Human Visual lperception relevant tto

Transcription

1 Human Visual lperception relevant tto 3D-TV Wa James Tam Communications Research Centre Canada

2 An understanding of Human Visual Perception is important for the development of 3D-TV

3 Ottawa Communications Research Centre Canada (CRC) is the primary federal laboratory for R&D in advanced telecommunications in Canada.

4 Stereovision 3D Research Stereo Inter-ocular Deficiencies Object-motion Surrogate & Averaging g Visual Comfort Depth Maps

5 What underlies 3D-TV? Depth from Disparity it Processing by the Human Visual System

6 Horizontal Disparity

7 Changing Horizontal Disparities

8 Two Types of Disparities Uncrossed ddisparityit Crossed Disparity

9 Differences/Implications Stereo deficiency Performance Visual comfort Visual display

10 Stereo fi i i Who can Deficiencies benefit from 3D TV?

11 6% vs. 30%

12 Display Duration and Depth Discrimination

13 Apparatus

14 Front or Back?

15 Left or Right?

16 Depth Discrimination Front/back? Left/Right closer? 0.25 deg & 0.125/0.375 deg 4.7 deg 10.6 deg Dynamic tracking (PEST) Duration varied msec 100 viewers 8 Estimates each

17 Results

18 Results breakdown

19 Age

20 Hours watching TV

21 Interpupillary p separation

22 Conclusion Individuals vary along a continuum in their ability to process stereoscopic depth information

23 Inter-ocular Averaging How can bandwidth be reduced? d?

24 3D-TV

25 Interocular masking of blur in one eye Julesz 71

26 3D-TV

27 Asymmetrical Quality

28 Subjective evaluation of Image Quality For different extents of asymmetry in image quality

29 Continuous Quality Scale

30 Subjective Assessment Original Quality age Q Ima Severe

31 Subjective Assessment Original Coarse

32 3D-TV Viable

33 Left eye Right eye Scene Cut Cross-switch

34 Conclusion Asymmetrical coding with cross-switching at scene cuts is a viable method for bandwidth savings

35 Motion & Visual Comfort How might visual comfort be affected by stereoscopic objects in motion?

36 Motion in Depth

37 Display

38 100 Visual Comfort 21 Viewers Very Comfortable 80 Comfo ort 60 Comfortable Mildly Uncomfortable Visual 40 Uncomfortable 20 Extemely Uncomfortable Disparity Magnitude (minutes of arc)

39 100 Visual Comfort 21 Viewers Very Comfortable 80 Comfor rt 60 Comfortable Mildly Uncomfortable Visual 40 Uncomfortable 20 Extemely Uncomfortable 0 Slow Medium Fast Velocity (cm/sec)

40 Horizontal Motion Crossed Disparity

41 Display Uncrossed Disparity

42 Visual Comfort 100 Very Comfortable 80 Comfor rt 60 Comfortable Mildly Uncomfortable Visual 40 Uncomfortable 20 Extremely Uncomfortable 0 Crossed Cross&Uncross Uncrossed Sequence Type 21 Viewers

43 Visual Comfort 100 Very Comfortable 80 Comfor rt 60 Comfortable Mildly Uncomfortable Visual 40 Uncomfortable 20 Extremely Uncomfortable Disparity Magnitude (minutes of arc) 21 Viewers

44 Visual Comfort 100 Very Comfortable 80 Visua al Comfo ort Comfortable Mildly Uncomfortable Uncomfortable 20 Extremely Uncomfortable 0 Slow Medium Fast Velocity (degrees per sec) 21 Viewers

45 Conclusion Aside from disparity, object motion can significantly ifi influence visual ldiscomfort

46 Surrogate Depth Maps How to create 3D contents from 2D?

47 2D-to-3D Conversion Technique Object segmentation and horizontal shifting

48 2D-to-3D Another Approach Depth Maps

49 2D + Depth Multiple Two views for autostereoscopic p displays displays

50 Advantages bandwidth-efficient, multiple l images & viewpoints

51 Create Depth Maps using information from 2D source images

52 Make use of human visual perception for 2D-to-3D conversion

53 Depth Filling/Interpolation Communications Research Centre Canada

54

55

56 Depth Surface Perception The region surrounded by a disparate boundary tends to take up the same depth as its boundary

57 First Study Sparse Depth Maps Communications Research Centre Canada

58 Cones

59 Meal

60 Cones Full Edge Boundary

61 Meal Full Edge Boundary

62 Double-stimulus Continuous Quality Scale Communications Research Centre Canada

63 Image Quality Viewers 20 Differen nce Ra ating (T Test - Refere ence) Real-Camera Full-Depth Edge-Depth Flat-Depth Boundary-Depth Experimental Condition Communications Research Centre Canada

64 55 Depth Quality Viewers 50 ting nce) nce Rat Referen ifferen Test - R D (T Bette er 0 Real-Camera Edge-Depth Constant-Depth Full-Depth Boundary-Depth Experimental Condition

65 Surrogate Depth Maps Clearly different but functionally equivalent in a critical way Communications Research Centre Canada

66 New Method based on the gray level intensities of a colour component Communications Research Centre Canada

67 Components of Colour Image

68 Components of Colour Image

69 Components of Colour Image

70 Experimental Study Surrogate Depth Maps based on colour component Communications Research Centre Canada

71 Test sequences Ten eight-second video clips ITU-601 format (720x480, 30fps)

72 Aquarium

73 Camera Test

74 Confetti

75 Crowd

76 Flowerpot

77 Red Leaves

78 Street Organ

79 Trapeze

80 Tulips

81 Waterslide

82 Single-stimulus Continuous Quality Scale Communications Research Centre Canada

83 Single-stimulus Continuous Quality Scale Communications Research Centre Canada

84 Depth Results Image Depth Quallity 2D 3D Converted-3D Depth Quality Rating Video Sequence

85 Visual Comfort Results Image Visual Comfort 2D 3D CONVERTED-3D Visual Comfor rt Rating Video Sequence

86 How can it possibly work? - Natural Images - Foreground-background separation - Shading within objects - Visual experience Communications Research Centre Canada

87 Depth reversal Familiarity it information

88 Surrogate Depth Maps can be useful for 2D-to-3D conversion

89 Summary 1. Stereopsis (& disparity type) 2. Stereoscopic deficiencies 3. Inter-ocular averaging 4. Object motion (& visual comfort) (& display duration) (& bandwidth reduction) 5. Surrogate depth maps g p p (& 2D-to-3D conversion)

90 Take Home Message

91 Human Visual Perception can be exploited for 3D-TV

92 Thanks!

93 Half Colour

94 Half Colour

95 Sony GDM-F CRT monitors 4H viewing i distance Field-sequential Stereo

96 Subjective Assessment Depth Quality Visual Comfort 21 Viewers