Geometric analysis and model construction of the relationship between color contrast and light and shade in plane visual design

With the continuous development of various industries, all kinds of product sales and management as well as a variety of goods publicity and advertising need graphic design technology to complete. Excellent graphic design can attract more eyeballs, so that people pay attention to the graphic design works to display the goods, which is particularly fierce for commercial competition in today’s society, the development of various industries undoubtedly plays an important role in advertising [1-4]. In packaging design, poster design, corporate image design, store design, exhibition board design and other designs are inseparable from the excellent graphic design technology as a support [5-6]. In these graphic design can not be separated from the color harmony, appropriate color matching and embellishment can highlight the theme, attention, the appropriate application of color to graphic design is the most basic professional knowledge of each graphic designer to master [7-10].

Color is an important element in graphic design, the use of color contrast can reflect the characteristics of the work [11]. Color contrast allows graphic design to better reflect the power and concept of contrast, the reasonable use of color contrast creation techniques can attract the attention of the audience to a certain extent, the overall hierarchy of the work and the sense of image to enrich the work, can enhance the visual impact of the work to people [12-15]. The contrast between light and dark color is mainly divided into long tone and material contrast, in which the long tone contrast is to use the main color of the design work with high brightness color, and contrast with the lower brightness color [16]. The visual stimulation of the light and dark contrast design technique is quite strong and obvious, and the designed work has relatively high resolution and strong contrast, which can both convey appreciation and bring a sense of achievement [17-19]. In graphic design, reasonable color matching and application is a very profound study, which needs to accumulate experience in the actual application process, in order to achieve a good use of color in graphic design, it is crucial to analyze the relationship between color contrast and light and dark.

In this paper, we propose a probability density function estimation method to find the local optimal solution of the objective function in a probabilistic stochastic way with the help of the sampling idea of simulated annealing, so as to converge to the global optimal solution, and we propose a dictionary-type strategy to optimize the geometric feature estimation of the probability model. In order to deal with design problems in planar visual images, geometric feature modeling is performed in terms of both layout and element color arrangement. Color space conversion and luminance enhancement are carried out for planar visual images, and the histogram equalization function is constructed to realize the enhancement processing in the generation of antagonistic color enhancement network, forming the color adaptive adjustment of planar visual images. Analyze the color enhancement effect of graphic visual images and evaluate the design of graphic visual images.

2

The use of color language in graphic design

Color language is a symbol that conveys emotions and messages through color, which is a very important and charming art language. Color has the advantage of being verbal, so designers can use color language to not only create a real visual scene in a two-dimensional space, but also to convey the designer’s thoughts and feelings.

Graphic design is about creating and combining symbols, pictures and words in a variety of ways to create a visual representation that conveys an idea or message. The language of color can help to improve the ability to convey information and visual appeal of graphic design, and graphic design can also provide a platform for the language of color to be displayed, which are complementary and mutually reinforcing.

2.1

Color Language

Color is the reflection of different objects received by the naked eye, and the recognition of color is a subjective consciousness of the visual nerve sensation caused by the stimulation of the eyes by electromagnetic wave radiation.

Color is the most sensitive formal element that can cause people’s common aesthetic pleasure. Color is one of the most expressive elements because its nature directly affects people’s feelings.

The rich variety of colors can be divided into two broad categories of colorless and colored. Colored colors have three basic characteristics: hue, purity (also known as color, saturation), brightness, the three echo, common correlation. Hue for the color “appearance”, a variety of colors by the combination of different hue extension, is the main carrier of information transfer. Purity for the inheritance of color “gene”, the design can be combined with the design intent to use the purity of the contrast, contrast design effect. Brightness for the interpretation of color “mutation”, the design of the combination of color light and dark changes, to create a colorful atmosphere, so that the work has a visual impulse.

When people look at colors, they will have associations related to their daily lives, and then a series of psychological changes will also occur. This is the emotion of color, which is divided into the sense of warmth and coldness, nearness and distance, hardness and softness, and lightness, and these color emotions are respectively turned into language to convey different signals to the audience. For example, the warmth and coolness of colors give people the imagination of a specific scene. The sense of distance and proximity of color conveys the sense of space. The softness and hardness of colors give people the imagination of the material of the picture. The sense of lightness and weight of color gives people a sense of visual pleasure or depression. All in all, the rational use of color to convey the language of emotion is a common technique used by graphic designers.

2.2

The relationship between color language and graphic design

Essentially, the computer graphic design of color changes in the adjustment and control, it makes the finished product of graphic design color language expression is richer, more infectious and attractive, which makes the finished product of graphic design artistic effect can be greatly highlighted to play, to meet the different aesthetic needs of different consumers and users. Considering the color itself has a strong personality connotation, different people for different colors will reflect different preferences and psychological characteristics, so in the use of color for computer graphic design process must consider its visual function and aesthetic aesthetic form, to be in the design process will be a perfect combination of practicality and aesthetics, give full consideration to the materials, technology and design techniques of the application of a reasonable combination to ensure that the last Design a relatively complete, beautiful, artistic computer graphic design works. To sum up, computer graphic design and color language should be presented in a complementary relationship, color language in computer graphic design will be constantly proclaimed, penetration, and ultimately flow to other social space areas, presenting its more diversified practical functions.

2.3

Analysis of visual geometry principles in design

The overall benefit of geometric scale analysis is that it shows the true importance of science and math in design. Proportions are crucial to composition because they establish visual relationships not only in terms of length and width, but also in terms of the relationship between the elements that make up the piece and the whole. The viewer is seldom aware of specific proportions, but rather enjoys the harmony created by the proportions and the relationships between the elements.

Geometric analysis assumes that a system of proportions and reference lines constitutes the entire composition of a work of art, building, product, or graphic design. This analysis does not take into account conceptual, cultural, or media influences, but it can shed light on some of the rules of composition and explain the favorable response of visitors through quantitative measurements of proportion and calibration. The value of geometric analysis lies in its ability to reveal the design concepts and principles behind the work of artists, architects and designers. They are the key elements that determine the direction of a design and can reveal how the designer has placed these elements in a composition and provide insight into the designer’s choices. The process of geometric analysis is one of exploration, experimentation and discovery.

3

Modeling based on geometric features

3.1

Probability density function estimation

There exists a set of observation samples $\{x_{1}, x_{2}, \dots, x_{n}\}$ of a random variable $X$ , where $n$ is the number of sample points, and in the study of this paper $x_{i} (i = 1, 2, \dots, n)$ are all one-dimensional sample points. Assume that $X$ obeys a particular distribution and its probability density function is $p (x)$ . In real practical applications, the exact form of $p (x)$ is mostly unknown, yet numerous machine learning and pattern recognition algorithms are often based on sample probability density distributions. Examples include the nearest neighbor classifier and the plain Bayesian classifier. Therefore, estimating the unknown probability density function based on known samples plays a key role in improving the performance of learning algorithms. The task of probability density function estimation is to estimate the unknown probability density function $p (x)$ from a known sample $\{x_{1}, x_{2}, \dots, x_{n}\}$ according to some mechanism, assuming that the density function obtained from the estimation is $\bar{p} (x)$ , and the ultimate goal of the estimation is to minimize the error between $\bar{p} (x)$ and $p (x)$ [20-21].

For the estimation of the probability density function there are three commonly used forms: full-participant, no-participant, and half-participant estimation. Full parametric estimation means that the probability density function is of a known form which contains unknown parameters. Non-parametric means that the form of the probability density function is unknown. Semi-parametric estimation is an estimation method that combines the full-parametric and non-parametric methods, and falls somewhere in between. The classical nonparametric estimation method is the Parzen window method, or kernel method for short. The kernel density estimation method is described in detail below.

For a given observation $\{x_{1}, x_{2}, \dots, x_{n}\}$ of random variable $X$ , the kernel method provides an estimate of the unknown probability density function by means of the following equation (1): (1) $\hat{p} (x) = \frac{1}{n h} \sum_{j = 1}^{n} K (\frac{x - x_{j}}{h})$ where the kernel function $K (u)$ satisfies the following two conditions: $\int_{- \infty}^{+ \infty} K (u) d u = 1$ . holds for any $u$ , $K (u) \geq 0$ .

Parameter $h$ is called the window width, which satisfies $\lim_{n \to + \infty} h (n) = 0$ and $\lim_{n \to + \infty} [N \times h (n)] = \infty$ , $h$ as a function of the number $n$ of sample points. A brief explanation of the introduction of equation (1) will be given below. For a given observation $x$ of a random variable $X$ , the probability that it falls into the interval $[a, b]$ is: (2) $P = \int_{a}^{b} \bar{p} (x) d x$

When $|b - a| \cdot ε$ , $ε$ is a very small value, then Eq. (2) can be converted approximately: (3) $P = \int_{a}^{b} \hat{p} (x) d x \approx \hat{p} (x) \int_{a}^{b} d x = \hat{p} (x) (b - a)$

To wit: (4) $\hat{p} (x) = \frac{P}{b - a}$

Assuming that $k$ out of $n$ observations $\{x_{1}, x_{2}, \dots, x_{n}\}$ of the current random variable $X$ fall into interval $[a, b]$ , the probability that any observation of random variable $X$ falls into interval $[a, b]$ is: (5) $P = \frac{k}{n}$

The equation (5) can be obtained by bringing the equation (4) into (4): (6) $\hat{p} (x) = \frac{k}{n (b - a)}$

For segmented functions: (7) $K (\frac{x - x_{i}}{h}) = \{\begin{matrix} 1 & \frac{|x - x_{i}|}{h} \leq \frac{1}{2} \\ 0 & O t h e r \end{matrix}$

It indicates that when the observed value $x_{i}$ of the random variable $X$ falls within an interval $[x_{i} - \frac{h}{2}, x_{i} + \frac{h}{2}]$ centered at $x$ and radiused at $\frac{h}{2}$ , where $h$ is a very small value, the value of this function $K (\frac{x - x_{i}}{h})$ is 1. Thus, for the value of $k$ in expression (6) can be expressed by expression (8): (8) $k = \sum_{i = 1}^{n} K (\frac{x - x_{i}}{h})$

Together with the very small value of $|b - a|$ , the following equation (9) holds approximately: (9) $h = b - a$

Bringing equations (8) and (9) into expression (6) gives: (10) $\hat{p} (x) = \frac{1}{n h} \sum_{j = 1}^{π} K (\frac{x - x_{j}}{h})$

Where function $K (\frac{x - x_{i}}{h})$ is the window function, i.e. the kernel function, and $h$ is the window width. It was later shown that any function that satisfies the conditions can be used as a kernel function. The Gaussian kernel function $K (x) = \frac{1}{\sqrt{2 π}} \exp (- \frac{x^{2}}{2})$ is used as the preferred choice in many practical applications due to its continuous nature. The expression for $K (x)$ can be obtained by bringing it into Eq. (11): (11) $\bar{p} (x) = \frac{1}{n h} \sum_{j = 1}^{n} K (\frac{x - x_{j}}{h}) = \frac{1}{n h} \sum_{j = 1}^{n} [\frac{1}{\sqrt{2 π}} \exp (- \frac{{(x - x_{j})}^{2}}{2 h^{2}})]$

Eq. (11) is the expression for the probability density of random variable $X$ obtained by estimating on observation sample $\{x_{1}, x_{2}, \dots, x_{n}\}$ using the Parzen window method.

3.2

Probabilistic model optimization

3.2.1

Feature sampling based on simulated annealing

Simulated annealing algorithm is a kind of large-scale sampling algorithm for solving distribution simulation problems that are difficult to sample directly, and it is a stochastic optimization search algorithm based on Monte Carlo iterative solution strategy.

Drawing on the sampling idea of simulated annealing, when solving the optimal solution of each kernel density estimation function, the function results of density estimation v( ) and ( | ) can be used as the evaluation score. Therefore, the process of model sampling is to find the appropriate design features through continuous iterative search to make the evaluation score as large as possible. The core of the idea is to find the local optimal solution of the objective function in the solution space based on probabilistic stochastic way, which tends to the global optimal solution.

3.2.2

Model optimization based on dictionary strategy

A flat visual image often consists of multiple elements or multiple colors, and when optimizing a feature model based on probabilistic estimation, the optimization problem of multiple probability density functions is involved. To address this problem, this paper adopts a dictionary-based strategy to evaluate the sampled features in a linear fashion from each individual design element so that they satisfy the following optimization function: (12) $\max ({\hat{f}}_{1} (X_{1}), P_{2} (X_{2} | C_{2}), \dots)$ where ${\hat{f}}_{1} (x_{1})$ is the first optimization object and $x_{1}$ is the corresponding design feature. Through the sampling method, each evaluation function can obtain the optimal solution result, denoted as $f_{1}^{*}$ . Using this result as a reference, all the design features that satisfy the following conditions during the sampling process are retained: (13) $R = [X |{\hat{f}}_{1} (X_{1}) \geq (1 - \frac{δ}{100}) f_{1}^{*}], δ \in [0, 100]$ where $δ$ is an adjustment factor that determines the number of sampled features to be retained. $X$ is a design feature that satisfies the condition during sampling and is kept in set $R$ .

3.3

Geometric feature estimation for probabilistic models

A print advertising image often consists of a combination of several types of design elements, including images, graphics, text, and symbols. In each design element, one or more colors are included. These complex forms of element combinations and color matching methods all make the model construction of design features very difficult.

In order to simplify the difficulty of modeling, this paper focuses on solving design problems in print visual images.

Print advertising images often consist of several basic design elements, such as background, product images and text. Taking e-commerce print advertisement images as an example, these images often contain simple and clear layout styles, and more uniform and regular color combinations (e.g., the background color tone is consistent, and the text color tends to be uniform). In the study of graphic visual images of text, the feature model is mainly used to model such graphic advertisement images to solve the task of graphic advertisement image design for products.

3.3.1

Layout Geometry Feature Estimation

The layout problem of graphic visual images is essentially to determine the geometric features such as size and position of each design element within the context of the image. The purpose of constructing a model oriented to layout-related geometric features is to estimate the pattern of change in the position of different elements.

Unlike document-based images, flat visual images include text and image-type elements, which increase the complexity of model construction and the difficulty of solving.

This paper refers to the container-based geometric feature modeling method. In the modeling of elemental geometric features, researchers often consider the elements in an image that are of the same type and adjacent to each other as a whole, and call the whole a container.

In order to obtain the design features of each type of container in a flat visual image, this paper clusters the design elements in the image.

3.3.2

Estimation of color geometric features of elements

Since each element in a flat visual image contains multiple colors. In order to simplify the complexity of the model, the color design problem of a flat visual image is transformed into solving the main color features of the subject, background and copy in the image, and ensuring that the main color features are in harmony with each other.

On the basis of the primary color features obtained by probabilistic estimation, the algorithm can further color reconstruct the elements to complete the color design task of a flat visual image.

Although there is no lack of additional design elements in planar images, meeting the color matching requirements of the above three main types of elements can solve the vast majority of color design tasks for planar visual images.

Under the above assumptions, the color design problem of a print advertisement image can be expressed as follows: from the joint probability density distributions of the color features in the subject, background and copy elements, find the optimal sequence of features such that the density function $P (X_{P}, X_{B}, X_{T})$ scores the highest. Where $X_{P}, X_{B}, X_{T}$ represents the color features in the subject, background and copy elements respectively. As the color change to the subject in the image will affect the consumer’s interpretation of the real content of the print advertisement image. Therefore, the color design of print advertising images often does not involve the color reconstruction of the subject elements to avoid misleading consumers. Thus, the color design problem of an image is further represented as: finding the optimal background and copy main color for a specific product main color $X_{p}$ , which makes the density function $P (X_{B}, X_{T} |X_{P})$ score the highest. To this end, during the construction of the color geometric feature model, the algorithm filters the matching training data according to the main color of the subject in the image, so as to ensure that the color geometric features in the data satisfy the conditional parameters of the model.

4

Graphic visual image color enhancement model construction

4.1

Spatial Transformation and Brightness Enhancement

4.1.1

Color space conversion

The RGB color model consists of $R$ , $G$ and $B$ primary colors. Although the RGB model has a comfortable color display, the difference between two similar color values in the model is too large, resulting in incomplete consistency in calculations and high correlation between the three components of the modified model. If one of the component values is changed, the other two component values are also changed. Unlike the RCB model, HSV is a color model based on the visual system of the human eye.The HSV model has two main distinctive features. One is that the change in the value of the $V$ -component is not related to the overall color information of the image. The second is that the $H$ or $S$ components are the same as the color perceived by human eye vision. According to these two characteristics. It can be further known that HSV color space model is more suitable for color image processing algorithms than RGB model. Therefore, before processing an image, the color space of the image to be processed is first converted from the RGB model to the HSV model3.

Assuming that $(r, g, b)$ is the red, green, and blue coordinates of a color, to get the value of $(h, s, v)$ in HSV space, the conversion formula can be written as: (14) $v = \max \{r, g, b\}$ (15) $h = \{\begin{matrix} (g - b) \times 60 / s & v = r \\ 180 + (b - r) \times 60 / s & v = g \\ 240 + (r - g) \times 60 / s & v = b \end{matrix}$ (16) $s = \{\begin{array}{l} (v - \min \{r, g, b\}) \times 255 & v \neq 0 \\ 0 & v = 0 \end{array}$

When a value of $(h, s, v)$ is given, the corresponding $(r, g, b)$ tristimulus color is: (17) $(r, g, b) = (r' + λ, g' + λ, b' + λ)$

Of these, $λ = v - c$ .

4.1.2

Luminance Component Enhancement

Since the image needs to be preprocessed to obtain a brighter image compared to the original image before applying the homomorphic filtering algorithm for color enhancement, the global histogram and local histogram equalization methods are used to achieve a uniform distribution of gray values in the image, and then the dynamic range of the pixel values in the image is extended by using the gray scale mapping. However, the preprocessing process is prone to color distortion and the generation of new noise, and has the disadvantages of difficulty in achieving satisfactory results, excessive computation, and the formation of mosaic effects.

To solve the above problem, the image is chunked and then the homomorphic filter based on DCT transform is divided into each sub-image used. The image to be processed is divided into $8 \times 8$ sub-image and the image chunk is $8 \times 8$ 2D matrix $f (m, n)$ , where $0 \leq m \leq 7$ , $0 \leq n \leq 7$ , then the DCT transform of $f (m, n)$ is defined as: (18) $F (u, t) = \frac{1}{4} C (u) C (t) \sum_{n = 0}^{7} \sum_{n = 0}^{7} f (m, n)$ (19) $\cos \frac{(2 m + 1) u π}{16} = \cos \frac{(2 n + 1) t π}{16}$ where $0 \leq u \leq 7$ , $0 \leq t \leq 7$ . And: (20) $C (u) = C (t) = \{\begin{array}{l} \frac{1}{\sqrt{2}}, u = t = 0 \\ 0 \end{array}$

After color space conversion and luminance component enhancement the brighter images can be extracted, resulting in a more uniform distribution of image gray values.

4.2

Flat visual image equalization processing

In order to achieve color adaptive adjustment of planar visual images, the enhancement process is performed in generating an adversarial color enhancement network, and the resulting mapping formula for the histogram equalization function is constructed as: (21) $R (α) = (\sum_{k = 0}^{α} \frac{h (k)}{m}) (β - 1)$

Where, $α$ denotes the gray scale factor. $k$ denotes the gray level, $h (k)$ denotes the histogram component. $β$ denotes the number of gray levels. In the case of gray level localization, the above equation can be used for mapping process. When the inverse color enhancement network is generated, the perceptible limit is set. In the case where the amount of change is less than the degree value, the network cannot perceive the change in image brightness. To enhance the image details and avoid the glare illusion, the image gradient is set to a value consistent with the threshold value, from which the image gray level is calculated: (22) $s (α) = μ [R' (α)]$ where $R' (α)$ denotes the post-shear histogram mapping function. $μ$ denotes the threshold value. Set the color contrast enhancement adjustment scheme to enhance the local histogram based on the global histogram. Ensure the relative brightness of the local area by outputting light intensity to obtain accurate mapping effect: (23) $W (x, y) = \frac{1}{φ} \sum_{k = 1}^{m} [\exp (- \frac{{(x - x_{k})}^{2} + {(y - y_{k})}^{2}}{2}) α_{k}]$ where $φ$ denotes the normalization factor. $B$ denotes the region where the pixel is located, and $(x_{i}, y_{k})$ denotes the coordinates of the region where the pixel is located. According to the adaptive process, the equalization of planar visual images is realized.

4.3

Perturbing aggressive adversarial processing

Due to the presence of a large amount of noise in the target image, although most of the noise can be eliminated by the luminance enhancement process, there are still perturbation offensive factors in the image, which need to be subjected to adversarial processing. The dimensionality of the image features is reduced by the generated adversarial color enhancement network, and the dimensional feature information is expressed using the visual representation layer, and the adversarial loss is calculated by equation (24) [22].

The dimensional feature information, the adversarial loss is calculated through equation (24): (24) $ϑ = J \log σ (I) + J \log (1 - σ (I + σ (I)))$ where $J$ denotes the expectation. $σ (I)$ denotes the judgment function. Using this formula, attack samples and initial samples can be imported into the judgment function separately and the judgment results can be output, from which accurate adversarial loss data can be obtained. The loss function is utilized to iterate the network and update it in real time to improve the anti-interference ability of the system.

5

Graphic visual image color design analysis

5.1

Gain and Luminance Linearity Tests

In order to assist the effect of the color enhancement model proposed in this paper in the application of color adjustment system, the data collector hardware is designed and developed on the basis of Tiny6410 core board. It is known that the board integrates the S3C6410 processor, JTAG interface, reset circuit and NAND FLASH, etc., so the design of the hardware should focus on the peripheral circuits. SP3232 is used as a level conversion chip, so as to realize the hardware design of the plane visual image color adaptive adjustment system.

Test system to adjust the color process, adjust the white gain value after the screen brightness value, gain and brightness test data shown in Figure 1, W is the width of the plane visual image.

According to the test data in the figure, draw the gain and brightness change relationship curve. According to the curve change trend, it can be proved that there is a linear relationship between gain and luminance. Therefore, it verifies that the color enhancement model of this design can be used to adjust the plane visual image for color defects.

5.2

Image color adjustment effect test

The design model of this paper is compared with two traditional color adjustment systems to adaptively adjust the same image with color defect problem.

The distribution states of the difference colors of the images under the application of different systems are shown in Fig. 2, and Figs. (a), (b), and (c) are the design model of this paper, the traditional system 1, and the traditional system 2, respectively.

From the three sets of test results, it can be seen that only the image under the application of the design model in the paper, the difference color coordinates are concentrated in the center position, mainly concentrated in [0.1500,0.1575]. And after the two traditional systems are adjusted, the distribution of the difference color coordinates of the image is relatively scattered, which shows that its color adjustment effect is not ideal.

5.3

Image Color Enhancement Performance Analysis

The experiment measures the image enhancement effect of this paper’s method through four evaluation indexes, namely, mean value, variance, information entropy and clarity of the plane image, and the larger the value of the evaluation indexes, the better the image enhancement effect.

The information entropy formula is: (25) $E = - \sum_{i = 1}^{M} z (x_{i}) \lg z (x_{i})$

The clarity of the image is evaluated in terms of sharpness, which is expressed by the formula: (26) $g = \frac{1}{M \times N} \sum_{i, j \in M, N} \sqrt{Δ l_{x}^{2} (i, j) + Δ l_{y}^{2} (i, j)}$ where $M \times N$ represents the size of the image. $Δ I_{x}$ and $Δ I_{y}$ are the differences of the pixel points in the $x$ and $y$ directions.

When calculating the mean, variance, information entropy and sharpness on the image sub-blocks using the above formula, the flat image is divided into pixel blocks of $10 \times 10$ and the mean value of each sub-block of each channel in the RGB color space is calculated. Then calculate the mean value. The calculated results are normalized in the range of 0 to 1. The image enhancement performance analysis is shown in Fig. 3. Figures (a) and (b) show the results of the color enhancement performance analysis of the three methods for street and flower images, respectively. The information entropy of this paper’s method in Fig. (a) is 0.7236, which is significantly higher than that of the original image and the traditional methods1 and 2.

From the objective evaluation criteria, it can be judged that the mean, variance, information entropy and clarity of the image processed using this paper’s method are better than those achieved by the other two methods, which accurately and objectively reflects that this method is more advantageous in image enhancement processing.

5.4

Graphic Visual Design Evaluation

5.4.1

Basis for Design Effectiveness Evaluation

Formal beauty standards for graphic visual image design include:

Harmony: broadly defined as in the visual sense, two and on the visual elements or various regions to give people a sense of overall coordination. The narrow definition is that the relationship between the visual elements of unity and contrast is not disordered.

Balance: It does not mean the balance of moments on the plane structure, but the balance of the size, color and other visual elements of things in the visual scene conveyed to the visual cognitive system. Image in general to the visual center as the fulcrum, each visual element fulcrum for the middle also realize the balance of visual perception.

Visual weight: People’s visual attention focus is related to the visual attractiveness of the quadrant and visual elements. Changes in image outline, color or visual stimulus area distribution will affect the visual center of gravity, so the distribution of visual center of gravity is especially important in graphic visual image design. In graphic visual image design, the theme or information to be expressed should be allocated in the visual center of gravity, nearby.

Psychological effect: Graphic visual images convey visual potential information through visual elements, which in turn make observers produce associations and eventually achieve the expected mood of the advertisement.

When receiving external visual information, the human eye visual cognitive system and the brain work together. Each single component of a thing or visual image is rearranged and combined to generate a whole that is easier to process and recognize. Rather than differentiating between the individual single components at the outset. Within a single visual scene or a single frame of reference, the visual perceptual system of the human eye can only process a few independent areas of the whole. If a visual scene contains too many separate units, the visual perception system automatically simplifies and reorganizes them into a whole that is easier to process visually. If this processing cannot be accomplished, the visual cognitive system will default to the scene being in a disorganized state, and ultimately will not be able to process and perceive it correctly. Thus, the visual cognitive system of the human eye and the operation of the brain is a continuous process of organization, simplification and unification. It is because of this process that the visual system can automatically generate an overall unit that is easy to visually process, thus ensuring the efficiency of the visual cognitive system.

The main role of graphic visual design is to convey information to the audience through the visual elements in it, which is also the focus of graphic visual evaluation.

5.4.2

Evaluation results and analysis

In order to verify the effectiveness of the method proposed in this paper, 500 images obtained by searching the keyword “classic graphic visual design” on the webpage are used as the database. The dictionary strategy and color enhancement method proposed in this paper are used for the secondary color adjustment of the graphic visual images.

In this paper, the eye-tracking device generated by SMI (Germany) is used to conduct experiments on the graphic visual image data formed based on the method in this paper. The obtained eye-tracking instrument experimental data is used as a reference standard for the evaluation of graphic visual image design. The subjects were selected as 10 students of graphic visual design in a university, including 5 subjects related to the research direction and 5 subjects related to the non-research direction. Before the experiment, the professional operators of the equipment connected and debugged the eye-tracking equipment, and monitored the whole process of the experiment from the side. If the subjects or experimental equipment problems, can be adjusted in real time and continue the experiment.

In order to quantitatively illustrate the effectiveness of the method in this paper, we conducted experiments on 500 plane visual images, counted the number of attentional focuses and the order of attentional focuses that fell into the pre-expressed information area of the plane visual image, and found its average value, analyzed and analyzed the results. The average value is also found, analyzed and conclusions are drawn.

Firstly, the order of attention focus is analyzed, the lower the order of attention focus is, the more the pre-expressed information area of the graphic visual image is noticed first, and the more it can express the graphic visual designer’s intention in the first time. The average attention order comparison data is shown in Table 1.

It can be seen that the pre-expression information region in the graphic visual image can be better detected by using this paper’s method, which indicates that the pre-expression information region in the graphic visual design image formed by this paper’s method is more capable of attracting visual attention in the first time than the original graphic image, which further proves the effectiveness of this paper’s method.

Table 1.

Average observation sequence comparison data

	The data of the eye movement instrument		After the color adjustment image
	The data of the eye movement instrument			The data of the eye movement instrument
Primary image	Experimenter 1	1.75	This method	Experimenter 1	5.52
	Experimenter 2	1.68		Experimenter 2	5.63
	Experimenter 3	1.32		Experimenter 3	5.21
	Experimenter 4	1.52		Experimenter 4	5.24
	Experimenter 5	1.41		Experimenter 5	5.33
	Experimenter 6	1.36		Experimenter 6	5.58
	Experimenter 7	1.39		Experimenter 7	5.94
	Experimenter 8	1.42		Experimenter 8	5.57
	Experimenter 9	1.33		Experimenter 9	5.51
	Experimenter 10	1.56		Experimenter 10	5.34

6

Conclusion

This paper analyzes the application of color language in graphic design and proposes the design of geometric features of planar visual images based on probability density function estimation. Using generative adversarial network technology to deal with the perturbation offensive factors in plane visual image, to generate the color enhancement model of plane visual image. Design the color adaptive adjustment system for planar visual image with linear relationship between gain value and luminance value (R, G, B, W). That is, the plane visual image color enhancement model proposed in this paper can be used to adjust the plane visual image for color defects, and can effectively adjust the difference color in the plane visual image. It points out the importance of visual elements communication in graphic visual design as the focus of graphic visual design evaluation. Combined with the eye gaze data of the eye-tracking device, it reflects the feasibility of the color enhancement model of graphic visual images in this paper.

Lingua:: Inglese

Frequenza di pubblicazione:: 1 volte all'anno
Argomenti della rivista:: Scienze biologiche, Scienze della vita, altro, Matematica, Matematica applicata, Matematica generale, Fisica, Fisica, altro

Feed RSS della rivista

Geometric analysis and model construction of the relationship between color contrast and light and shade in plane visual design

Junjie He

Sheng Rong

Weidan Xie

Pubblicato online: 24 set 2025

Ricevuto: 31 gen 2025

Accettato: 02 mag 2025

DOI: https://doi.org/10.2478/amns-2025-0994

Parole chiaveProbability density function estimation, Image color enhancement, Geometric features, Planar visual image

© 2025 Junjie He et al., published by Sciendo

This work is licensed under the Creative Commons Attribution 4.0 International License.

Parole chiave
Probability density function estimation, Image color enhancement, Geometric features, Planar visual image