The Spectral Printer: From Technical Challenge To Business Case 
1 like721 views
This document discusses the technical and business challenges of spectral printing. Spectral printing aims to match the spectral reflectance of original content under any light source, rather than just matching colorimetry under a single light source. The document outlines the key technical challenges in achieving spectral printing, including developing suitable colorants, gamut mapping algorithms, and halftoning techniques. It also discusses whether customers would perceive the improved spectral matching. Finally, it considers the potential business case by estimating the size of the addressable printing market and whether spectral printing capabilities could enable premium pricing or new applications.
The Spectral Printer: From Technical Challenge To Business Case 
- 1. THE SPECTRAL PRINTER: FROM TECHNICAL CHALLENGE TO BUSINESS CASE Dr. Ján Morovič Senior Color Scientist / Master Technologist, Hewlett–Packard Company Director of CIE Division 8: Image Technology 8th ROND Conference, 14th March 2013, Örnsköldsvik, Sweden http://fotonerdz.deviantart.com/art/Matrix-spectrum-278731679 © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 2. Can we make it? Can we make money from it? WHICH COMES FIRST? © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 3. OUTLINE • What is spectral printing? (Again!) • Can we make it? • Technology: inks, substrates, gamut, separation, workflow, content • Customer benefit: “Will I see it?” • Can we make money from it? • Market size: how many billion pages/$? • Added value: premium pricing / new applications? • Conclusions © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 4. WHAT IS SPECTRAL PRINTING? Match original under any light source. Light source 1 Match for all observers (human, animal). → Reproduction changes with illumination like the O P original does and for all viewers. ‘I can see what the Mona Lisa would look like Light source 2 Light source in my living room’ Light source O P BUT: “Match” is aim, not necessarily property! © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 5. SO, WHAT IS COLOR PRINTING? Designed to match color under one light source (D50) and Light source 1 for human observers with certain cone sensitivities. O P BUT: • ∆Es even under chosen light source • Uncontrolled/unknown ∆Es under other light sources Light source 2 Light source Light source Assumption: for a spectral print, color differences under variety of light sources are lower than for colorimetric (metameric) print. O ? P P © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 6. ANY OTHER ‘SPECTRAL’? • Focus here on spectral printing (i.e., aiming at a match of spectral reflectance between original and reproduction). • Spectral data also has benefits for metameric reproduction • Metameric match under non-standard / multiple light sources • Modeling of colorant interactions (e.g. spot color overprinting) • Modeling of printer behavior © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 7. HOW DO I MAKE A SPECTRAL PRINT? Spectral Gamut Matching content mapping Halftoning Colorants Substrate © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 8. WHERE CAN I GET SPECTRAL CONTENT? Spectral content Gamut mapping Matching Halftoning Colorants Substrate • Spectral image capture → broadly applicable, but still scarce • Spectral measurement → applicable to color palettes (e.g., Pantone) • Predicted from original device’s model → when original is a print, spectra can be predicted using model of it’s colorants and halftoning • Predicted from colorimetry / trichromatic capture → colorimetric capture + measurements to characterize spectral basis (e.g., of artwork) • Without spectral content, spectral printing is inapplicable → not going to replace metameric color reproduction wholesale © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 9. HOW DO I ADAPT IT TO MY MATERIALS? Spectral content Gamut mapping Matching Halftoning Colorants Substrate • Single-shot mapping • Sequential color mapping: match under illuminant U, then to V in remaining • Direct spectrum → colorant amounts metamer set, then to W, etc. (Urban et al., mapping (optimization/regression, 2008) computationally expensive, disregards visual system entirely) (e.g., Bastani et al., • Workflow considerations 2006) • LabPQR (Derhak and Rosen, 2004) → extension • Gamut mapping + Matching of ICC framework, easy WF integration • Color gamut mapping to gamut of 3 fundamental bases + spectral mapping in • Result: spectrum to colorant amounts metameric blacks (Chau and Cowan, 1996) © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 10. TWO-STAGE MAPPING & LABPQR © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 11. HOW DO I BUILD COLORANT PATTERNS? Spectral content Gamut mapping Matching Halftoning Colorants Substrate • Challenges: • Error diffusion (Norberg, 2012) • Multiple colorants (CMYK+) • BUT: “arbitrary” multi-colorant patterns challenging • Difficulty of applying threshold matrix halftoning (AM/FM) → moire • Potentially abrupt changes in colorant space → artifacts • Solutions: © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 12. AN ALTERNATIVE: HANS Spectral content Gamut mapping Matching Halftoning Colorants Substrate printer (inks, media, WS, FW, pipeline)! artist! Printing as mosaic assembly Separation: colorant amount selection → Neugebauer Primary mosaic! (NP) statistics print! Halftoning: per colorant continuous levels to discrete drops → NP per pixel from local statistics tesserae/ Neugebauer Primaries (NPs) / tiles! © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice. at-pixel ink drop states!
- 13. AN ALTERNATIVE: HANS Spectral content workflow Gamut mapping Matching Halftoning Colorants Substrate colorant colorant colorant colorant space matching vector halftoning drops / gamut reflectance reflectance pixel / mapping colorant workflow NP area NP NP space HANS coverage matching halftoning vector © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 14. AN ALTERNATIVE: HANS Spectral content Gamut mapping Matching Halftoning Colorants Substrate Alternative basic Alternative basic • The reflectance of a pattern is the convex, area-coverage- ink combinations:! Print color! NPacs:! 2! 6086! weighted combination of it’s NPs’ reflectances (corrected for (in 4D)! (in 34=81D)! dot gain) → YNSN 3000x possible patterns (same • A reflectance can be matched if there exists a simplex in color)! W=70%! C1=2%! reflectance space, that has the reflectances of a printing M1=5%! Y1=5%! system’s NPs as its vertices W=79%! Y1=2%! K1=6%! C1M1=1%! W=77%! K1=14%! C1K1=2%! • The area coverages that need to be assigned to each NP C=34%! M=27%! C=7%! Y=1%! included! C1=1%! Y1=2%! Y1K1=2%! C2=1%! M1K1=3%! Y1K1=2%! match the convex weights with which it needs to be Y=28%! K=27%! K1=20%! K2=2%! C2=4%! combined to match the target spectrum • Operating in NP space offers vastly more choice: n versus kn dimensions (for n colorants and k levels per colorant per pixel) • E.g., for 6-ink printer with up to 3 drops per ink per pixel, HANS has 46=4096D versus the 6D of a colorant workflow • The rest is brute (or not so brute) force! © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 15. AN ALTERNATIVE: HANS Spectral content Gamut mapping Matching Halftoning Colorants Substrate • HANS error diffusion: • Error is computed in area coverage terms • Area coverages account for optically- additive aspect of print color formation • Arbitrary NP combinations can be formed and transitioned between • Non-linearity of colorant interactions is handled prior to halftoning, which can then operate in an additive space © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 16. COLORANTS AND SUBSTRATE Spectral content Gamut mapping Matching Halftoning Colorants Substrate • What colorants are best for spectral • what is physically realizable reproduction? • what can be manufactured in bulk • Several studies (e.g., Tzeng, 1999) • what can be successfully jetted or • Narrow-band, inverted Gaussians deposited using other marking (from first principles; Chen, 2004) engines • Substrate’s colorant limit (rules out Gaussians) • BUT, constraints from: © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 17. COLORANTS AND SUBSTRATE Spectral content Gamut mapping Matching Halftoning Colorants Substrate © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 18. WILL MY CUSTOMERS SEE THE DIFFERENCE? © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 19. WHAT MAKES A GOOD SPECTRAL PRINT? • Many metrics defined in the literature • Purely spectral, taking into account CMFs, focusing on metamerism between specific light sources, looking at sets of color difference statistics per light source + combinations of these • BUT: how does an observer experience the goodness of a spectral match? • Original – print ← observer • Observer views original-print-pair under many light sources (not arbitrary → database of measured lights) • Under each light source they see certain levels of color difference (→ most accurately predicted by ∆E2000, with JND units) • Surveying their experiences from under multiple light sources gives rise to a variety of color difference magnitudes (→ choice of relevant statistics) © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 20. MIPE: AN EXPERIENCE–BASED METRIC Color difference with visually meaningful units (~JND) Paramers (>0 ∆E even under ‘reference’ illuminant) 20 Non-parametric descriptive statistics Median: how close a match can be expected for an arbitrary, but realistic, light source 95th percentile – median: how much this Light sources / illuminants match varies CIE standard and recommended illuminants Maximum: how far apart the two can get + measured light sources at worst © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 21. SPECTRAL ‘GOODNESS’ OF METAMERIC PRINT Colorimetric ‘goodness’ Spectral ‘goodness’ • Similar performance under D50 (for Spot color: PANTONE uncoated, matte and coated, 1224 patches per substrate (3672 total samples; mixtures of which colorimetry was matched) and Pantone system’s 15 base inks) Fine art: 1168 measurements taken from paintings all other 172 illuminants (color gamut differences a big contributor already) matte glossy • MIPE errors higher than D50 ∆Es HP Designjet Z3100 (sanity check for maximum) 10 inks: cMmYnNKRGB • → Gap between MIPE and D50 ∆Es + glossy/matte black + gloss enhancer indicates room for improvement from spectral matching © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 22. SPECTRAL ‘GOODNESS’ OF SPECTRAL PRINT © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 23. TECHNICAL CHALLENGES • Degree of “spectral goodness” benefit versus metameric reproduction • spectral workflow + inherent limits of materials used • more colorants ≠ better spectral performance (necessary but not sufficient) • Way forward: • Dedicated spectral reproduction colorants • Efficient spectral workflow • Experience-focused optimization • Availability of content © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 24. STANDARDS? (A.K.A. THE CIE SALES PITCH) • CIE Technical Committee 8-07 Multispectral Imaging (chair: Masahiro Yamaguchi) • Terms of reference: To study, develop and recommend encoding techniques and data formats for the exchange of multispectral images, and to provide test procedures for the evaluation of multispectral imaging systems. • Technical report due end ’13: will provide basics of multispectral format requirements & compare four alternative formats © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 25. CAN I MAKE MONEY FROM IT? © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 26. ADDRESSABLE MARKET ($846B IN 2016 WW) $118B 3% 1% 2% 14% 4% 4% Potentially 5% spectral 8% 43% $364B 13% Non-spectral 18% 86% Packaging Commercial Advertising Magazines Newspapers Books Catalogues Office Directories Security Source: PIRA © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 27. SPECTRAL APPLICATIONS Application Feasibility Salability higher (similarity to production print is higher (content available, matching has Proofing essence of proof → spectral is clearly good chances ← similar materials) better than metameric proof) medium (spectral match to original may not medium (content availability limited, material Fine-art reproduction always be desired → Mona Lisa in Louvre v. my differences greater) kitchen) higher (color differences are source of many lower (content scarce, materials potentially Catalogues returned good, better reproduction → more very different) profitable operation) higher (persistent need for ever broadening higher (spectral differences are created Security variety of features; closer ties to original among available metamer sets, not v. original) materials if spectral properties exploited) © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 28. WHERE DOES THE MONEY COME FROM? • “Better” reproduction → proof, fine art print • New features → security • Secondary benefits → reduction lost profit from returned goods (catalogues) • Premium pricing v. differentiation from competitors (print as commodity) © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 29. A TALE OF TWO PRINTERS • Printer A: the hunting knife • Dedicated spectral printing colorants & printing technology + spectral workflow • + Excels at spectral printing; limited addressable market • - Not suitable for ‘color’ printing (ink use, grain, cost) • Printer B: the Swiss Army knife • General purpose materials & technology (+ spectral workflow) • + Suitable for spectral or ‘color’ printing; large addressable market • - Less spectrally accurate than Printer A © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 30. CONCLUSIONS • There are multiple applications that can benefit from spectral printing • Colorants developed specifically for spectral printing would boost its potential • Solutions need to focus on customer/viewer experience © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 31. ACKNOWLEDGEMENTS • Peter Morovič • Ole Norberg • Juan Manuel García-Reyero • Martí Rius • Jordi Arnabat • Johan Lammens • Carlos Amselem © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.
- 32. THANK YOU! © Copyright 2013 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice.