Docstoc

UBC Computer Science at UBC

Document Sample
UBC Computer Science at UBC Powered By Docstoc
					Depth Perception
and Visualization
   Matt Williams




                    From: http://www.cs.washington.edu/homes/cassidy/tele/index.html
Depth Perception
and Visualization
   References and borrowed images:
   Ware, C., Chapter 8 of Information Visualization: Perception for Design. 2000, San
    Fancisco: Morgan Kaufmann.
   J.D. Pfautz, Depth Perception in Computer Graphics, Doctoral Dissertation, University of
    Cambridge, UK, 2000.
   C. Ware, C. Gobrecht, and M.A. Paton, "Dynamic Adjustment of Stereo Display Parameters,"
    IEEE Transactions on Systems, Man and Cybernetics---Part A: Systems and Humans, Vol. 28, No. 1,
    Jan. 1998, pp. 56-65.
   www.wlu.ca/~wwwpsych/tsang/8Depth.ppt(no author provided)
   Robertson,G.,Mackinlay,J.,&Card,S.ConeTrees: Animated 3D visualizations of hierarchical
    information. In Proceedings of CHI'91 (New Orleans, LA), ACM, 189-194.
   WANGER, L., FERWANDA, J., AND GREENBERG, D. 1992. Perceiving spatial relationships in
    computer generated images. IEEE Computer Graphics and Applications (May) 44-58.
Depth Perception
and Visualization
     Depth Perception
         Cues
         How do we combine these cues to perceive
          depth
     InfoVis Application
         Which cues are helpful?
         Which cues may be important in your
          project?
Depth Cues
      Monocular
        Perspective Cues
        Size

        Occlusion

        Depth of Focus

        Cast Shadows

        Shape from Motion

      Binocular
        Eye Convergence
        Stereoscopic depth
Structure from Motion
    Motion Parallax
    Kinetic Depth
      n

                                                                                               b
                                               a




                                                                  c

             Ware, C., Chapter 8 of Information Visualization: Perception for Design. 2000, San Fancisco: Morgan Kaufmann.
Structure from Motion
       Kinetic Depth Effect
       Assumption of rigidity allows us to
        assume shape as objects
        move/rotate
                                                                                 b
                                 a




                                                    c

       Ware, C., Chapter 8 of Information Visualization: Perception for Design. 2000, San Fancisco: Morgan Kaufmann.
Perspective Cues
   Parallel lines converge
   Distant objects appear smaller
   Textured Elements become smaller
    with distance




             Ware, C., Chapter 8 of Information Visualization: Perception for Design. 2000, San Fancisco: Morgan Kaufmann.
Perspective Cues




http://www.wlu.ca/~wwwpsych/tsang/8Depth.ppt
Perspective Cues
   Taking advantage of linear perspective
    in visualization




             Ware, C., Chapter 8 of Information Visualization: Perception for Design. 2000, San Fancisco: Morgan Kaufmann.
 Perspective Cues
               Size Constancy
               Perception of actual size versus retinal size.
               Can perceive 2D picture plane size for sketchy
                images (see below)




http://www.wlu.ca/~wwwpsych/tsang/8Depth.ppt
Perspective Cues




  http://www.wlu.ca/~wwwpsych/tsang/8Depth.ppt
Perspective Cues




  http://www.wlu.ca/~wwwpsych/tsang/8Depth.ppt
Perspective Cues
    Usually we percieve images on the
     computer from the wrong viewpoint
    Robustness of linear perspective (Kubovy, 1986)
        e.g Movie Theatre




    Why might we want to correct for
     viewpoint changes (head movement)
     anyway?

                   Motion Parallax
                   Placement of virtual hand or object
Perspective Cues
     Placement of virtual hand or object
     Need for head coupled perspective




             vrlab.postech.ac.kr/vr/gallery/edu/vr/display.ppt
  Occlusion
                 The strongest depth cue.




http://www.wlu.ca/~wwwpsych/tsang/8Depth.ppt

                                      Ware, C., Chapter 8 of Information Visualization: Perception for Design. 2000, San Fancisco: Morgan Kaufmann
Depth of Focus
     Strong Depth Cue
     Must be coupled with user input (e.g.
      point of fixation)
     Computationally expensive




          Ware, C., Chapter 8 of Information Visualization: Perception for Design. 2000, San Fancisco: Morgan Kaufmann.
    Cast Shadows
   Important cue for height of an object above a
    plane
   An indirect depth cue
   Shown to be stronger than size perspective
    (Kersten, 1996)




            Ware, C., Chapter 8 of Information Visualization: Perception for Design. 2000, San Fancisco: Morgan Kaufmann.
      Shape From Shading




   Ware, C., Chapter 8 of Information Visualization: Perception for Design. 2000, San Fancisco: Morgan Kaufmann.


                                                                                  http://www.wlu.ca/~wwwpsych/tsang/8Depth.ppt
                           Ware Chapter 7
Eye Convergence




      Better for relative depth than for
       absolute depth
              Ware, C., Chapter 8 of Information Visualization: Perception for Design. 2000, San Fancisco: Morgan Kaufmann.
Stereoscopic Depth
   How it works
   Two different views fuse to one
    perceived view (try it)




          Ware, C., Chapter 8 of Information Visualization: Perception for Design. 2000, San Fancisco: Morgan Kaufmann.
Stereoscopic Depth
   Panum’s fusional area
   Range before diplopia occurs(worst case):
       Fovea – 1/10 of a degree (3 pixels)
       Periphery – 1/3 of a degree (10 pixels)
   Factors for Fusion
       Moving images
       Blurred images
       Size
       Exposure
    Stereoscopic Depth




velab.cau.ac.kr/lecture/Stereo.ppt
Stereoscopic Depth
   Problems with stereoscopic displays
   Diplopia occurs when images don’t fuse (try it)
       Diplopia reduced for blurred images – great for the real
        world but …
       Stereoscopic displays only contain sharp images. Close-
        up unattended items can be obtrusive.

   Vergence Focus Problem
       Everything on the computer screen is on the same focal
        plane.
       Causes eyestrain
   Frame Cancellation:
Stereoscopic Depth
    Frame Cancellation:




           Ware, C., Chapter 8 of Information Visualization: Perception for Design. 2000, San Fancisco: Morgan Kaufmann.




    Solution?
Stereoscopic Displays
       Cyclopean Scale
              Move virtual environment close to the
               display plane
                   NoCancellation
                   Reduced Vergence-focus problem




       Ware, C., Chapter 8 of Information Visualization: Perception for Design. 2000, San Fancisco: Morgan Kaufmann.
    Stereoscopic Displays
   Virtual Eye Separation
    (Telestereoscope)
   Allows for a decrease or
    increase in disparity
   Allows for an increase or
    decrease in the depth of
    the virtual environment




           http://www.cs.washington.edu/homes/cassidy/tele/index.html
Depth Perception Theory
    General Unified Theory
        Perceived Depth = Weighted sum of all Depth Cues
        Rank the cues in importance
        e.g.
             Occlusion
             Motion Parallax
             Stereo
             Size constancy
             Etc.
Depth Perception Theory
          Importance changes with distance

                     Motion
                     parallax
                                                            Occlusion

                                    Cast Shadows
  Depth Contrast




                                                          Size constancy
                                          , 96
                                 Stereo
                                          , 96
                   Convergence

                                                 Aerial

                       1            10              100
                   Depth (meters)                             Cutting, 1996
Space Perception Theory
    Task Dependant Model
       Cues weights are combined differently based on the task

       Evidence?

            Task: Orientation of a virtual Object
               • Cast Shadows and Motion Parallax help
               • But …Linear Perspective hinders such orientation
            Task: Object translation
               • Linear perspective was the most useful cue




                                                                Wanger, 1992
InfoVis Tasks:
      Tracing 3D data paths
      Judging 3D surfaces
      Finding 3D patterns of points
      Relative Position in 3D space
      Judging movement of Self
      Judging Up Direction
      Feeling a “sense of Presence”
      Tracing 3D Data Paths
                       Benefits of 3D Trees
                              More nodes can be displayed (Robertson et al.,
                               1993)

                              Reduced errors in detecting Paths
                               (Sollenberger and Milgram, 1993)




   Ware, C., Chapter 8 of Information Visualization: Perception for Design. 2000, San Fancisco: Morgan Kaufmann.
Tracing 3D Data Paths
     Beneficial Cues:
       Kinetic Depth and Stereoscopic
        Depth reduced errors in path
        detection
       Kinetic Depth was the stronger cue

       Occlusion Is helpful
         (Ware and Franck, 1996)
         3D Patterns of Points




http://neutrino.kek.jp/~kohama/sarupaw/sarupaw_html/fig/nt_3d.gif   http://www-pat.fnal.gov/nirvana/plot_wid.html
3D Patterns of Points
     Beneficial Cues:
       Structure from motion
       Stereo Depth

     Not Beneficial:
       Perspective
       Size

       Cast Shadows

       Shape from Shading (How?)
              3D Patterns of Points
                          Add shape to clouds of points




   Ware, C., Chapter 8 of Information Visualization: Perception for Design. 2000, San Fancisco: Morgan Kaufmann.
            Judging Relative Position

   Small Scale (Threading a needle)
       Beneficial: Stereo
       Not Beneficial: Motion Parallax
   Large Scale ( > 30 m)
       Beneficial: motion parallax,
        perspective, cast shadows, texture
        gradients
       Not Beneficial: stereo


                       Ware, C., Chapter 8 of Information Visualization: Perception for Design. 2000, San Fancisco: Morgan Kaufmann.
Conclusion      

                
                     Depth Cues
                     Existing Theories
                    Application to InfoVis

                                                                       Occlusion
                                                                       Texture Gradient
                                                                       Size Constancy
                                                                       Cast Shadows
                                                                       Stereo




             From: http://www.cs.washington.edu/homes/cassidy/tele/index.html

				
DOCUMENT INFO