MAULUDDIN, MUHAMMAD HIDAYAT and Supardi, Julian (2025) KLASIFIKASI DUA TAHAP PADA IMAGE DEBLURRING WAJAH BERDASARKAN JENIS DAN TINGKAT KEPARAHAN BLUR DENGAN CONVOLUTIONAL NEURAL NETWORKS DAN U-NET. Masters thesis, Sriwijaya University.
![[thumbnail of RAMA_55101_09012682327010_cover.jpg]](https://repository.unsri.ac.id/179282/17/RAMA_55101_09012682327010_cover.jpg) |
Image
RAMA_55101_09012682327010_cover.jpg Download (329kB) |
![[thumbnail of RAMA_55101_09012682327010.pdf]](http://repository.unsri.ac.id/style/images/fileicons/text.png) |
Text
RAMA_55101_09012682327010.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (7MB) |
|
Text
RAMA_55101_09012682327010_TURNITIN.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (15MB) |
|
Text
RAMA_55101_09012682327010_0010077210_01_front_ref.pdf - Accepted Version Available under License Creative Commons Public Domain Dedication. Download (881kB) |
|
Text
RAMA_55101_09012682327010_0010077210_02.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (239kB) |
|
Text
RAMA_55101_09012682327010_0010077210_03.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (362kB) |
|
Text
RAMA_55101_09012682327010_0010077210_04.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (965kB) |
|
Text
RAMA_55101_09012682327010_0010077210_05.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (10kB) |
|
Text
RAMA_55101_09012682327010_0010077210_06_ref.pdf - Bibliography Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (77kB) |
|
Text
RAMA_55101_09012682327010_0010077210_07_lamp.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (5MB) |
Abstract
Image blur, particularly in facial images with varying types and severity levels, can significantly degrade visual quality and hinder the performance of facial recognition systems. This study proposes a two-stage classification approach for image deblurring based on Convolutional Neural Networks (CNN) and U-Net. The approach separates the classification of blur type and blur severity before applying a specialized deblurring model tailored for each identified combination. The three main blur types addressed are Gaussian Blur, Motion Blur, and Average Blur, each divided into five severity levels. Each deblurring model is independently developed according to the identified blur category through the two-stage classification system. Experimental results show that this separated approach significantly improves image restoration quality especially at low to moderate severity levels compared to conventional methods that use a single model for all severity levels. The average PSNR and SSIM scores, which reached 30.946 and 0.898 respectively, confirm the effectiveness of this strategy. Furthermore, the proposed framework enables more adaptive and specific processing tailored to the characteristics of the blurred image. In conclusion, integrating two-stage classification with deblurring model separation based on blur type and severity has been shown to enhance overall image restoration performance.
| Item Type: | Thesis (Masters) |
|---|---|
| Subjects: | T Technology > T Technology (General) > T1-995 Technology (General) |
| Divisions: | 09-Faculty of Computer Science > 55101-Informatics (S2) |
| Depositing User: | MUHAMMAD HIDAYAT MAULUDDIN |
| Date Deposited: | 20 Jul 2025 13:22 |
| Last Modified: | 20 Jul 2025 13:22 |
| URI: | http://repository.unsri.ac.id/id/eprint/179282 |
Actions (login required)
 |
View Item |