Shape and Luminance Cues for the Visual Perception of Glow

1. Shape and Luminance Cues for the Visual Perception of Glow Minjung Kim Master of Arts Candidate York University, Toronto August 12, 2011 Committee Chair: Dr. Laurie Wilcox Thesis Advisor: Dr. Richard Murray External Examiner: Dr. Keith Schneider Dean’s Representative: Dr. Laurence Harris

2. Source: NASA Astronomy Picture of the Day, December 19, 2009

3. + shape luminance =

4. + shape luminance = • Directly lit from above • Lambertian

5. + shape luminance =

6. + shape luminance =

7. Previous Work

8. Previous Work • Lightness-based: more luminous than white • e.g., Bonato & Gilchrist (1999); Li & Gilchrist (1999)

9. Previous Work • Lightness-based: more luminous than white • e.g., Bonato & Gilchrist (1999); Li & Gilchrist (1999) • Blur-based: smooth luminance gradients • May be less luminous than white • e.g., Zavagno & Caputo (2001, 2005) white grey

10. Glow in Complex Objects • Previous studies examined simple shapes • Some forms of glow unique to complex objects • Convexities, concavities • e.g., Langer (1999), translucent materials

11. Langer (1999) • Shadows, glow related via interreflections? • Interreflections: lighting unique to concavities • Entirely theoretical; no psychophysics normal contrast-reversed

12. Translucent Materials Concavity Bright Nearer to glow Convexity Dark Farther from glow http://www.mpi-inf.mpg.de/~lensch/ proj/3DScanTranslucent/

13. Bright-Means-Deep Hypothesis • Complex glowing objects are… • … dark in convexities • … bright in concavities • -> “bright-means-deep”

14. Bright-Means-Deep Hypothesis • Complex glowing objects are… • … dark in convexities • … bright in concavities • -> “bright-means-deep” • Bright-means-deep… • … in real-life glow? • … for perceptual cue to glow?

15. Outline

16. Outline Experiment 1: induce glow by mimicking bright-means-deep relationship (“diffuse countershading”)

17. Outline Experiment 1: induce glow by mimicking bright-means-deep relationship (“diffuse countershading”) Experiment 2: reduce glow by interfering with bright-means-deep relationship

18. Outline Experiment 1: induce glow by mimicking bright-means-deep relationship (“diffuse countershading”) Experiment 2: reduce glow by interfering with bright-means-deep relationship Experiment 3: perceive shape using bright-means-deep relationship

19. EXPERIMENT 1Can we induce glow by mimicking the bright-means-deep pattern?

23. Diffuse Countershading

24. Diffuse Countershading direct diffuse

25. Diffuse Countershading direct diffuse contrast-reversed

26. contrast-reversed direct diffuse (weight) Diffuse Countershading + + =

27. “countershaded” direct diffuse (weight) Diffuse Countershading + + =

34. Analysis Obtained probability of choosing higher-weight stimulus Ordering ignored when aggregating trials e.g., trial with w=2, w=8 equivalent to w=8, w=2

35. Results

38. 55 df

42. Results

43. Results

44. Control Experiments Control for influences of low-level features Mean luminance Contrast 2AFC with randomized low-level features

45. Mean Luminance Control

46. Contrast Control

47. Conclusion Diffuse countershading effective for glow Response probabilities Thurstone scaling analysis Participant response not driven by low-level properties Mean luminance Contrast Consistent with bright-means-deep hypothesis

48. EXPERIMENT 2Can we reduce glow by destroying the bright-means-deep pattern?

49. Depth Mismatching Some glowing objects already follow bright-means-deep pattern Shape modification while holding luminance constant removes this pattern Depth inversion Depth randomization Perceived glow should be reduced

50. Stimulus Generation Needed shape to appear equally good when depth-normal and depth-modified Needed to avoid using pre-rendered left- and right-eye images Monocular occlusion issues for depth-inversion experiment 1 experiment 2