Super-Resolution Image Restoration from Image Sequence

超解析度有許多種類的應用，為了要使低解析度的內容呈現在高解析度的顯示器上，市場對於超解析度的演算法有著迫切需求。本論文中提出了一個基於凸集投影的超解析度演算法。在超解析度處理中，會給定一個低品質的影像序列，而輸出我們想要得到的高品質影像。我們採用頻率域方法來估計畫面間的運動位移，也使用對數影像處理來增強輸出影像的結果。The field of super-resolution has a wide area of applications. In order to display relatively low-quality content on high-resolution displays, the need for super resolution algorithms has become an urgent market priority. A method of super-resolution based on project-onto-convex-sets (POCS) is proposed in this thesis. In the super-resolution process, a set of low-quality images is given, and a single improved-resolution image is desired. We adopt frequency-domain method to estimate motion and enhance the result of high-resolution image by logarithmic image processing model.口試委員會審定書 i 誌謝 ii 中文摘要 iii ABSTRACT iv CONTENTS v LIST OF FIGURES vii Chapter 1 Introduction 1 1.1 Motivation 1 1.2 Super Resolution 1 1.3 Single-Frame Restoration 4 1.4 Multi-Frame Restoration 4 1.5 Applications of Super-Resolution 5 Chapter 2 Related Works 6 2.1 How Is Super-Resolution Possible? 6 2.2 Observation Model 7 2.2.1 Down-sampling 8 2.2.2 Warping 9 2.2.3 Blurring 10 2.3 The Point Spread Function 11 2.4 Sub-Pixel Shift 12 Chapter 3 State-of-the-Art SR Methods 14 3.1 Frequency Domain Methods 14 3.2 Spatial Domain Methods 16 3.2.1 Bicubic Interpolation 16 3.2.2 Iterative Back-Projection Techniques 17 3.2.3 Projection Onto Convex Sets 18 Chapter 4 Methodology 20 4.1 Overview 20 4.2 Motion Estimation 21 4.3 Project onto Convex Sets 23 4.3.1 Up-Sample 24 4.3.2 Shift the Up-Sampled Low-Resolution Images 24 4.3.3 Combine the Up-Sampled Images. 25 4.3.4 Linear Filtering and Iteration until It Converges 26 4.4 Image Enhancement 27 Chapter 5 Experimental Results 29 5.1 Overview 29 5.2 Experimental Environment 29 5.3 Evaluation Results 30 Chapter 6 Conclusion and Future Work 48 REFERENCES 4