Jzazbz Current Proposal

June 27, 2017

ICC-CMRF (2016-2017)

An Efficient Uniform Color Space for High Dynamic Range and Wide Gamut Imagery

Presenter: Muhammad Safdar (PhD) Supervisor: Prof. M. Ronnier Luo

Zhejiang University, China

1 Uniform Color Space  Predicts perceptual color difference  No inter-dependence b/w lightness, chroma, and hue

ELCH ()()() 2   2   2

Lightness Hue

• K. A. G. Smet et al., LEUKOS, 2016 • I. Lissner et al., IEEE Trans. Image Process., 2012 2 Criteria for a New UCS

1 Uniformity (Local & Global) 2 Hue Linearity 3 Grey-scale Convergence

Lightness

Chroma

4 To uniformly encode high dynamic range signals e.g., 0-10,000 cd/m2

5 To uniformly encode wide gamut image signals e.g., Rec.2020

3 Test Spaces

1) CIELAB CIE/ISO standard

2) CAM16-UCS Accurately predicts small color difference data

3) ICTCP Dolby’s proposal for HDR and WCG

4) Jzazbz Current Proposal

• CIE, Colorimetry, 2004 • C. Li et al., Color Imaging Conf., 2016 • Dolby, 2016 4 Criteria and Visual Data

No. Criteria Data sets Purpose 1 Perceptual Color i) COMBVD i) Training Difference ii) OSA ii) Testing 2 Perceptual Uniformity i) COMBVD Ellipses i) Reference ii) MacAdam Ellipses ii) Testing iii) Munsell Data iii) Testing 3 Hue Linearity i) Hung & Berns i) Reference ii) Ebner & Fairchild ii) Testing iii) Xiao et al. iii) Testing

4 Wide-range Lightness i) RIT SL1 i) Testing Prediction ii) RIT SL2 ii) Training

5 Grey-scale Chroma-ratio metric (%) 3C Chroma-Ratio 100 w convergence CCCr g b

• D. L. A. MacAdam, JOSA, 1942 • D. L. A. MacAdam, JOSA, 1974 • M. D. Fairchild, Electronic Imaging, 2011 5 Proposed Jzazbz space

X ' bX  ( b 1) Z  D65 D65 D65 '     (1) YD65 gYD65  ( g 1) X D65 

LX  0.41478972 0.579999 0.0146480  D65  MY  -0.2015100 1.120649 0.0531008        D65  (2) SZ  -0.0166008 0.264800 0.6684799  D65 

n p LMS,,  14 cc12  wheren  2610 / 2 10,000 '''  5 (3) LMS,,,  n p 1.7 2523 / 2 LMS,, 1 c 3  10,000

IL  0.5 0.5 0  '  aM  3.524000 -4.066708 0.542708  ' (4) z         ' bSz  0.199076 1.096799 -1.295875  

1 dI   z (5) J z   whered  0.56 1 dI z

6 Results (Uniformity) CIELAB CAM16-UCS

100 40

50 20

M b* b 0 0 -10

-20

-50 -30

-40 -50 0 50 100 -40 -20 0 20 40 a a* M

IC C J a b T P z z z 0.15

0.2 0.1

0.1 0.05

b -C 0 0

-0.1 -0.05

-0.2 -0.1 -0.2 -0.1 0 0.1 0.2 -0.1 -0.05 0 0.05 0.1 0.15 C a P z 7 Results (Uniformity in WCG)

ICTCP Jzazbz

-CT bz

az CP (a) (b)

8 Results (Quantitative)

50 CIELABCIELAB 40 CAM16-UCSCAM16-UCS

30 ICTCPICTCP

JzazbzJzazbz 20

0 COMBVD OSA COMBVD COMBVD MacAdam MacAdam Local Global Local Global

9 Results (Hue Linearity)

10 Results (Quantitative)

8 CIELABCIELAB CAM16-UCSCAM16-UCS 6 ICTCPICTCP

4 JzazbzJzazbz

0 Hung & Berns Hung & Berns Ebner & Fairchild Xiao et al. (blue)

11 Results (Lightness Prediction)

12 Results (Lightness)

20 CIELABCIELAB CAM16-UCSCAM16-UCS 15 ICTCPICTCP 10 JzazbzJzazbz 5

0 RIT SL1 RIT SL2 Munsell Value Chroma Ratio (%)

13 Conclusions

Jzazbz  Second best for small color difference  Best for large color difference  Best for wide gamut uniformity  Best for hue linearity  Best for wide-range lightness prediction  Plausible grey-scale convergence

Jzazbz gave overall best performance and should be confidently used

14 Demonstration (Image Segmentation) Segmentation (7 regions) using K-means clustering algorithm Original CIELAB CAM16-UCS

YCbCr ICTCP Jzazbz Deliverables 1. Journal Paper: Optics Express (IF=3.3) https://www.osapublishing.org/oe/abstract.cfm?uri=oe-25-13-15131

2. MATLAB Code for Jzazbz

15 Any Questions? Model Development

 Initially, we start using a structure similar to ICTCP

LX 1,1  1,21  1,1  1,2  D65   MY  1       (1)  2,1 2,2 2,1 2,2  D65      SZ 3,1  3,21  3,1  3,2  D65 

n p LMS,, cc12  ''' 10,000 LMS,,,  n (2) LMS,, 1 c 3  10,000

' IL1,1  1,21  1,1  1,2   aM 1     ' z 2,1 2,2 2,1 2,2  (3) ' bSz 3,1  3,21  3,1  3,2 

• Dolby, 20016 • M. Safdar et al., Color Imaging Conf., 20016 Model Development  Then we introduced a linear given below to correct hue linearity

' X D65 bXD65  ( b 1) Z D65  '     (1) YD65 gYD65  ( g 1) X D65 

' LX1,1  1,21  1,1  1,2 D65  MY 1     ' (2) 2,1 2,2 2,1 2,2 D65  SZ3,1  3,21  3,1  3,2 D65

n p LMS,, cc12  ''' 10,000 (3) LMS,,,  n LMS,, 1 c 3  10,000

' ILz 1,1  1,21  1,1  1,2   aM 1     ' z 2,1 2,2 2,1 2,2  (4) ' bSz 3,1  3,21  3,1  3,2 

• G. Kuehni., Col. Res. Appl., 1999 • G. Cui et al., Col. Res. Appl., 2002 Model Development  Another simple equation to tune perceptual lightness

1 dI z J z  (5) 1 dI z

after Eq.5

• M. R. Luo et al., Col. Res. Appl., 2006