IMPLEMENTASI REAL-ESRGAN UNTUK PENINGKATKAN RESOLUSI CITRA BANGUNAN BERSEJARAH DI INDONESIA

Ananda Feri Setyawan; Sri Mulyono

doi:10.29407/noe.v8i02.26604

Authors

Ananda Feri Setyawan Universitas Islam Sultan Agung Semarang
Sri Mulyono

DOI:

https://doi.org/10.29407/noe.v8i02.26604

Keywords:

Real-ESRGAN, High-Order Degradation Model, Historical Building Images

Abstract

The preservation of digital images of historical buildings in Indonesia faces visual quality challenges due to degradation such as blur, noise, resizing, and compression. Images from Leiden University Libraries experience visual degradation due to the digitization process, the age of the documents, and the limitations of documentation technology, so the type of degradation in the images is not known with certainty. Therefore, to address these problems, this study implements the Real-ESRGAN method to realistically increase the resolution of historical building images without explicit information on the type of degradation. Real-ESRGAN is a development of ESRGAN that combines the RRDB architecture in the generator and U-Net in the discriminator. This model is trained using the Flickr2K dataset whose degradation is simulated with a high-order degradation model to resemble real-world images. Evaluation using the no-reference CLIP-IQA metric shows that the Real-ESRGAN model is capable of producing high-resolution images with good perceptual quality without the need for reference images. The highest average CLIP-IQA score obtained was 0,70863 at the 119th epoch. This value indicates that the resulting image results are in accordance with good quality images visually and show the potential to increase the resolution of images of historic buildings in Indonesia.

Abstract views: 0 , PDF downloads: 0

References

Agustinova, D.E. (2022) ‘Strategy for Maintaining Cultural Heritage Objectives’, 18(2), pp. 60–68.

Goodfellow, I. et al. (2014) ‘Generative adversarial networks’, Communications of the ACM, 63(11), pp. 139–144. Available at: https://doi.org/10.1145/3422622.

Hardiansyah, B., Armin, A.P. and Yunanda, A.B. (2019) ‘Rekonstruksi Citra Pada Super Resolusi Menggunakan Interpolasi Bicubic’, INTEGER: Journal of Information Technology, 4(2), pp. 1–12. Available at: https://doi.org/10.31284/j.integer.2019.v4i2.684.

Ledig, C. et al. (2017) ‘Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network’, pp. 4681–4690.

Nandal, P. et al. (2024) ‘Super-resolution of medical images using real ESRGAN’, IEEE Access, (December), pp. 176155–176170. Available at: https://doi.org/10.1109/ACCESS.2024.3497002.

Rohim, M.I.A. et al. (2024) ‘Peningkatan Performa Pengenalan Wajah pada Gambar Low-Resolution Menggunakan Metode Super-Resolution’, Jurnal Teknologi Informasi dan Ilmu Komputer, 11(1), pp. 199–208. Available at: https://doi.org/10.25126/jtiik.20241117947.

Safira, F. et al. (2020) ‘Peran Arsip Dalam Pelestarian Cagar Budaya Di Indonesia: Sistematika Review’, Baca: Jurnal Dokumentasi Dan Informasi, 41(2), p. 289. Available at: https://doi.org/10.14203/j.baca.v41i2.593.

Sarwito, J., Kesia, B. and Vania, C. (2024) ‘PENDEKATAN DIGITAL DALAM PENELITIAN SEJARAH ARSITEKTUR : TEKNOLOGI PEMODELAN DAN REKONSTRUKSI 3D PADA BANGUNAN KOLONIAL DI SEMARANG’, 4(1), pp. 3–8. Available at: https://doi.org/10.24167/joda.v4i1.12771.

Shearer, C. (2000) ‘The CRISP-DM model: the new blueprint for data mining’, Journal of data warehousing, 5(4), pp. 13–22.

Wang, J., Chan, K.C.K. and Loy, C.C. (2022) ‘Exploring CLIP for Assessing the Look and Feel of Images’, Proceedings of the 37th AAAI Conference on Artificial Intelligence, AAAI 2023, 37, pp. 2555–2563. Available at: https://doi.org/10.1609/aaai.v37i2.25353.

Wang, X. et al. (2018) ‘ESRGAN: Enhanced super-resolution generative adversarial networks’, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 11133 LNCS, pp. 63–79. Available at: https://doi.org/10.1007/978-3-030-11021-5_5.

Wang, X. et al. (2021) ‘Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Data’, Proceedings of the IEEE International Conference on Computer Vision, 2021-Octob, pp. 1905–1914. Available at: https://doi.org/10.1109/ICCVW54120.2021.00217.

Zhang, Y. and Yu, W. (2022) ‘Comparison of DEM Super-Resolution Methods Based on Interpolation and Neural Networks’, Sensors, 22(3). Available at: https://doi.org/10.3390/s22030745.