《超分辨率重建技术研究--山东大学博士学位论文.doc》由会员分享，可在线阅读，更多相关《超分辨率重建技术研究--山东大学博士学位论文.doc（147页珍藏版）》请在金锄头文库上搜索。
1、山东大学博士学位论文目 录中文摘要1ABSTRACT5缩略语简表9第一章 绪论111.1概述111.2 国内外研究现状121.2.1 基于重构的超分辨率技术131.2.2 基于学习的超分辨率技术161.3 超分辨率技术的应用171.4 本论文的研究内容和主要贡献191.4.1多帧图像超分辨率重建技术研究191.4.2单帧超分辨率重建与图像增强技术研究201.4.3盲超分辨率重建技术研究211.4.4基于ICA的人脸幻想算法研究21第二章 图像及视频序列的超分辨率重建技术232.1引言232.2基本概念232.3数学模型252.4实现超分辨率的可行性分析282.5基本算法312.5.1基于重构的
2、多帧超分辨率算法312.5.2基于学习的单帧超分辨率算法382.5.3超分辨率重建图像质量评价402.6 小结41第三章 多帧超分辨率图像重建433.1引言433.2 基于平稳小波变换的超分辨率重建方案443.2.1平稳小波变换443.2.2 基于SWT的超分辨率重建453.2.3 规整化参数的估计473.2.4 仿真试验及分析473.3基于峭度的超分辨率重建方案503.3.1峭度图像的定义及性质503.3.2低噪环境下的超分辨率重建533.3.3仿真试验及分析563.4基于分割的超分辨率重建方案603.4.1基于高阶统计的图像分割603.4.2具有边缘保持特性的超分辨率重建613.4.3仿真
3、试验及分析633.5 小结65第四章 基于学习的单帧超分辨率图像重建674.1 引言674.2流形学习684.2.1流形学习的基本理论684.2.2流形学习的基本算法704.2.3流形学习与超分辨率的内在联系724.3光照变化下的低分辨率图像获取模型734.4基于自商图像的超分辨率重建744.4.1自商图像744.4.2 基于SQI超分辨率重建754.4.3仿真试验及分析774.5基于对数小波变换的超分辨率重建794.5.1对数小波变换804.5.2基于Log-WT的超分辨率重建814.5.3 仿真试验及分析834.6 小结84第五章 盲超分辨率图像重建875.1引言875.2参数化模糊函数模
4、型875.3基于矢量量化的盲超分辨率重建885.3.1矢量量化895.3.2模糊参数辨识915.3.3盲超分辨率图像重建945.3.4仿真试验及分析955.4基于支撑向量机的盲超分辨率重建995.4.1支撑向量机995.4.2模糊辨识及超分辨率重建1005.4.3仿真试验及分析1035.5两种方法的性能分析1065.6小结108第六章 基于独立分量分析的人脸超分辨率重建1096.1 引言1096.2 独立分量分析1106.2.1 数学模型1116.2.2 FastICA算法1126.3统计独立的基图像1136.4基于ICA的人脸超分辨率重建1146.5仿真试验及分析1166.7 小结119第七
5、章 结论与展望120附 录123附录A：基于总变分的图像去噪算法123附录B：基于最小二乘的运动估计算法124参考文献126致 谢139攻读学位期间发表的论文及专利141附件：外文论文两篇143CONTENTSAbstract1Abbreviation List9Chapter 1 Introduction111.1 Background111.2 Previous Work121.2.1 Reconstruciton-based SR Technique131.2.2 Learning-based SR Technique161.3 Applications171.4 Main Contri
6、butions191.4.1 Multi-frame SR Technique191.4.2 Single-frame SR and Image Enhancement Technique201.4.3 Blind SR Technique 211.4.4 ICA-based Face Hallucination Technique21Chapter 2 Image/Video Super-Resolution Reconstruction Technique232.1 Introduction232.2 Basic Concept232.3 Mathematical Model252.4 F
7、easibility Analysis of SR282.5 Classical Algorithm312.5.1 Reconstruciton-based Multi-frame SR Technique312.5.2 Learning-based Single-frame SR Technique382.5.3 Quality Avaluation402.6 Summary41Chapter 3 Multi-frame SR Reconstruction433.1 Introduction433.2 SR Reconsruction Based on Stationary Wavelet
8、Transform443.2.1 Stationary Wavelet Transform443.2.2 SWT Based SR Algorithm453.2.3 Estimation of the Regularization Parameter473.2.4 Experiments and Analysis473.3 SR Reconsruction Based on Kurtosis503.3.1 Definition and Characters of the Kurtosis Image503.3.2 SR Reconsruction in Low SNR Environment5
9、33.3.3 Experiments and Analysis563.4 SR Reconsruction Based on Image Segmentation603.4.1 Image Segmentation Based on High Order Statistics603.4.2 SR Reconsruction with Edge Preservation613.4.3 Experiments and Analysis633.5 Summary65Chapter 4 Learning-based Single-frame SR Reconsruction674.1 Introduc
10、tion674.2 Manifold Learning684.2.1 Basic Theory684.2.2 Classical Algorithm704.2.3 Inherent Relation Between Manifold Learning and SR724.3 Imaging Model with Illumination Changing734.4 SR Reconsruction Based on Self Quotient Image744.4.1 Self Quotient Image744.4.2 SQI Based SR Algorithm754.4.3 Experi
11、ments and Analysis774.5 SR Reconsruction Based on Logarithmic-Wavelet Transform794.5.1 Logarithmic-Wavelet Transform804.5.2 Log-WT Based SR Algorithm814.5.3 Experiments and Analysis834.6 Summary84Chapter 5 Blind Super-Resolution Reconstruction875.1 Introduction875.2 Parameterized Model of Blur Funct
12、ion875.3 Blind SR Based on Vector Quantization885.3.1 Vector Quantization895.3.2 Blur Identification915.3.3 VQ Based Blind SR Reconstruction945.3.4 Experiments and Analysis955.4 Blind SR Reconstruction Based on Support Vector Machines995.4.1 Support Vector Machines995.4.2 Blur Identification and SR
13、Reconstruction1005.4.3 Experiments and Analysis1035.5 Performance Analysis of the Two Algorithms1065.6 Summary108Chapter 6 Face SR Based on Independent Component Analysis1096.1 Introduction1096.2 Independent Component Analysis1106.2.1 Mathematical Model1116.2.2 FastICA Algorithm1126.3 Independent Im
14、age Basis1136.4 Face SR Reconstruction Based on ICA1146.5 Experiments and Analysis1166.7 Summary119Chapter 7 Conclusions and Future Works120Appendix123Appendix A: Image Noise Renoval Based on Total Variation123Appendix B: Motion Estimation Based on Least Square124References126Acknowledgements139Publications and Patents During the Research141Annex: Two Papers Related to This Research143141中文摘要随着多媒体通信和信号处理技术的发展，人们越来越需要高分辨率的图像。然而许多成像系统受其物理条件的限制，所得到的图像往往会发生退化变形和受噪声污染，因而分辨率不可能很
