Revitalizing ancient murals in the Shekhawati region through image inpainting techniques

India has an extensive range of ancient artistic expressions in the form of mural paintings. These murals incorporate architectural features representing the cultural heritage of ancient civilizations. These paintings have declined over time due to human negligence and have deteriorated. The traditional approaches to restore this significant artwork are less effective and have been declared impotent. Thus, a practical digital inpainting framework is suggested to breathe new life into degraded murals. The suggested framework employs a two-step approach to attain optimal results in reconstructing murals for restoration purposes. In the first step, a semantic model is utilized to generate damage, defining masks that serve as inpainting parameters in various instances. The second step evaluates the effectiveness of three distinct inpainting architectures: PCONV, Conditional Texture Structure Dual Generation, and T-former.Additionally, this study provides a qualitative and quantitative comparison, utilizing metrics such as style loss, perceptual loss, SSIM, PSNR, and the perceptual quality of completeness. The results of the comparative analysis reveal that the T-former model outperforms others significantly, attaining an average SSIM of 0.9947 and a PSNR score of 45.3815 on test images. Moreover, the experiment's reconstruction, achieved by constraining all patches, successfully passes the visual inspection test, faithfully replicating the original mural.