The automatic arrangement algorithm in image generation design has become the key to improving generation efficiency and quality. This study focuses on an automatic orchestration algorithm based on Directed Acyclic Graph (DAG), and compares it with Genetic Algorithm (GA) and Ant Colony Optimization (ACO) to evaluate its application performance in intelligent vision image generation. The research methods include data preprocessing and feature extraction, selection and training of image generation models, and implementation and optimization of automatic arrangement algorithms. Through a series of comparative experiments, this study comprehensively evaluated four core indicators: algorithm efficiency, image quality, diversity, and robustness. The experimental results show that the DAG based algorithm outperforms GA and ACO algorithms in terms of the Fréchet Inception Distance (FID) value of image quality, time consumption of generation efficiency, and robust peak signal-to-noise ratio. Its maximum FID value is only 63, the minimum is 41, the minimum image generation time is only 73.3 s, and the peak signal-to-noise ratio reaches 45.5 dB. DAG based algorithms can significantly improve generation efficiency and robustness while maintaining generation quality. The conclusion of this study provides strong performance proof for automatic arrangement algorithms in intelligent visual image generation design, and lays the foundation for future research directions and practical applications.