Mural image restoration involves repairing damaged sections of murals to attain desirable visual outcomes. In recent years, the development of mural restoration algorithms has emerged as a key area of interest, driven by the need to preserve murals as valuable artifacts of human historical heritage. Despite their significance, murals have suffered various degrees of deterioration over time. The limited availability of mural-specific datasets and the complexity of mural textures pose significant challenges for contemporary image restoration algorithms, rendering them less effective in mural restoration tasks. To address this, we have compiled a dataset encompassing 3,492 murals and introduced a novel mural image restoration approach, the Edge Assistance and Aggregated Contextual Transformations GAN (EAAOT-GAN). This approach is structured around two phases: edge generation and image restoration. Initially, it generates complete edges of murals, followed by the restoration of the entire mural images through the integration of these edges. Comparative analysis with leading image restoration techniques demonstrates that our method competes favorably with the most advanced mural restoration models, as evidenced by both qualitative and quantitative evaluations.