multi perspective panoramas
play

Multi-perspective Panoramas Slides from a talk by Lihi Zelnik-Manor - PowerPoint PPT Presentation

Multi-perspective Panoramas Slides from a talk by Lihi Zelnik-Manor at ICCV07 3DRR workshop Pictures capture memories Panoramas Registration: Brown & Lowe, ICCV05 Blending: Burt & Adelson, Trans. Graphics,1983


  1. Multi-perspective Panoramas Slides from a talk by Lihi Zelnik-Manor at ICCV’07 3DRR workshop

  2. Pictures capture memories

  3. Panoramas Registration: Brown & Lowe, ICCV’05 Blending: Burt & Adelson, Trans. Graphics,1983 Visualization: Kopf et al., SIGGRAPH, 2007

  4. Bad panorama? Output of Brown & Lowe software

  5. No geometrically consistent solution

  6. Scientists solution to panoramas: Single center of projection No 3D!!! Registration: Brown & Lowe, ICCV’05 Blending: Burt & Adelson, Trans. Graphics,1983 Visualization: Kopf et al., SIGGRAPH, 2007

  7. From sphere to plane Distortions are unavoidable

  8. Distorted panoramas Actual appearance Output of Brown & Lowe software

  9. Objectives 1. Better looking panoramas 2. Let the camera move: • Any view • Natural photographing

  10. Stand on the shoulders of giants Cartographers Artists

  11. Cartographic projections

  12. Common panorama projections Perspective Stereographic φ Cylindircal θ

  13. Global Projections Perspective Stereographic Cylindircal

  14. Learn from the artists Sharp Multiple view points discontinuity perspective perspective De Chirico “Mystery and Melancholy of a Street”, 1914

  15. Two horizons!

  16. Renaissance painters solution “School of Athens”, Raffaello Sanzio ~1510 Give a separate treatment to different parts of the scene!!

  17. Personalized projections “School of Athens”, Raffaello Sanzio ~1510 Give a separate treatment to different parts of the scene!!

  18. Multiple planes of projection Sharp discontinuities can often be well hidden

  19. Single view Our multi-view result

  20. Single view Our multi-view result

  21. Single view Our multi-view result

  22. Applying personalized projections Input images Foreground Background panorama

  23. Single view Our multi-view result

  24. Single view Our multi-view result

  25. Objectives - revisited 1. Better looking panoramas 2. Let the camera move: • Any view • Natural photographing Multiple views can live together

  26. Multi-view compositions David Hockney, Place Furstenberg, (1985)

  27. Why multi-view? Multiple viewpoints Single viewpoint David Hockney, Melissa Slemin, Place Furstenberg, 1985 Place Furstenberg, 2003

  28. Multi-view panoramas Single view Multiview Zomet et al. (PAMI’03) Requires video input

  29. Long Imaging Agarwala et al. (SIGGRAPH 2006)

  30. Smooth Multi-View Google maps

  31. What’s wrong in the picture? Google maps

  32. Non-smooth Google maps

  33. The Chair David Hockney (1985)

  34. Joiners are popular Flickr statistics (Aug’07): 4,985 photos matching joiners. 4,007 photos matching Hockney. 41 groups about Hockney Thousands of members

  35. Main goals: Automate joiners Generalize panoramas to general image collections

  36. Objectives • For Artists: Reduce manual labor Manual: ~ 40min. Fully automatic

  37. Objectives • For Artists: Reduce manual labor • For non-artists: Generate pleasing-to-the-eye joiners

  38. Objectives • For Artists: Reduce manual labor • For non-artists: Generate pleasing-to-the-eye joiners • For data exploration: Organize images spatially

  39. What’s going on here?

  40. A cacti garden

  41. Principles

  42. Principles • Convey topology Correct Incorrect

  43. Principles • Convey topology • A 2D layering of images Blending: Graph-cut: Desired joiner blurry cuts hood

  44. Principles • Convey topology • A 2D layering of images • Don’t distort images translate rotate scale

  45. Principles • Convey topology • A 2D layering of images • Don’t distort images • Minimize inconsistencies Bad Good

  46. Algorithm

  47. Step 1: Feature matching Brown & Lowe, ICCV’03

  48. Step 2: Align Large inconsistencies Brown & Lowe, ICCV’03

  49. Step 3: Order Reduced inconsistencies

  50. Ordering images Try all orders: only for small datasets

  51. Ordering images Try all orders: only for small datasets complexity: (m+n) m = # images n = # overlaps = # acyclic orders

  52. Ordering images Observations: – Typically each image overlaps with only a few others – Many decisions can be taken locally

  53. Ordering images Approximate solution: – Solve for each image independently – Iterate over all images

  54. Can we do better?

  55. Step 4: Improve alignment

  56. Iterate Align-Order-Importance

  57. Iterative refinement Initial Final

  58. Iterative refinement Initial Final

  59. Iterative refinement Initial Final

  60. What is this?

  61. That’s me reading

  62. Anza-Borrego

  63. Tractor

  64. Art reproduction Paolo Uccello, 1436

  65. Art reproduction Zelnik & Perona, 2006 Paolo Uccello, 1436

  66. Art reproduction Zelnik & Perona, 2006 Single view-point

  67. Manual by Photographer

  68. Our automatic result

  69. Failure?

  70. GUI

  71. The Impossible Bridge

  72. Homage to David Hockney

  73. Take home • Incorrect geometries are possible and fun! • Geometry is not enough, we need scene analysis • A highly related work: "Scene Collages and Flexible Camera Arrays,” Y. Nomura, L. Zhang and S.K. Nayar, Eurographics Symposium on Rendering, Jun, 2007.

  74. Thank You

  75. 15-463 Class Project from 2007 http://www.cs.cmu.edu/afs/andrew/scs/cs/1 5-463/f07/proj_final/www/echuangs/

Recommend


More recommend