This study examines the role of image-based generative artificial intelligence (AI) in the restoration and utilization of cultural heritage. Leveraging models such as GANs, VAEs, and NeRF, AI enables the realistic reconstruction of damaged or fragmented heritage data. These technologies, combined with digital twins and virtual restoration, have significantly enhanced the preservation and visualization of cultural assets. Case studies include the colorization of historical glass plate negatives at the National Museum of Korea, the restoration of Dunhuang murals and ancient sculptures, and AI-driven monitoring systems for site conservation. Creative applications also extend to archaeological site detection, ancient manuscript decoding, and reinterpretation of traditional crafts through media art. Despite these advances, challenges remain, including data bias, authenticity concerns, ethical implications, and intellectual property issues. The study highlights the importance of inclusive datasets, transparent restoration practices, and interdisciplinary collaboration. It argues for the development of clear ethical guidelines to ensure responsible and culturally respectful use of AI in heritage contexts. This research offers strategic insights into how AI can support sustainable and meaningful cultural heritage preservation.