Skripsi

PERBAIKAN KUALITAS CITRA BAWAH AIR MENGGUNAKAN CONVOLUTIONAL NEURAL NETWORK (CNN) DENGAN ARSITEKTUR U-NET

---

Underwater images suffer from degradation caused by light refraction and suspended particles, resulting in the appearance of noise, color casts, low contrast, and loss of fine details, hindering further vision tasks like object detection. This study implements Residual U-Net Model, incorporating Residual blocks onto each convolutional block in the encoder side to improve feature extraction and preserve finer details. Using the EUVP dataset, with a total of 13.678 underwater images, the model is trained using a combined L1 and SSIM Loss Function. Quantitative evaluation for the paired images achieved MSE score of 5.3×10⁻⁵, PSNR of 24.97dB, and SSIM of 0.861. Meanwhile for the unpaired images the model achieved the average UIQM and UCIQE score of 2.849 and 22.17 respectively. Furthermore, Visual assessments and object detection testing also demonstrated improvements in visual clarity and detection accuracy on the reconstructed images, confirming the model’s effectiveness in restoring degraded underwater images.

---

Availability

Detail Information

Series Title

-

Call Number

T1929542026

Publisher

Indralaya : Prodi Teknik Informatika, Fakultas Ilmu Komputer Universitas Sriwijaya., 2026

Collation

xiv, 103 hlm.; ilus.; tab.; 29 cm.

Language

Indonesia

ISBN/ISSN

-

Classification

006.370 7

Content Type

Text

Media Type

-

Carrier Type

-

Edition

-

Subject(s)

Prodi Teknik Informatika
Citra Bawah Air--Convolutional Neural Network

Specific Detail Info

-

Statement of Responsibility

MI

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment