UID:
edoccha_9960885586002883
Umfang:
1 online resource (vii, 214 pages) :
,
illustrations.
Serie:
Karlsruher Institut für Technologie
Inhalt:
In this work, spatio-spectrally coded multispectral light fields, as taken by a light field camera with a spectrally coded microlens array, are investigated. For the reconstruction of the coded light fields, two methods, one based on the principles of compressed sensing and one deep learning approach, are developed. Using novel synthetic as well as a real-world datasets, the proposed reconstruction approaches are evaluated in detail.
Anmerkung:
Preface i -- Nomenclature v -- 1 Introduction . 1 -- 1.1 Contributions . 7 -- 1.2 General remarks 10 -- 1.3 Thesis outline . 11 -- 2 Spectral Light Fields 13 -- 2.1 Light field acquisition 15 -- 2.2 Light field applications 18 -- 2.3 Coded light fields . 20 -- 2.3.1 Spatio-spectrally coded light fields . 21 -- 2.4 Light field data 23 -- 3 Reconstruction from Coded Light Fields 25 -- 3.1 Compressed sensing-based reconstruction . 25 -- 3.1.1 Fixed basis-based reconstruction 29 -- 3.1.2 Dictionary-based reconstruction 32 -- 3.2 Principal reconstruction via multi-task deep learning 38 -- 3.2.1 Related work . 39 -- 3.2.2 Network architectures 41 -- 3.2.3 Training strategies 50 -- 3.3 Mask optimization via neural fractals . 55 -- 4 Experimental Setup . 65 -- 4.1 Synthetic dataset . 65 -- 4.1.1 Dataset properties 66 -- 4.1.2 Random scene generation 68 -- 4.1.3 Challenges . 71 -- 4.2 Real-world dataset 74 -- 4.2.1 Spectral light field camera 77 -- 4.2.2 Radiometric calibration . 79 -- 4.2.3 Geometric calibration 85 -- 4.2.4 Light field decoding . 101 -- 4.3 Evaluation metrics 104 -- 4.4 Training and implementation details 109 -- 4.4.1 Loss functions and weights . 109 -- 4.4.2 Mask generation, augmentation, and seeding . 112 -- 4.4.3 Training and implementation 116 -- 5 Results 119 -- 5.1 Compressed sensing-based reconstruction . 120 -- 5.2 Principal reconstruction . 123 -- 5.2.1 Adaptive auxiliary loss approaches 124 -- 5.2.2 Adaptive multi-task approaches 128 -- 5.2.3 Reconstruction using random coding masks . 129 -- 5.2.4 Noise, angular resolution, and depth 137 -- 5.3 Mask optimization 144 -- 5.3.1 Predefined coding masks 144 -- 5.3.2 End-to-end optimized coding masks 150 -- 6 Conclusion 157 -- 6.1 Summary 157 -- 6.2 Limitations and outlook . 160 -- A Light Field Camera Depth and Disparity . 167 -- B Geometric Calibration Evaluation . 173 -- B.1 Parameter choice . 173 -- B.2 Evaluation metrics 174 -- B.3 Results . 176 -- C Spectral Light Field Camera Technical Details . 183 -- Bibliography . 189 -- List of publications . 212 -- List of supervised theses 213.
Weitere Ausg.:
ISBN 1000148072
Sprache:
Englisch
DOI:
10.5445/KSP/1000148072
Bookmarklink