MUMBAI, India, Aug. 17 -- Intellectual Property India has published a patent application (202641094174 A) filed by Vardhaman College Of Engineering on August 04, 2026, for Deep Learning-Based Restoration Of Ancient Artworks Using Style Reconstruction.
Inventors include Ms. Ruqqaiya Begum; Dr. N Deepika Rani; Dr. Mohammad Gouse Baig; Ms. Sowmya M; Mr. Rajkumar; and Ms. Ashima Jain.
The application for the patent was published on August 14, 2026, under issue no. 33/2026.
Abstract: Deep Learning-Based Restoration of Ancient Artworks Using Style Reconstruction is the proposed invention. The invention proposed acquires high resolution images of the damaged paintings, murals, manuscripts and cultural artefacts and does pre-processing including noise removal, illumination correction, colour normalisation and damaged region detection. Semantic features that represent artistic structures, brushstroke patterns, texture characteristics, and contextual information are extracted from preserved regions of the artwork. A diffusion model progressively reconstructs the missing or deteriorated areas while maintaining artistic style, texture continuity, colour harmony, and historical consistency through adaptive style reconstruction constraints. A restoration verification module evaluates reconstruction quality using structural similarity, semantic coherence, texture preservation, and historical compatibility metrics. An explainable artificial intelligence module generates restoration confidence maps, quality reports, and visual comparisons to assist conservators in validating restored content. The invention proposed allows to preserve valuable historic artefacts by means of an accurate, non-destructive, transparent and digitally reversible restoration of ancient artworks, suitable for museums, archaeological institutions, cultural heritage organisations and laboratories for the conservation of art.
Disclaimer: Curated by HT Syndication.
