Title Page
Contents
Abstract 8
1. Introduction 10
2. Related works 14
2.1. Depth Video Acquisition Methods 14
2.1.1. Passive methods 14
2.1.2. Active Methods 16
2.1.3. Hybrid methods 17
2.2. Depth Map Estimation in DERS 19
2.2.1. Depth Map Features 19
2.2.2. Depth Map Representation in DERS 19
3. Depth Video Relabeling Algorithm 22
3.1. Previous Relabeling Algorithm 22
3.2. Proposed Relabeling Method 24
3.3. Comparison among Subjective Quality Assessment, PSNR and VQM 28
3.3.1. Mean Opinion Score (MOS)and Absolute Category Rating (ACR) 28
3.3.2. Video Quality Metric (VQM) 29
3.3.3. Experimental Results 30
3.4. Experimental Results 32
3.4.1. Experimental Setup 32
3.4.2. Objective Evaluation (PSNR) 32
3.4.3. 3D video quality evaluations 33
3.4.4. Subjective evaluations 34
4. Depth Video Filtering 36
4.1. Conventional Depth Video Filtering 36
4.1.1. Sobel Edge Detection 36
4.2.1. Bilateral Filter 36
4.2.2. Median Filter 37
4.2. Proposed Two-Tier Filtering 39
4.3. Experimental Results 41
4.3.1. Experimental Setup 41
4.3.2. The objective evaluation 42
4.3.3. The Subjective Evaluation 43
5. Conclusion 45
References 46
논문요약 50
Table 1. Number of labels in each group (20th frame)(이미지참조) 25
Table 2. Experimental sequences 30
Table 3. Experimental results of all sequence and metric 31
Table 4. Correlation coefficient of VQM and PSNR about MOS 31
Table 5. The characteristics of the test sequences 32
Table 6. Synthesis results (PSNR) using depth maps 33
Table 7. Synthesis results (VQM) using depth maps 33
Table 8. Experimental Setup 41
Table 9. Synthesis results (PSNR) using depth maps (100frames) 42
Table 10. Depth map filtering results (complexity) (100frames) 42
Fig 1. Video-plus-Depth representation type of 3D video 10
Fig 2. Block diagram of the depth image based rendering 12
Fig 3. Passive estimation with triangulation methods 14
Fig 4. Captured check board patterns 15
Fig 5. Current commercial TOF cameras (Model/ Resolution) 16
Fig 6. Process of the structured light-based method 17
Fig 7. Hybrid camera system 18
Fig 8. 2D color video and accompanying 8-bit depth-images 19
Fig 9. Array of depth intensity in DERS 21
Fig 10. Grouping the total depth range 22
Fig 11. Weighting the interval based on moving object(G3) 22
Fig 12. Flow chart of the adjustment of parameters 26
Fig 13. Stimulus presentation in the ACR method 28
Fig 14. 5-grade quality scale 29
Fig 15. A compressed still images, 300KB (a) and 700KB (b) 34
Fig 16. Synthesized results for Pantomime 35
Fig 17. Synthesized results for Kendo 35
Fig 18. Block diagram of proposed method 39
Fig 19. Boundary depth map of Ballet sequence 40
Fig 20. Synthesized results for Lovebird1 44
Fig 21. Synthesized results for Newspaper 44