Digital prediction of ancient ceramic images missing areas based on deep adversarial and reverse diffusion

Ancient ceramics are subjected to natural disasters, temporal erosion, and other environmental impacts. These historical ceramics exhibit particular vulnerability to glaze wear, disc damage, and fragmentation phenomena. Such degradation frequently results in ceramic images with missing texture, blurred patterns and other issues. To address these challenges, a digital prediction for missing areas in ancient ceramic images through a coupled deep adversarial and reverse diffusion method is proposed. Using deep adversarial patterns prediction and reverse diffusion complementary prediction to construct characteristic equations for patterns and complementary prediction in ancient ceramics. Experimental results demonstrate that model achieves a discriminator loss of 0.620, with MSE and LPIPS metrics reaching 0.0137 and 0.0166, respectively. The proposed method enables effective prediction of missing regions in ancient ceramic images, providing valuable insights into the artifacts’ historical context and cultural significance. This advancement holds substantial implications for both the preservation and interpretative transmission of cultural heritage.