Computer Vision for Interactive Computer Graphics

1
Computer Vision for Interactive Computer Graphics

• Mrudang Rawal

2
Introduction

• Human-computer interaction
• Computers interpret user movements, gestures and
  glances via fundamental visual algorithms.
• Visual algorithms tracking, shape recognition
  and motion analysis
• Interactive apps response time is fast,
  algorithms work for different subject and
  environment, and economical.

3
Tracking Objects

• Interactive applications track objects large
  and small
• Different methods and techniques used.

4
Large Object Tracking

• Large objects like hand or body tracked.
• Object is in front of camera.
• Image properties (Image moments), and artificial
  retina chip do the trick.

5
Step 1 Shape recognition

• Training and Testing of object.
• Technique Orientation Histogram
• Set of each shape oriented in possible direction.
• Match current shape orientation with the ones in
  the set.

6
Step 2 Shape recognition

• Optical flow sense movements gestures
• Frequency of alternation of horizontal and
  vertical velocity (frame avgs) used to determine
  gestures.
• Fast Flow Optical algorithm
• Temporal difference, current previous frame
• If pixel temporal diff ! 0  if -ve motion
  towards adj pixel with greater luminance in
  current frame  if ve towards lower luminance in
  current frame
• Apply the 1-d direction estimation rules to four
  orientations at each pixel
• Average out motion estimates at each pixel, then
  average flow estimate compared to its neighboring
  8 pixels

7
Small Object Tracking

• Large objects tracking techniques not adequate.
• Track small objects through template based
  technique normalized correlation

8
Normalized Correlation

• Examine the fit of an object template to every
  position in the analyzed image.
• The Location of maximum correlation gives the
  position of the candidate hand.
• The value of that correlation indicates how
  likely the image region is to be a hand.

9
Example Television Remote

• To turn on the television, the user holds up his
  hand.
• A graphical hand icon with sliders and buttons
  appears on the graphics display.
• Move hand to control the hand icon

10
Conclusion

• Simple vision algorithms with restrictive
  interactivity allows human-computer interaction
  possible.
• Advances in algorithms and availability of
  low-cost hardware will make interactive
  human-computer interactions possible in everyday
  life.

11
References

• 1 R. Bajcsy. Active perception. IEEE
  Proceedings, 76(8)996-1006, 1988.
• 2 A. Blake and M. Isard. 3D position, attitude
  and shape input using video tracking of hands and
  lips. In Proc. SIGGRAPH 94,pages 185192, 1994.
  In Computer Graphics, Annual Conference Series.
• 3 T. Darrell, P. Maes, B. Blumberg, and A.
  P.Pentland. Situated vision and behavior for
  interactive environments. Technical Report 261,
  M.I.T. Media Laboratory, Perceptual Computing
  Group, 20 Ames St., Cambridge, MA 02139, 1994.
• 4 I. Essa, editor. International Workshop on
  Automatic Face- and Gesture- Recognition.IEEE
  Computer Society, Killington, Vermont, 1997.
• 5 W. T. Freeman and M. Roth. Orientation
  histograms for hand gesture recognition. In M.
  Bichsel, editor, Intl. Workshop on automatic face
  and gesture-recognition, Zurich, Switzerland,
  1995. Dept. of Computer Science, University of
  Zurich, CH-8057.
• 6 W. T. Freeman and C. Weissman. Television
  control by hand gestures. In M. Bichsel, editor,
  Intl. Workshop on automatic face and gesture
  recognition, Zurich, Switzerland, 1995. Dept. of
  Computer Science, University of Zurich, CH-8057.

12

• 7 B. K. P. Horn. Robot vision. MIT Press,1986.
• 8 M. Krueger. Articial Reality. Addison-Wesley,
  1983.
• 9 K. Kyuma, E. Lange, J. Ohta, A. Hermanns,B.
  Banish, and M. Oita. Nature, 372(197),1994.
• 10 R. K. McConnell. Method of and apparatus for
  pattern recognition. U. S. Patent No.4,567,610,
  Jan. 1986.