ePrints@IIScePrints@IISc Home | About | Browse | Latest Additions | Advanced Search | Contact | Help

Transforming Pixels into a Masterpiece: AI-Powered Art Restoration using a Novel Distributed Denoising CNN (DDCNN)

Sankar, B and Saravanan, M and Kumar, K and Dubakka, S (2023) Transforming Pixels into a Masterpiece: AI-Powered Art Restoration using a Novel Distributed Denoising CNN (DDCNN). In: 3rd International Conference on Emerging Techniques in Computational Intelligence, ICETCI 2023, 21-23 septembar 2023, Hyderabad, pp. 164-175.

pro_2023_int_con_eme_tec_com_int_ICETCI_2023.pdf - Published Version

Download (56MB) | Preview
Official URL: https://doi.org/10.1109/ICETCI58599.2023.10331299


Art restoration plays a pivotal role in preserving and revitalising cultural heritage. However, conventional art restoration methods are often fraught with limitations, including the challenge of faithfully reproducing the original artwork's essence while addressing issues such as fading, staining, and physical damage. In this context, we present a pioneering approach in this paper that harnesses the potential of deep learning, particularly Convolutional Neural Networks (CNNs), coupled with Computer Vision techniques, to revolutionize the art restoration process. Our method begins by generating artificially induced deteriorated art images, creating a comprehensive dataset encompassing various forms of distortions and multiple levels of degradation. This dataset is the foundation for training a Distributed Denoising CNN (DDCNN), capable of effectively removing distortions while preserving the intricate details inherent to artworks. This integration of Computer Vision and CNN-based denoising enables the restoration of artworks with high accuracy, ensuring that the original artistic essence is faithfully preserved. One of the key strengths of our approach lies in its adaptability to different kinds of distortions at varying levels of degradation. Utilizing a versatile training dataset, our Distributed Denoising CNN can address a wide spectrum of distortion types, ranging from subtle colour variations to more severe structural damage. This adaptability empowers our method to cater to a diverse array of deteriorated artworks, including paintings, sketches and photographs. Through extensive experimentation on a diverse dataset, our results consistently demonstrate the efficiency and effectiveness of our proposed approach by comparing it against other De-noising CNN models. We showcase the substantial reduction of distortion in the art image, transforming deteriorated artworks into masterpieces. Quantitative evaluations further underscore the superiority of our method compared to traditional restoration techniques, reaffirming its potential to reshape the landscape of art restoration and contribute significantly to the preservation of our cultural heritage. In summary, our paper introduces a groundbreaking AI-powered solution that leverages the synergy of Computer Vision and deep learning, exemplified by Distributed Denoising CNN, to restore artworks with unprecedented accuracy and fidelity. This transformative approach not only overcomes the limitations of existing methods but also paves the way for future advancements in the field of art restoration, ensuring the enduring legacy of our cultural treasures. © 2023 IEEE.

Item Type: Conference Paper
Publication: Proceedings of the 2023 International Conference on Emerging Techniques in Computational Intelligence, ICETCI 2023
Publisher: Institute of Electrical and Electronics Engineers Inc.
Additional Information: The copyright for this article belongs to Author.
Keywords: Convolution; Convolutional neural networks; Deep learning; Historic preservation; Image reconstruction; Restoration, Art restoration; Arts image; Computer vision techniques; Convolutional neural network; Cultural heritages; De-noising; Distributed denoising convolutional neural network; Physical damages; Restoration methods; Restoration process, Computer vision
Department/Centre: Division of Mechanical Sciences > Mechanical Engineering
Date Deposited: 04 Mar 2024 04:55
Last Modified: 04 Mar 2024 04:55
URI: https://eprints.iisc.ac.in/id/eprint/84092

Actions (login required)

View Item View Item