An Integrated Repainting System for Digital Restoration of Vijayanagara Murals

Abstract

An integrated repainting system is proposed in this paper for digital restoration of images of heritage murals, which have historical significance in their painting styles and ritualistic contents. The repainting system uses an ensemble of conventional image processing tools, in tandem with some state-of-the-art image rendition techniques, such as scaled bilateral filtering, source-constrained inpainting, tonal processing, and texture mapping based on gradient fusion. Murals that are old by nearly four centuries, have been imaged in situ from the walls of temples under a controlled environment, and then they have been fed to our repainting system. As the work of mural art is highly subjective, and so is its interpretation, a battery of tests for subjective evaluation has been performed to compare the different stages of restoration. Three different tournament strategies have been followed to make the test result devoid of any subjective bias as far as possible. The overall evaluation result is quite encouraging, as the restored images exhibit a gradually improving quality through the different stages of restoration.