Construction of multi-style face models based on artistic image generation algorithms

• Published in: Nonlinear engineering Vol. 14; no. 1; pp. 43 - 6
• Main Author: Xu, Zhaoyi
• Format: Journal Article
• Language: English
• Published: Berlin De Gruyter 21.08.2025; Walter de Gruyter GmbH
• Subjects: Algorithms; artistic image generation algorithms; clarity; Data processing; face models; Image processing; Modules; multi-scale attention modules; receptive fields
• ISSN: 2192-8029; 2192-8010; 2192-8029
• DOI: 10.1515/nleng-2025-0149

In response to the issues of insufficient clarity and data processing capabilities in the current face model construction process, a method for constructing multi-style face models based on artistic image generation algorithms is proposed. The new method incorporates the image decomposition and normalization model to enhance the algorithm’s ability to process image data. Subsequently, receptive fields, core attention modules, multi-scale attention modules, and dynamic selection attention modules are introduced to enhance the algorithm’s handling of details and model clarity in the process of constructing face data. The research results indicated that the new algorithmic model could improve the data metrics of face models. The incorporation of distinct modules has the potential to enhance various algorithmic metrics. However, the multi-scale attention module emerged as a pivotal component, demonstrating superior stability compared to numerous alternative algorithms. When dealing with face images of different styles, the performance of the new algorithm was better compared to that of other algorithms. Therefore, the new algorithm can enhance the numerical values of face models to a certain extent, improve parameters such as facial texture clarity, and have important research significance in the direction of constructing multi-style face models.