are analyzed and marked by scanning the boundary area pixel by pixel to obtain the key node marking results;

[0134] Through image processing methods, the geometric features of boundary pixels are extracted one by one, and boundary intersections, sharp edge turning points, and geometric connection points are marked as key nodes. When marking boundary intersections, the system identifies points where multiple directions intersect by calculating the directional differences of adjacent pixels and uses them as boundary intersections. When marking sharp edge turning points, the system obtains the curvature value of each pixel, and uses the curvature analysis function of OpenCV to mark pixels with curvatures higher than the specified threshold to confirm that they are sharp turning points. For the marking of geometric connection points, the system analyzes the angle changes of adjacent pixels and marks the connection areas where the angles change suddenly. After the entire marking process is completed, the system integrates the annotation data of these key nodes into the boundary information, providing basic data for boundary smoothing processing and ensuring the accuracy and consistency of node annotation.

[0135] According to the key node marking results, the formula is used:

[0136]

[0137] Calculate the change rate Δq of the boundary curve to obtain the change rate data;

[0138] Among them, Δq is used to quantify the smoothness of the boundary curve. The higher the value, the more drastic the curve change is, and further smoothing is needed. M is the total number of boundary nodes, which indicates how many nodes there are on the curve. It is usually obtained by analyzing the number of nodes in the boundary refinement result. Image processing software such as MATLAB or OpenCV library in Python is used to count nodes and extract coordinates. X m is the position of the mth boundary node, usually represented as a coordinate point (x m ,y m ), extracted from the boundary refinement result of the graphics, obtained through image processing tools such as OpenCV, X m-1 is the position of the m-1th node. The distance between each pair of adjacent nodes is calculated by the difference in node coordinates. m -X m-1 | represents the distance between adjacent boundary nodes, which can be calculated using computing tools such as NumPy or MATLAB.

[0139] If the current boundary curve contains 5 consecutive nodes, their position coordinates are (1,1), (2,3), (4,5), (5,8), (7,10) in sequence. The change values ​​of adjacent nodes are calculated as follows:

[0140]

[0141] Sum and then average:

[0142]

[0143] The results show that the curve change rate is 2.77.

[0144] Based on the rate of change data, the boundary curve is gradually adjusted to eliminate abrupt changes in the curve by optimizing the curvature distribution between key nodes, and the details are gradually narrowed down to the local area for smoothing to obtain the global composition optimization result;

[0145] Through the analysis of the above curve change rate Δq, the system recognizes that the current Δq value is 2.77. Assuming the baseline value is 2, it shows a larger curvature change compared with the baseline value 2, indicating that the curve has a more prominent curvature fluctuation in the local area. The system first performs a preliminary smoothing operation on the large-scale boundary curve. During the adjustment process, the key node positions are trimmed one by one to reduce the curvature change and match the curvature values ​​between adjacent nodes. In the specific operation, the system translates the key nodes in sequence, continuously fine-tunes the areas with higher Δq values, gradually reduces the curvature change, and ensures that the adjusted curvature is more uniform. As the large-scale smoothing process proceeds, the system gradually narrows the smoothing range to the local area, and judges the curvature improvement by re-analyzing the local Δq value. The remaining areas with large curvature fluctuations are further refined to ensure the connection between the local area and the overall curve. After completing the large-scale and smooth local detail integration, the system finally obtains a boundary curve with consistent curvature, ensuring the continuity and smoothing effect of the global composition, and obtaining the global composition optimization result.

[0146] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. AI-based graphic design optimization system, characterized by: The system comprises: The information entropy distribution evaluation module obtains the information of each element in the design graphics, calculates the information entropy of each element feature item to obtain the information entropy value of the feature, determines the visual importance difference between elements, sorts the elements according to the information importance, and generates the importance sequence of the design elements; The color harmony optimization module analyzes the current color feature data of the element information in the importance sequence of the design elements, establishes a color relationship model, extracts the color coordination data in the model, adjusts the color matching of the elements using the color coordination data, and generates an optimized graphic color matching result; The graphic boundary optimization module marks the boundaries in the design graphic based on the optimized graphic color matching result, integrates the associated element feature data into a boundary pixel feature matrix, constructs a relationship diagram representing the dependency of the graphic boundary conditions according to the association strength, and obtains the graphic boundary refinement result by moving pixel points; The global composition adjustment module integrates boundary information, marks key nodes, analyzes the change rate of boundary curves and performs smoothing processing according to the graphic boundary refinement result, refines boundary details, and obtains a global composition optimization result.

2. The AI-based graphic design optimization system according to claim 1, characterized in that: The steps for obtaining the information entropy value of the feature are specifically as follows: Based on the element feature information in the graphic design, the text, icons and background elements of the graphic are identified, and the size, position, color and contrast are verified. By detecting the boundaries and contours of the elements, analyzing the color composition and brightness of the elements, and sorting and classifying them, an element feature data set is generated; According to the element characteristic data set, the formula is adopted: H(x)=-∑(P i log P i +αD i ) Calculate the information entropy H(x) to obtain the information entropy value of each element feature; Among them, P i represents the probability of the i-th element, log is used to measure the discreteness of the element distribution, α is an adjustment parameter used to control the effect of visual distance on information entropy, D i Indicates the visual distance from the ith element to the center of the design graphic.

3. The AI-based graphic design optimization system according to claim 2, characterized in that: The steps of obtaining the order of elements according to the importance of information are specifically as follows: According to the information entropy value information of each element feature, the information entropy value of the element in the current design is compared with the information entropy value of the similar element in the previous design image, the information amount and visual importance of the element in the current design are determined, and the visual importance difference analysis result is generated; Based on the visual importance difference analysis results, the elements are sorted according to the size of the information entropy value, and the information entropy value of each element is compared and arranged in order from high to low to generate an importance sequence of the design elements.

4. The AI-based graphic design optimization system according to claim 3, characterized in that: The acquisition steps of establishing the color relationship model are specifically as follows: Based on the color feature data extracted from the importance sequence of the design elements, including hue, brightness and saturation, the color feature of each design element is analyzed, and the hue, brightness and saturation values ​​of each element are extracted and normalized to generate a color feature data set; Based on the color feature data set, the formula is adopted: Calculate color coordination C d , get color coordination data; Among them, n represents the number of element pairs that need to be compared. and Represent the normalized hue values ​​of the i-th and j-th elements respectively, and Represent the normalized brightness values ​​of the i-th and j-th elements respectively, and Represent the normalized saturation values ​​of the i-th and j-th elements, respectively, w t 、w l and w s Represent the weights of hue, brightness and saturation respectively; Based on the color coordination data, the relationships between all color features are sorted out, and the color coordination information between the elements in the graphic is presented in a visualized manner through a matrix. By analyzing the differences in hue, brightness and saturation of each pair of elements in the matrix, combined with the degree of color coordination, the color associations between the elements are identified and classified, and a color relationship model is established.

5. The AI-based graphic design optimization system according to claim 4, characterized in that: The steps for obtaining the color matching of the adjustment element are specifically as follows: Based on the data in the color relationship model, extract the color coordination information of each design element, including hue, brightness and saturation data, identify the color-inharmonious parts in the graphics, adjust the hue, brightness and saturation, and generate a color matching result; According to the color matching result, the adjusted color combination is applied to the corresponding elements of the design graphic, and the hue, brightness and saturation of each element are updated to obtain an optimized graphic color matching result.

6. The AI-based graphic design optimization system according to claim 5, characterized in that: The steps of obtaining the relationship diagram representing the dependency of the graphic boundary conditions are specifically as follows: Based on the optimized graphic color matching result, the boundaries in the design graphic are marked, the initial positions of the boundaries of the elements in the graphic are identified, and the color, brightness and texture features of each pixel are extracted by scanning the pixel information of each design element point by point, and a boundary pixel feature matrix is ​​established; According to the information of the boundary pixel feature matrix, the formula is adopted: Calculate the correlation strength R(i′,j′) between pixel points i′ and j′ to obtain the correlation information of the pixel pair; Among them, A i′ and A j′ Represent the color values ​​of pixel i′ and j′ respectively, B i′ and B j′ Represents the brightness value F of pixel i′ and j′ respectively i′ and F j′ Represent the texture features of pixel i′ and j′ respectively, and N represents the total number of neighboring pixels; According to the association information of the pixel pairs, each pixel point is taken as a node in turn, and the pixel pairs with significant association strength are represented by edge connections, and the weights of the edges are assigned according to the strength differences to construct a relationship diagram of the graphic boundary condition dependence.

7. The AI-based graphic design optimization system according to claim 6, characterized in that: The steps for obtaining the graphic boundary refinement result are specifically as follows: According to the data of the relationship diagram of the graphical boundary condition dependency, the formula is adopted: Calculate the position adjustment amount Δp to obtain the position adjustment amount data of the boundary pixel point; Among them, k(i′,j′) is the relative distance between pixel points i′ and j′; Based on the position adjustment amount data of the boundary pixel points, the adjustment amount is used to filter the boundary pixel points. By capturing the coordinates of the pixel points and moving the positions unit by unit, the color and brightness of the moved pixel points are detected in real time to be consistent with the features of the adjacent pixels, so as to obtain the graphic boundary refinement result.

8. The AI-based graphic design optimization system according to claim 7, characterized in that: The steps for obtaining the global composition optimization result are specifically as follows: According to the graphic boundary refinement result, the boundary information is integrated, and the boundary intersection points, edge sharp turning points and geometric shape connection points are analyzed and marked by scanning the boundary area pixel by pixel to obtain the key node marking result; According to the key node labeling results, the formula is used: Calculate the change rate Δq of the boundary curve to obtain the change rate data; Where M is the total number of boundary nodes, X m is the position of the mth boundary node, X m-1 is the position of the m-1th node; Based on the change rate data, the boundary curve is gradually adjusted, and the abrupt changes of the curve are eliminated by optimizing the curvature distribution between key nodes, and the local area is gradually narrowed down for detail smoothing to obtain a global composition optimization result.

Images