Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor | Department of Electrical and Electronic Engineering | en_US |
| dc.contributor.advisor | Wang, Yi (EEE) | en_US |
| dc.creator | Hu, Qiongyang | - |
| dc.identifier.uri | https://theses.lib.polyu.edu.hk/handle/200/14630 | - |
| dc.language | English | en_US |
| dc.publisher | Hong Kong Polytechnic University | en_US |
| dc.rights | All rights reserved | en_US |
| dc.title | JPEG images restoration with thumbnail guidance and super-resolution | en_US |
| dcterms.abstract | Images are essential for communication, documentation, and artistic expression. However, during storage and transmission, bit errors may occur due to storage or transmission issues, leading to pixel distortion and affecting overall visual quality. Image restoration is crucial for applications in photography, medical imaging, and security. High-quality images enhance the perception of visual content and improve decision-making processes in various fields. As technology advances, the demand for robust restoration methods grows, making this research timely and significant. The primary goal of this thesis is to develop an optimized deep learning framework for image restoration, focusing on addressing challenges posed by bit errors and low resolution images. We first explore the challenges associated with restoring images affected by bitstream damage, highlighting prevalent issues such as loss of detail and the introduction of artifacts. For those type of bit errors, we utilize a three-step Mamba-based thumbnail-guided network designed to restore damaged images effectively. The proposed network features two parallel branches: one for direct image restoration and another utilizing thumbnails as auxiliary information to guide the restoration process. The methodology demonstrates significant improvements in restoration quality, showcasing the effectiveness of combining thumbnail data with deep learning techniques. Second, we address the issue of insufficient image resolution, aiming to enhance the image size while maintaining clarity and detail upon enlargement. We propose a graph-based framework to accomplish the single image super resolution task. The results reveal notable advancements in super-resolution performance, validating the proposed graph-based network's effectiveness in overcoming existing challenges. In summary, this thesis presents a novel dual-branch network to address bit error issues, along with a graph-based approach to tackle critical challenges related to low-resolution images, thereby making significant contributions to the field of image restoration. The findings underscore the potential of advanced deep learning techniques in improving image quality, paving the way for future research and applications in various domains of image processing technology. | en_US |
| dcterms.extent | x, 70 pages : color illustrations | en_US |
| dcterms.isPartOf | PolyU Electronic Theses | en_US |
| dcterms.issued | 2026 | en_US |
| dcterms.educationalLevel | M.Phil. | en_US |
| dcterms.educationalLevel | All Master | en_US |
| dcterms.accessRights | open access | en_US |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/14630