Ancient Text Image Inpainting Algorithm via Edge Guide and Laplacian Pyramid Decomposition

Current image inpainting methods are often perform poorly on ancient text images, producing results with blurred textures or incomplete structural content. To address this problem, we propose an inpainting algorithm for ancient text images via edge guide and laplacian pyramid decomposition. We first use an edge restoration network to restore the edge structure for the damaged regions and construct an edge-guided map. Then, we employ the pre-trained text learning network to restore the local damaged regions and obtain a local inpainting image, which is decomposed into a content image and a detail map through laplacian pyramid transform. At last, we build a laplacian pyramid restoration network to inpaint the damaged image by applying the content restoration network on the two layers of laplacian pyramid respectively. The content restoration network introduces a dual cross-encoder and multi-scale fusion blocks to prompt the network to obtain more effective feature information and generate desirable image inpainting results. The superiority quantitative results on the benchmark dataset demonstrate the effectiveness and feasibility of the proposed method, that peak signal to noise ratio (PSNR) is 34.322, structural similarity (SSIM) is 0.970 and root mean square error (RMSE) is 5.203.