Pattern reconstruction method for unearthed ceramic artifacts based on improved DDPM and StyleGAN

The digital restoration of unearthed ceramic artifacts often faces difficulties, such as incomplete patterns, weathering and fading, and blurred textures. To enhance the structural reconstruction ability and style adaptability of pattern reconstruction, a framework for reconstructing unearthed ceramic artifacts patterns based on an improved denoising diffusion probability model and style generation adversarial network is proposed. A multi-scale efficient channel attention mechanism is introduced into the denoising diffusion probability model to enhance the ability to capture detailed textures. In addition, a progressive generation and style injection module is constructed to enhance the flexibility of the style generation adversarial network in pattern diversity and style control. The experimental results show that the improved denoising diffusion probability model achieves a structural similarity index of 0.945, with peak signal-to-noise ratio and average color distance of 28.6dB and 3.1, respectively, which is superior to the comparison model. The improved style generative adversarial network achieves a Diversity Index (DI) of 0.885 and an average generation delay controlled within 1.52s. The two constitute a dual-pathway restoration toolkit that complements the overall digital preservation workflow in terms of structural reconstruction and style generation, providing an effective technical path for high-quality restoration and diversity modeling of artifacts patterns.