Search

CN-121999072-A - Color scheme generation method, system and product based on multidimensional semantic and visual perception model

CN121999072ACN 121999072 ACN121999072 ACN 121999072ACN-121999072-A

Abstract

The invention belongs to the field of computer graphics and artificial intelligence, and relates to a color scheme generation method based on a multidimensional semantic and visual perception model, which determines a dominant color based on the multidimensional semantic and user preference; and based on a parameterized perception rule base, carrying out perception optimization on the candidate color matching scheme set and outputting a final scheme, wherein the final scheme comprises optimal area proportion parameters of main colors, auxiliary colors and interspersed colors. The invention precisely quantifies and correlates abstract design requirements with specific color parameters, and ensures that the generated color scheme accords with the design intent input by a user from the source. The invention can also generate the personalized color matching scheme which is more in line with the aesthetic of the target user on the basis of meeting the general design principle, thereby greatly improving the user acceptance of the scheme.

Inventors

  • LIU XINXIONG
  • LI LIN
  • GAN YAN
  • WANG YALI

Assignees

  • 华中科技大学

Dates

Publication Date
20260508
Application Date
20260119

Claims (10)

  1. 1. The color scheme generation method based on the multidimensional semantic and visual perception model is characterized by comprising the following steps of: Determining a dominant color based on multi-dimensional semantics and user preferences; Receiving multidimensional semantic input, namely receiving one or more color image words selected by a user from a preset multidimensional semantic space and receiving importance weights distributed by the user for each dimension; calculating a semantic gravity point, namely calculating the semantic gravity point in the HSB three-dimensional color space according to the semantic words selected by the user and the weights thereof; correcting by applying a user preference model, namely searching in a neighborhood range by taking the calculated semantic gravity center as a center and scoring the preference degree of a plurality of candidate colors in the neighborhood according to the user preference; Determining a final dominant color, namely selecting the color with the highest user preference score in the neighborhood as the final dominant color; generating a candidate color matching scheme set based on the semantic guided harmonious template; Around the determined dominant color, generating a series of candidate multicolor schemes which are harmonious in theory and accord with design intent according to preset semantic-harmonious template mapping rules; thirdly, based on a parameterized perception rule base, carrying out perception optimization and outputting a final scheme; carrying out multi-constraint scoring and sequencing on the candidate schemes, namely comprehensively scoring each candidate multicolor scheme generated in the step two from a plurality of dimensions such as harmony degree, user preference degree, semantic consistency and the like, and selecting Top-N schemes with highest scores; Optimizing the Top-N schemes by applying a perception rule, and acquiring optimal area proportion parameters of corresponding main, auxiliary and embellishment colors according to a preset parameterized perception rule according to a harmonious template used by the current scheme; And outputting a final scheme, namely finally presenting one or more comprehensively optimized color schemes to a user.
  2. 2. The method for generating a color scheme based on a multi-dimensional semantic and visual perception model according to claim 1, wherein the multi-dimensional semantic space comprises at least one of style, function, emotion and cultural dimension.
  3. 3. A method of generating a color scheme based on a multi-dimensional semantic and visual perception model according to claim 1, characterized in that in the step of calculating the semantic centroid a pre-trained semantic-color mapping model is invoked, which is based on an adaptive mapping algorithm comprising non-linear activation functions for mapping semantic requirements of different intensities to specific areas of the color space.
  4. 4. The method for generating a color scheme based on a multi-dimensional semantic and visual perception model according to claim 1, wherein in the step of correcting by applying a user preference model, a pre-trained user preference prediction model is called to score the preference degree of a plurality of candidate colors in the neighborhood, and the user preference prediction model is obtained based on machine learning training of user investigation data and is used for predicting the preference probability of a user on the colors according to the input user portrait and design situation.
  5. 5. The method for generating a color scheme based on a multi-dimensional semantic and visual perception model according to claim 1, wherein the preset semantic-harmony template mapping rule is a pre-constructed semantic-harmony template mapping rule library, and the semantic-harmony template mapping rule library associates the multi-dimensional semantic with the color harmony templates, so as to select one or more color harmony templates which can most express the design intent of the user input according to the received multi-dimensional semantic input; The method also comprises the steps of generating auxiliary colors and interspersed colors, namely, calculating HSB value ranges of the theoretical auxiliary colors and the interspersed colors by using the selected color harmony templates based on the main colors, and sampling in the ranges to generate a plurality of three-color candidate color matching schemes comprising the main colors, the auxiliary colors and the interspersed colors.
  6. 6. The method for generating a color scheme based on a multidimensional semantic and visual perception model according to claim 1, wherein the parameterized perception rule base quantifies visual perception effects caused by factors such as area ratio, simultaneous comparison and the like of multicolor combination under a specific background in a parameterized mode.
  7. 7. The method of claim 6, wherein the step of applying the perception rules for optimization comprises core optimization and/or risk prompting: the core optimization comprises recommending optimal area proportion layout, namely inquiring a parameterized perception rule base, and directly acquiring optimal area proportion parameters verified by experimental data and driven by the data according to a harmonious template used by the current scheme; the risk prompt comprises analyzing the perceived risk in the scheme according to a parameterized perceived rule base, and generating a natural language prompt; the output final solution includes presenting one or more fully optimized color schemes to the user, each scheme including HSB color values, optimal area ratio layout suggestions recommended based on human perception experimental data, and perceived risk cues.
  8. 8. A computer system comprising a memory, a processor and a computer program stored on the memory, wherein the processor executes the computer program to implement a method of generating a color scheme based on a multi-dimensional semantic and visual perception model as claimed in any one of claims 1 to 7.
  9. 9. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of a method for generating a color scheme based on a multidimensional semantic and visual perception model according to any one of claims 1-7.
  10. 10. A computer program product comprising a computer program, characterized in that the computer program, when being executed by a processor, implements the steps of a method for generating a color scheme based on a multi-dimensional semantic and visual perception model according to any one of claims 1-7.

Description

Color scheme generation method, system and product based on multidimensional semantic and visual perception model Technical Field The invention belongs to the field of computer graphics and artificial intelligence, and in particular relates to an intelligent and personalized multicolor color scheme generation method and a related system combining natural language processing, machine learning and psychophysics experimental data. Background Color design is a key link in the fields of industrial design, UI/UX design, brand image construction and the like. The traditional color design flow is highly dependent on personal experience and subjective judgment of designers, and has the problems of low efficiency, poor reproducibility, difficulty in meeting the needs of diversified users and the like. To solve this problem, the prior art mainly adopts the following two modes: color matching tools based on fixed harmony rules, such as Adobe Color and the like, which generate Color matching schemes based on classical Color harmony theory (such as complementary colors, similar colors, trichromatic groups and the like). Such tools can guarantee the harmony of the scheme in the geometrical relationship of the color wheel, but the main defects are that: Disjoint from design intent, the generated solution lacks direct relevance to specific design goals (e.g., "technological feel", "safety"), and the designer needs to perform extensive manual screening and secondary authoring. The proposal is universal and cannot be adjusted in a personalized way according to the specific portraits (such as age, cultural background and professional experience) of the target user. Color recommendation systems based on data mining, some systems mine popular color matching combinations by analyzing color matching of existing design works. The drawbacks of this type of method are: The recommended result is statistically relevant, lacks the design theory and the color psychology basis of the bottom layer, and has poor interpretation. The innovation is insufficient, the circulation of repeated popular modes is easy to fall into, and a prospective and innovative color scheme is difficult to generate. Furthermore, the above prior art generally ignores a key issue of perceived consistency of color. That is, a set of color schemes that are numerically (physically) harmonious may, in practical applications, create visual illusions due to interactions between colors (e.g., synchronization contrast effects) and layout factors (e.g., area ratios), resulting in deviations of the visual effect ultimately perceived by the user from the designer's initial intent. For example, the theoretical primary-secondary relationship may be reversed due to improper area allocation, and the key stippling may be "engulfed" by the background due to insufficient contrast. Therefore, a new technical solution is urgently needed in the art, and abstract design intent, personalized preference of users and objective visual perception rules can be systematically integrated, so that a high-quality multicolor color scheme which meets design goals, satisfies aesthetic requirements of users and maintains consistency of visual effects in practical application is automatically generated. Disclosure of Invention Aiming at the defects or improvement demands of the prior art, the invention provides a color scheme generation method, a system and a product based on a multidimensional semantic and visual perception model, thereby solving the technical problems of disjoint of the color scheme generation and the design intention, lack of individuation and neglect of visual perception consistency in the prior art. To achieve the above object, according to one aspect of the present invention, there is provided a color scheme generating method based on a multi-dimensional semantic and visual perception model, comprising the steps of: Determining a dominant color based on multi-dimensional semantics and user preferences; Receiving multidimensional semantic input, namely receiving one or more color image words selected by a user from a preset multidimensional semantic space and receiving importance weights distributed by the user for each dimension; calculating a semantic gravity point, namely calculating the semantic gravity point in the HSB three-dimensional color space according to the semantic words selected by the user and the weights thereof; correcting by applying a user preference model, namely searching in a neighborhood range by taking the calculated semantic gravity center as a center and scoring the preference degree of a plurality of candidate colors in the neighborhood according to the user preference; Determining a final dominant color, namely selecting the color with the highest user preference score in the neighborhood as the final dominant color; generating a candidate color matching scheme set based on the semantic guided harmonious template; Around the determined dominant color,