The application of a variational partial differential equation for mural image inpainting is mainly studied in this paper. Squared L2-norm of the fractional Laplacian is used as regularization term, the existence and uniqueness of the model are proved within the framework of fractional Sobolev spaces. Discretization of fractional Laplacian is adjusted for mural images and stability is analyzed. Experimental results show that, compared with commonly used methods, the proposed model significantly improves the PSNR and SSIM of mural inpainting, and effectively enhances the authenticity of restoring high-frequency detail information of mural images while maintaining structural consistency.
Maimaitiaili et al. (Wed,) studied this question.