To address the issue of detail loss and low inpainting efficiency caused by complex damages in image inpainting (such as wrinkles, mildew spots, etc.), as well as the limitations of using artificially synthesized masks that do not match the distribution of real damaged images in existing methods, we propose a physical degradation simulation framework and an efficient two-stage sampling inpainting method based on a lightweight diffusion model. First, we construct the physical degradation simulation framework to enhance damage features such as wrinkles and mildew spots on the image, and design corresponding mask images for the simulated damaged images. We then build a lightweight diffusion backbone network, reducing the parameter count by 83%. We employ an efficient two-stage sampling mechanism, reconstructing the main structure at low resolution, and then transform the result into a 256×256 latent space for detail optimization. This two-stage sampling method not only improves inpainting efficiency but also enhances inpainting quality.