Color Seamlessness in Multi-Projector Displays Using Constrained Gamut Morphing IEEE Visualization, 2009 Behzad Sajadi Maxim Lazarov Aditi Majumder M. Gopi

2 Registration Problem

3 Color: Brightness & Chrominance  Brightness: 1D  Chrominance (x, y): 2D  3D color gamut

4 Color Variation Visualization

5 Overview  Prior Art  Motivation  Algorithm  Results

6 6 Prior Art: Overlap Blending Proj1 Proj2 Proj1 Proj2 Overlap Region Raskar et al SIGGRAPH 1998; Li et al IEEE Computer Graphics and Applications, 2000; Chen et al SPIE Projection Displays, 2001

7 7 Prior Art: Overlap blending  Assumes Uniform brightness in each projector All projectors have similar brightness Projectors are linear devices  Addresses the overlaps only  No measurement or correction of intra or inter projector brightness variation

8 8 Before Correction Overlap Blending Prior Art: Overlap Blending

9 9 Prior Art: Measurement with High Resolution Camera Single Projector Brightness Profile Multi Projector Brightness Profile

10 Prior Art: Strict Brightness Uniformity u L IEEE TVCG 2003, PROCAMS 2003 Majumder and Stevens

11 Prior Art: Strict Brightness Uniformity L u L Significant Contrast/ Dynamic Range Compression

12 Before After Strict Brightness Uniformity Prior Art: Strict Brightness Uniformity

13 Prior Art: Constrained Brightness Smoothing  Smoothing is sufficient for perceptual seamlessness Non-linear filtering Maximize dynamic range Solved using dynamic programming ACM Transactions on Graphics 2005 Majumder and Stevens

14 Before After Strict Brightness Uniformity Prior Art: Constrained Brightness Smoothing IEEE TVCG 2003, PROCAMS 2003 Majumder and Stevens

15 Before After Constrained Brightness Smoothing ACM Transactions on Graphics 2005 Majumder and Stevens Prior Art: Constrained Brightness Smoothing

16 Overview  Prior Art  Motivation  Algorithm  Results

17 Motivation: Does it solve the problem? 17

18 Motivation: Our Contribution  Chrominance is constant within projector  Chrominance varies across projectors and in overlaps  Brightness smoothing does not guarantee chrominance smoothing  Constrained chrominance smoothing in addition to brightness smoothing 18 3D Gamut Morphing

19 Motivation: Key Insight  Smooth transition of chrominance across overlap region Blending of the chromaticity coordinates Only need to manipulate the brightness proportions of overlapping projectors  Manipulate brightness to address chrominance variation

20 Overview  Prior Art  Motivation  Algorithm  Results

21 Algorithm: Chrominance profile before registration

22 Algorithm: Chrominance Gamut Morphing Chrominance Gamut Morphing Horizontal Blending Vertical Blending New Brightness Profiles Horizontal Blending Attenuation Map Vertical Blending Attenuation Map

23 Algorithm: Horizontal Blending Before Chrominance Blending After Horizontal Blending

24 Algorithm: Vertical Blending After Horizontal Blending After Vertical Blending

25 Algorithm: Pipeline Projector Brightness Profiles Projector Chrominance Gamut Chrominance Morphing Attenuation Map Chrominance Gamut Morphing

26 Algorithm: Chrominance gamut morphing  Modifies brightness profiles  Removes C 0 brightness discontinuity Before Correction After Horizontal Blending After Vertical Blending

27 Algorithm: Perceptual Brightness Constraining  Does not guarantee imperceptible brightness changes  Apply Majumder et. al to constrain the brightness variations  Retains chrominance gamut morphing After Vertical BlendingAfter Brightness Constraining

28 Algorithm: Bezier-based Brightness Smoothing  Brightness profile is not derivative continuous after Majumder et. Al  Assures C n intensity continuity  Retains chrominance gamut morphing After Vertical BlendingAfter Brightness ConstrainingAfter Bezier-based Smoothing

29 Algorithm: Brightness Smoothing Brightness Smoothing Perceptual Brightness Constraining Vertical Bezier-based Brightness Smoothing New Brightness Profiles Brightness Smoothing Attenuation Map

30 Algorithm: Offline Correction Pipeline Chrominance Gamut Morphing Projector Brightness Profiles Reconstruct Chrominance Gamut New Brightness Profiles Brightness Smoothing Final Alpha Mask (A) White Balancing Prior to Correction

31 Color Variation Visualization

32 Algorithm: Online Image Correction Apply Alpha Mask Apply Projector Transfer Function Input Image Custom Linearization Function Linearize Input Image Final Alpha Mask (A)

33 Overview  Prior Art  Motivation  Algorithm  Results

34 Results: More General Pictures

35 Results: Extends to any Geometry

36 Before Correction Overlap Blending Final Result Majumder and Stevens 2005 Results: Comparison

37 Conclusion  First method that Complete 3D color registration  Addresses spatial variations in both chrominance and brightness High quality display with commodity projectors

38 Future Work  Extend to non-developable surfaces  Address intra-projector color variations

39 Questions?

40 Prior Art: Color Seamlessness  Brightness  Chrominance 40