Publications
Detailed Information
Single Image Super-Resolution using Spatial LSTM : 공간적 LSTM을 사용한 단일 영상 초해상도 기법
DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | 이경무 | - |
dc.contributor.author | 이정권 | - |
dc.date.accessioned | 2017-07-14T02:44:56Z | - |
dc.date.available | 2017-07-14T02:44:56Z | - |
dc.date.issued | 2017-02 | - |
dc.identifier.other | 000000142368 | - |
dc.identifier.uri | https://hdl.handle.net/10371/122860 | - |
dc.description | 학위논문 (석사)-- 서울대학교 대학원 : 전기·정보공학부, 2017. 2. 이경무. | - |
dc.description.abstract | In this thesis, a single image super-resolution method (SR) using spatial long short term memory (SR-SLSTM) is proposed. For a highly ill-posed problem like SR, the
receptive field size determining the amount of contextual information is the most important factor to solve the problem. While convolutional layer is limited to enlarge the receptive field size to gather more surrounding information and yields local features, spatial LSTM efficiently increases the receptive field to an image-level within a single layer and yields global image-level features. SR-SLSTM utilizes both local and global features using spatial LSTM residual unit at the same time to predict highly accurate high-resolution (HR) image from low-resolution image (LR). SR-SLSTM shows state -of-the-art results for SR benchmarks and especially yields better results from images with structural objects which benefit from contextual information. That result convince our proposed method using spatial LSTM efficiently extracts contextual information to handle a ill-posed problem like SR. | - |
dc.description.tableofcontents | 1 Introduction 1
1.1 Background and Research Issues 1 1.2 Outline of the Thesis 3 2 Related work 6 2.1 Super-Resolution using Deep Learning 6 2.1.1 Super-Resoltuion Convolutional Neural Network 6 2.1.2 Super-Resolution Neural Networks with larger receptive field 8 2.2 Spatial LSTM in computer vision 8 2.3 Contributions 10 3 Proposed method 11 3.1 Spatial LSTM unit for Super-Resolution 11 3.1.1 Long Short Term Memory 11 3.1.2 Basic Spatial LSTM 12 3.1.3 Spatial LSTM residual unit 16 3.2 Network Architecture 17 3.3 Training 20 4 Discussion 21 5 Experimental results 22 5.1 Datasets 22 5.2 Training Setup 23 5.3 Quantative Results 23 5.4 Qualitative Results 26 5.4.1 Natural images 26 5.4.2 Structural images 30 6 Conclusion 36 6.1 Summary of the Thesis 36 6.2 Future directions 36 6.2.1 Generative Modeling 36 6.2.2 Discrete Prediction 37 6.2.3 Using High-level Information 37 Bibliography 39 Abstract (In Korean) 44 | - |
dc.format | application/pdf | - |
dc.format.extent | 38909471 bytes | - |
dc.format.medium | application/pdf | - |
dc.language.iso | en | - |
dc.publisher | 서울대학교 대학원 | - |
dc.subject | Super-resolution | - |
dc.subject | Ill-posed problem | - |
dc.subject | Contextual information | - |
dc.subject | Receptivel field | - |
dc.subject | Spatial LSTM | - |
dc.subject.ddc | 621 | - |
dc.title | Single Image Super-Resolution using Spatial LSTM | - |
dc.title.alternative | 공간적 LSTM을 사용한 단일 영상 초해상도 기법 | - |
dc.type | Thesis | - |
dc.contributor.AlternativeAuthor | Lee Jung Kwon | - |
dc.description.degree | Master | - |
dc.citation.pages | 44 | - |
dc.contributor.affiliation | 공과대학 전기·정보공학부 | - |
dc.date.awarded | 2017-02 | - |
- Appears in Collections:
- Files in This Item:
Item View & Download Count
Items in S-Space are protected by copyright, with all rights reserved, unless otherwise indicated.