Peningkatan Kualitas Citra Bawah Air Menggunakan GAN dengan Mekanisme Residual dan Attention

Penulis

  • Nursanti Abdurrachman Universitas Indonesia, Depok
  • Dina Chahyati Universitas Indonesia, Depok

DOI:

https://doi.org/10.25126/jtiik.2025126

Kata Kunci:

peningkatan citra bawah air, GANs, attention mechanism, residual mechanism

Abstrak

Citra bawah air sering mengalami penurunan kualitas yang disebabkan oleh proses redaman dan hamburan cahaya yang dipengaruhi oleh panjang gelombang serta jarak antara objek dan kamera. Faktor-faktor seperti gangguan pencahayaan dan kompleksitas latar bawah air sering kali menyebabkan citra menjadi buram, mengalami perubahan warna, dan mengalami berbagai bentuk degradasi visual lainnya. Upaya peningkatan kualitas citra bawah air tidak hanya bertujuan untuk memperbaiki tampilan visual, tetapi juga untuk menghasilkan citra yang lebih baik sebagai masukan bagi proses pengolahan citra lanjutan. Keunikan dan kompleksitas dari citra bawah air membuat metode peningkatan konvensional yang dirancang untuk kondisi seperti cahaya rendah dan berkabut menjadi kurang efektif bila diterapkan dalam konteks bawah air. Untuk mengatasi hal ini, penelitian ini memanfaatkan Generative Adversarial Networks (GANs) yang dilengkapi dengan mekanisme attention dan residual pada bagian generator. Penggunaan attention dan residual mechanism memungkinkan jaringan untuk fokus pada bagian penting dari gambar dan membantu dalam pemulihan informasi yang hilang selama proses peningkatan gambar. Penelitian ini menggunakan dataset EUVP dengan data latih sebanyak 3330 citra, data uji sebanyak 1110 citra, dan data validasi sebanyak 1110 citra. Pendekatan yang diusulkan pada penelitian ini mampu menjawab tantangan-tantangan utama dalam peningkatan citra bawah air. Citra yang dihasilkan memiliki keseimbangan yang cukup baik antara kualitas alami gambar dan kesamaan struktural dengan gambar target. Pendekatan yang diusulkan dalam penelitian ini juga mampu menyeimbangkan pemulihan warna dan preservasi tekstur sehingga menghasilkan gambar yang lebih alami dan realistis, tanpa artefak warna yang berlebihan atau kehilangan detail tekstur. Hasil evaluasi menunjukkan metode yang diusulkan mencapai nilai PSNR 23.7966, SSIM 0.7219, UIQM 1.4485, dan UCIQE 0.2389.

 

Abstract

Underwater images inevitably suffer from quality degradation issues caused by wavelength- and distance-dependent attenuation and scattering. Light interference and complex underwater backgrounds frequently cause blurriness, color distortion, and other degradation problems in underwater images. Enhancing underwater image quality not only aims to improve visual perception but also to provide higher-quality inputs for other image processing techniques. The uniqueness and complexity of underwater images make enhancement methods designed to address issues like low light and foggy conditions unsuitable for underwater image enhancement tasks. Attention and residual mechanisms allow the network to focus on important parts of the image and aid in recovering lost information during the enhancement process. This research employs Generative Adversarial Networks (GANs) for underwater image enhancement, incorporating attention and residual mechanisms into the generator part. This study uses the EUVP dataset with 3330 training images, 1110 test images, and 1110 validation images. This research’s proposed method can address the challenges faced by underwater images. The resulting images achieve a good balance between natural image quality and structural similarity with the ground truth. The proposed method in this study is also able to balance color recovery and texture preservation, producing more natural and realistic images without excessive color artifacts or loss of texture details. Evaluation results show that the proposed method achieves PSNR 23.7966, SSIM 0.7219, UIQM 1.4485, dan UCIQE 0.2389.

 

Downloads

Download data is not yet available.

Referensi

CHAKRABORTY, T. et al. 2024 ‘Ten years of generative adversarial nets (GANs): a survey of the state-of-the-art’, Machine Learning: Science and Technology, 5(1), pp. 1–29. Available at: https://doi.org/10.1088/2632-2153/ad1f77.

CONG, R., YANG, N., et al. 2023 ‘Global-and-Local Collaborative Learning for Co-Salient Object Detection’, IEEE Transactions on Cybernetics, 53(3), pp. 1920–1931. Available at: https://doi.org/10.1109/TCYB.2022.3169431.

CONG, R., YANG, W., et al. 2023 ‘PUGAN: Physical Model-Guided Underwater Image Enhancement Using GAN With Dual-Discriminators’, IEEE Transactions on Image Processing, 32, pp. 4472–4485. Available at: https://doi.org/10.1109/TIP.2023.3286263.

FABBRI, C., ISLAM, M.J. AND SATTAR, J. 2018 ‘Enhancing Underwater Imagery Using Generative Adversarial Networks’, Proceedings - IEEE International Conference on Robotics and Automation, pp. 7159–7165. Available at: https://doi.org/10.1109/ICRA.2018.8460552.

HUANG, Z. et al. 2022 ‘Underwater Image Enhancement via Adaptive Group Attention-Based Multiscale Cascade Transformer’, IEEE Transactions on Instrumentation and Measurement, 71, pp. 1–18. Available at: https://doi.org/10.1109/TIM.2022.3189630.

ISLAM, M.J., XIA, Y. AND SATTAR, J. 2020 ‘Fast Underwater Image Enhancement for Improved Visual Perception’, IEEE Robotics and Automation Letters, 5(2), pp. 3227–3234. Available at: https://doi.org/10.1109/LRA.2020.2974710.

KRICHEN, M. 2023 ‘Generative Adversarial Networks’, 2023 14th International Conference on Computing Communication and Networking Technologies, ICCCNT 2023, pp. 1–7. Available at: https://doi.org/10.1109/ICCCNT56998.2023.10306417.

KUMAR, N. et al. 2023 ‘Underwater Image Enhancement using Deep Learning’, Multimedia Tools and Applications, 82(30), pp. 46789–46809. Available at: https://doi.org/10.1007/s11042-023-15525-4.

LI, C. et al. 2020 ‘An Underwater Image Enhancement Benchmark Dataset and beyond’, IEEE Transactions on Image Processing, 29, pp. 4376–4389. Available at: https://doi.org/10.1109/TIP.2019.2955241.

LI, C. et al. 2021 ‘Underwater Image Enhancement via Medium Transmission-Guided Multi-Color Space Embedding’, IEEE Transactions on Image Processing, 30, pp. 4985–5000. Available at: https://doi.org/10.1109/TIP.2021.3076367.

LI, C., GUO, J. AND GUO, C. 2018 ‘Emerging from Water: Underwater Image Color Correction Based on Weakly Supervised Color Transfer’, IEEE Signal Processing Letters, 25(3), pp. 323–327. Available at: https://doi.org/10.1109/LSP.2018.2792050.

MARASAKATLA, D. et al. 2023 ‘Under Water Image Enhancement Using CNN’, IEEE International Conference on Advances in Electronics, Communication, Computing and Intelligent Information Systems, ICAECIS 2023 - Proceedings, pp. 211–217. Available at: https://doi.org/10.1109/ICAECIS58353.2023.10170509.

PENG, W. et al. 2023 ‘RAUNE-Net: A Residual and Attention-Driven Underwater Image Enhancement Method’, 1. Available at: http://arxiv.org/abs/2311.00246.

WANG, Z. et al. 2022 ‘Adaptive feature fusion network based on boosted attention mechanism for single image dehazing’, Multimedia Tools and Applications, 81(8), pp. 11325–11339. Available at: https://doi.org/10.1007/s11042-022-12151-4.

YUE, G. et al. 2022 ‘Boundary Constraint Network With Cross Layer Feature Integration for Polyp Segmentation’, IEEE Journal of Biomedical and Health Informatics, 26(8), pp. 4090–4099. Available at: https://doi.org/10.1109/JBHI.2022.3173948.

ZAIDI, S., SINGH, P. AND GUHA, P. 2024 ‘Designing a U-Net Architecture for Underwater Image Enhancement’, 2024 National Conference on Communications, NCC 2024, pp. 1–6. Available at: https://doi.org/10.1109/NCC60321.2024.10485836.

ZHOU, J. et al. 2023 ‘Multi-view underwater image enhancement method via embedded fusion mechanism’, Engineering Applications of Artificial Intelligence, 121(February), p. 105946. Available at: https://doi.org/10.1016/j.engappai.2023.105946.

ZHOU, X. et al. 2023 ‘Edge-Guided Recurrent Positioning Network for Salient Object Detection in Optical Remote Sensing Images’, IEEE Transactions on Cybernetics, 53(1), pp. 539–552. Available at: https://doi.org/10.1109/TCYB.2022.3163152.

Diterbitkan

17-12-2025

Terbitan

Bagian

Ilmu Komputer

Cara Mengutip

Peningkatan Kualitas Citra Bawah Air Menggunakan GAN dengan Mekanisme Residual dan Attention. (2025). Jurnal Teknologi Informasi Dan Ilmu Komputer, 12(6), 1281-1290. https://doi.org/10.25126/jtiik.2025126