Optimizing the Spatial Coordination Between Graphics and Text in Visual Design Using GCN

At present, the optimization of spatial coordination between graphics and text in visual design still faces great challenges. Traditional methods have limitations in dealing with complex layout relationships, and it is difficult to achieve dynamic coordination and global optimization of design elements. To solve this problem, this study proposes an optimization method based on graph convolutional networks (GCN) to improve the spatial coordination of graphics and text in design. First, a graph structure is constructed for the graphic and text elements in visual design, and graphics and text are represented as graph nodes, with their spatial relationships connected as edges. Then, GCN is used to extract features from the graph structure, and multi-layer convolution is used to capture the nonlinear relationship between nodes and global layout characteristics. Then, combined with the objective function design, the loss function is used to balance the spatial distribution of graphics and text, so that they can achieve the best match in terms of aesthetics and functionality. The experiments are conducted on three actual design case data sets, including advertising design, user interface design, and publication typesetting. The results show that this method significantly improves the coordination of the design, the average distance between graphics and text is reduced by 21.2, the layout symmetry is improved by 8.48, and the user’s subjective satisfaction score is improved by 0.52 points. Research shows that the GCN-based method can not only effectively capture the complex interactions in the spatial relationship between images and text but also achieve global optimization, thus providing a new intelligent solution for visual design with broad application potential.