Researchers Yunshan Geng and Xiaoning Wang have introduced a specialized art generation model centered on visual element communication, known as VCE-GM, which demonstrates significant advantages over recent deep learning models across multiple performance indicators. Detailed in peer-reviewed research published in Discover Applied Sciences, the framework addresses critical generative limitations that have emerged as artificial intelligence and digital art become increasingly intertwined. While the rapid development of artificial intelligence technology has established the integration of AI and digital art as a major new research trend, and existing generative models already possess powerful image synthesis capabilities, these systems frequently encounter serious generation obstacles. Specifically, generative synthesis often suffers from empty content or semantic defocus because the underlying models generate imagery without adhering to basic visual principles.
To resolve the issues of empty content and semantic defocus, the researchers constructed VCE-GM to systematically formalize and parameterize the relevant elements of image generation, embedding these parameters directly into the entire generation process. These parameterized dimensions specifically comprise points, lines, surfaces, colors, space, rhythm, proportion, and texture. To connect these visual parameters with user prompts, the architecture achieves a mapping relationship from natural language to a structured vector through a specialized visual element embedding module, ensuring that descriptive text translates directly into structured visual elements.
Beyond the initial embedding of visual elements, the VCE-GM architecture utilizes a multimodal fusion module to achieve collaborative optimization across both global and local potential spaces. In terms of training objectives, the authors modified the loss function by incorporating an element consistency loss together with an integrated perceptual feedback module.
Experimental results reported in the study show that the constructed VCE-GM model achieves significant advantages over the latest deep learning models in multiple evaluation indicators, specifically showing marked improvements in element consistency and reduction in generation error. In addition to these primary technical indicators, the experimental findings demonstrated that the model possesses good cultural adaptability across creative contexts. The architecture also exhibited significant advantages in multi-resolution image generation and human-computer collaborative creation. Based on these validated outcomes, the authors conclude that the VCE-GM model provides a reliable technical path for digital image art creation.
Subject of Research: Visual communication-guided artificial intelligence in digital art design
Article Title: Exploration of art and digital design technologies guided by visual communication elements
Article References: Geng, Y., & Wang, X. (2026). Exploration of art and digital design technologies guided by visual communication elements. Discover Applied Sciences. https://doi.org/10.1007/s42452-026-09686-8
Image Credits: AI Generated
DOI: 10.1007/s42452-026-09686-8
Keywords: artificial intelligence, generative models, digital art, visual communication, VCE-GM, multimodal fusion, image synthesis, human aesthetics, Exploration, digital, design, technologies
News Source: Denise Maddox. (November 5, 2026). Visual Element-Guided AI Model Enhances Image Synthesis and Digital Art Creation. Scienmag.
