2010 © HP

HANSA New Color Separation and Halftoning Paradigm

Ján Morovič, Peter Morovič & Jordi ArnabatHewlett–Packard CompanyBarcelona, Spain

2010 © HP

What makes printed colors?

2010 © HP

Anatomy of a color halftone print

2010 © HP

Anatomy of a color halftone print

2010 © HP

Anatomy of a color halftone print

2010 © HP

Print color formation

Side view

70% W13% C10% K 6% M 1% CM

Neug

ebauer p

rimaries

Rel

ativ

e ar

ea c

ove

rag

es

Subtractive

Additive

2010 © HP

Analog from digital

2010 © HP

From color to halftone pattern

CIE L*a*b*sRGB

SWOP CMYK...

color managementcolor appearance modeldevice characterization

color enhancementgamut mapping

color separationlinearizationcalibrationhalftoning

2010 © HP

Controlling print color – a first principles approach

2010 © HP

How do we get from color to halftone pattern?

20% W30% C20% M 0% Y20% CM 0% CY 0% MY 0% CMY

Source color Color management Printable color

Halftone pattern’s Neugebauer Primary

statisticsHalftone pattern

2010 © HP

How does this differ from traditional color control?

Traditional New

Color separation

‘How much of each ink should I use for each

color?’Output: ink amounts

How much area should I cover with each

Neugebauer Primary’Output: Neugebauer

primary area coverages

Halftoning

Decides where to place ink drops based on color separation constraints.Controls: spatial and

overprinting properties

Decides where to place ink drops based on

color separation constraints.

Controls: spatial properties only

Ink amounts v.

patterns1:1 1:many

Specifying Neugebauer Primary area coverages provides access to vastly greater space of printable patterns.

kn v. n(for system where up to k-1 ink drops per pixel can be specified for n inks)

2010 © HP

Neugebauer primary area coverages: nightmare or walk in the park?

• Specifying Neugebauer Primary area coverages (NPacs)

• Selecting point in kn dimensions versus n (e.g., 46=4096D NPac space for CMYKcm printer with max. 3 drops per pixel per ink versus 6D ink space)

• Efficiently and effectively traversing high dimensional space

• Accurately predicting NPac colorimetry

• Obtaining NPac statistics on paper

• Trivial if ink drops were tessellating, uniform, perfectly–square and not subject to optical dot–gain :)

• BUT: difficult to do accurately due to dot gain, colorant layer thickness variation, substrate surface properties, ink-substrate interaction, ink–ink interaction, drop shape, drop placement errors, mis-registration, ...

2010 © HP

What if we can’t account for / eliminate obstacles

Printable color

Col

or s

epar

atio

n

Printed pattern NPacs (matching color)

Digital pattern NPacs (resulting in printed

patterns matching color)

Dig

ital

Prin

ted

[W,C,M,CM]=[0,0.5,0.5,0]

[W,C,M,CM]=[0.05,0.45,0.35,0.15]

[W,C,M,CM]=[0.5,0,0,0.5]

[W,C,M,CM]=[0.35,0.15,0.1,0.40]

digital NPac vectors > printable NPac vectors >> ink vectors

2010 © HP

From theory to practice

2010 © HP

A minimal Halftone Area Neugebauer Separation setup (CMYK, 1bpp)

Print & measure Neugebauer primary

(NP) CIE XYZs

Compute convex hull & tetrahedralize

hull NPs

Find printable color’s enclosing

tetrahedron

Printable color

20% W30% C20% M 0% Y20% CM 0% CY 0% MY 0% CMY 0% K 0% KC 0% KM 0% KY 0% KCM 0% KCY 0% KMY 0% KCMY

Barycentric coordinates are vertex NP areas

Select one NP per pixel & diffuse NPac-NP error

2010 © HP

Does it work?

2010 © HP

Test setup: ‘Can we find NPacs that use less ink?’

• Printer: HP Designjet L65500

• Inks: CMYKcm latex

• Substrate: Avery Self-Adhesive Vinyl

• Color samples: 748 Lab-uniform ISO coated v. 2 samples

• Color workflows compared:

• Ink space separation, GCR optimized for low grain, ink space halftoning (current default)

• Ink space separation, maximum GCR optimized for low grain, ink space halftoning (current optimal)

• NPac space separation (optimized for minimum ink use) and halftoning (HANS)

2010 © HP

Results –!ink use

2010 © HP

Results – image quality

Current(optimal ICC) HANS

2010 © HP

What next?

2010 © HP

Challenges and benefits

• Challenges:

• printer model accuracy (the more accurate the better the optimization)

• computational efficiency (weeks of computation per substrate)

• optimization (efficient models of print attributes, efficient traversal of NPac space)

• Benefits:

• greater & direct optimization (more from the same printer-ink-substrate)

• explicit trade–off among print attributes (grain v. ink use v. color constancy)

• inkset agnosticism (same process for CMY 1bpp and CMmYKkNnRGB 2bpp)

2010 © HP

Acknowledgements

– Dudi Bakalash– Lahav Langboim

– Shay Maoz– Amir Sheinman

– Igor Yakubov

– Gary Dispoto– I-Jong Lin

– John Recker– Ingeborg Tastl

– Bob Ulichney

– Michel Encrenaz

– Eduard Garcia– Joan Manel Garcia

– Oriol Gasch– Rafael Gimenez– Rafael Goma

– Andrés Gonzalez– Jacint Hument

– Johan Lammens– Alan Lobban

– Scott Norum

– Aleix Oriol– Ramon Pastor

– Yvan Richard– Aurora Rubio– Albert Serra

– Jep Tarradas– Joan Uroz

– Jordi Vilar

2010 © HP

Thank you!