megastereo
play

Megastereo: Constructing High-Resolution Stereo Panoramas 1 - PowerPoint PPT Presentation

Megastereo: Constructing High-Resolution Stereo Panoramas 1 Christian Richardt 1,2 Yael Pritch 1 2 Henning Zimmer 1,3 3 Alex Sorkine-Hornung 1 1 2 3 Structure of my talk 1. capturing stereoscopic panoramas 2. our image alignment pipeline


  1. Megastereo: Constructing High-Resolution Stereo Panoramas 1 Christian Richardt 1,2 Yael Pritch 1 2 Henning Zimmer 1,3 3 Alex Sorkine-Hornung 1 1

  2. 2

  3. 3

  4. Structure of my talk 1. capturing stereoscopic panoramas 2. our image alignment pipeline 3. our flow-based ray interpolation 4. results + live demo 5

  5. How to capture stereo panoramas? 6

  6. How to capture stereo panoramas? 6

  7. How to capture stereo panoramas? 10

  8. How to capture stereo panoramas? 10

  9. How to capture stereo panoramas? 10

  10. Omnistereo panoramas [Peleg et al., 2001] 11

  11. Omnistereo panoramas [Peleg et al., 2001] 12

  12. Omnistereo panoramas [Peleg et al., 2001] 12

  13. Omnistereo panoramas [Peleg et al., 2001] 13

  14. Omnistereo panoramas [Peleg et al., PAMI 2001] (our result) 14

  15. Related work panoramas: stereo panoramas: – Szeliski (2006) – Hum & He (1999) – Brown & Lowe (2007) – Peleg et al. (2001) generalised cameras: image alignment: – Gupta & Hartley (1997) – Lucas & Kanade (1981) – Zomet et al. (2003) – Snavely et al. (2006) – Yu & McMillan (2004) – Wu et al. (2011) multi-perspective: image stitching: – Agarwala et al. (2006) – Burt & Adelson (1983) – Rav-Acha et al. (2008) – Shum & Szeliski (2000) – Yu et al. (2010) – Kang et al. (2004) – Kopf et al. (2010) – Levin et al. (2004) 15

  16. Motivation unresolved practical issues in stereo panoramas: visible seams misalignment vertical parallax this is unpleasant in 2D, but intolerable in stereo 16

  17. Contributions a general and efficient solution for creating high-quality, high-resolution (stereo) panoramas revised image stabilisation and alignment: correcting camera orientations removing undesired vertical parallax interpolating continuous ray space from discrete views: resolving sampling artefacts virtually & on the fly 17

  18. Structure of my talk 1. capturing stereoscopic panoramas 2. our image alignment pipeline 3. our flow-based ray interpolation 4. results + live demo 18

  19. Input video circular motion challenging: hand-held 720 × 1280 (Canon S95) 19

  20. Image alignment: comparison image-based alignment our alignment approach 20

  21. Image alignment: comparison image-based alignment our alignment approach 20

  22. Raw input video input video omnistero panorama (crop) 21

  23. Lens undistortion undistorted images omnistero panorama (crop) 22

  24. Orientation stabilisation stabilised images omnistero panorama (crop) 23

  25. Vertical parallax cancellation compensated images omnistero panorama (crop) 24

  26. Structure of my talk 1. capturing stereoscopic panoramas 2. our image alignment pipeline 3. our flow-based ray interpolation 4. results + live demo 25

  27. Without strip blending far: duplication ‘refaim’ dataset near: truncation [Rav-Acha et al., 2008] 26

  28. Linear strip blending far: duplication ‘refaim’ dataset near: truncation [Rav-Acha et al., 2008] 27

  29. Duplication + truncation panoramic imaging surface far objects near objects 28

  30. Our flow-based ray interpolation panoramic imaging surface far objects near objects 29

  31. Our flow-based ray interpolation panoramic imaging surface far objects near objects 29

  32. Without strip blending far: duplication ‘refaim’ dataset near: truncation [Rav-Acha et al., 2008] 30

  33. Linear strip blending far: duplication ‘refaim’ dataset near: truncation [Rav-Acha et al., 2008] 31

  34. Our flow-based blending far: stretching ‘refaim’ dataset near: squeezing [Rav-Acha et al., 2008] 32

  35. Blending comparison no blending 33

  36. Blending comparison linear blending 34

  37. Blending comparison our flow-based blending 35

  38. Structure of my talk 1. capturing stereoscopic panoramas 2. our image alignment pipeline 3. our flow-based ray interpolation 4. results + live demo 36

  39. 360 º zoom 37

  40. 360 º zoom 38

  41. Street panorama (linear motion) ‘refaim’ dataset [Rav-Acha et al., 2008] 39

  42. 360 º 140 MP stereo panorama 100% zoom 40

  43. Live demo 41

  44. Conclusion a general and efficient solution for creating high-quality, high-resolution stereo panoramas Future work: extension to more general multi-perspective images handling changing exposures stereo panorama videos richardt.name/megastereo disneyresearch.com/project/megastereo/ 42

  45. please see Conclusion my poster a general and efficient solution for creating high-quality, high-resolution stereo panoramas Future work: extension to more general multi-perspective images handling changing exposures stereo panorama videos richardt.name/megastereo disneyresearch.com/project/megastereo/ 42

Recommend


More recommend