A Novel Feature Enhancement Approach for Image Design Through Transformer-Integrated Generative Adversarial Network

Traditional image feature enhancement methods often struggle to capture and precisely enhance the complex design elements in images. Hence, a transformer-based generative adversarial network (Transformer-GAN) designed in this paper fully utilizes the self-attention mechanism of the Transformer to capture the deep-level relationships between image design features. The proposed method generates images that not only have a more structurally similar appearance to real images but also have more prominent and distinct design features. In the experiments, we compared various existing image design feature enhancement methods. The results show that the proposed Transformer-GAN significantly enhances design features such as edge details, texture complexity and color contrast while maintaining the overall consistency of the image. The proposed method of fusing a Transformer with a generative adversarial network has brought new ideas and methods to the field of image design feature enhancement. This method not only has theoretical innovation but also shows significant advantages and potential in practical applications.