Spectral and color prediction for arbitrary halftone patterns: a drop-by-drop, WYSIWYG, “ink on display” print preview
3 likes458 views
The document presents an innovative approach to improving print simulation and preview accuracy for digital print production through a drop-by-drop, 'ink on display' method. It addresses challenges in existing print modeling, particularly the need for accurate color formation predictions in inkjet technologies, and discusses a new framework that incorporates halftone data for better visual representation. Key findings indicate that this method enhances customer assurance and reduces wastage by ensuring previews closely mimic final printed outputs.
![Background: print spectral & color models
• Prediction of integrated halftone
color: Yule-Nielsen modified
Neugebauer
• Reflectance of entire halftone pattern
predicted from constituent ink dot
overlap colors:
• where R(λ) is reflectance of optically-
integrated halftone pattern
neighborhood, wj is relative area
coverage of the j-th Neugebauer
Primary – P, and n is Yule-Nielsen non-
linearity accounting for optical dot gain.
Side view
Colorants
Substrate
Appearance
W
C M Y
Neugebauer primaries
Materials
Example halftone
C M Y CMY
+ + =
NP areas
W=1/9
C=1/9
M=2/9
CM=2/9
CY=1/9
MY=1/9
CMY=1/9Colorant vector
[C,M,Y]=[5/9, 6/9,3/9]](https://image.slidesharecdn.com/cic23mimir20151013-151202143212-lva1-app6892/85/Spectral-and-color-prediction-for-arbitrary-halftone-patterns-a-drop-by-drop-WYSIWYG-ink-on-display-print-preview-6-320.jpg)
Spectral and color prediction for arbitrary halftone patterns: a drop-by-drop, WYSIWYG, “ink on display” print preview
- 1. Spectral and color prediction for arbitrary halftone patterns: a drop-by-drop, WYSIWYG, “ink on display” print preview Peter Morovič, Ján Morovič, Xavier Fariña, Pere Gasparin, Michel Encrenaz, Jordi Arnabat Presented at 23rd IS&T Color and Imaging Conference, 21st October 2015, Darmstadt, Germany
- 2. Overview • Motivation – why do we keep working on print modeling? • Background – existing approaches & limitations • An end-to-end print simulation approach • Test setup • Results & discussion • Conclusions
- 3. Motivation • Why? • Digital print production: many short-run jobs • Bottleneck: operator intervention • Web-to-print workflows increase cost of manual overrides + greater need for accurate print preview before placing order • To avoid re-printing • To provide customer assurance • Challenge: Make print preview close to final printed output as possible • If not, decisions made on basis of preview are unreliable & inaccuracies further aggravate getting to right print
- 4. Lacking components • First, accurate model of print color formation, applicable to technologies like inkjet • Ink-media interactions are complex, highly non-linear & not well represented by models developed for analog printing • Second, framework for applying scaling to print preview that reflects nature of print color formation, instead of assuming display color formation.
- 5. Background: print spectral & color models • Prediction of ink overprinting: Kubelka Munk (1931) • Per-ink K (absorption) and S (scattering), derived from opacity and reflectance: • where R∞ is reflectance of infinitely thick sample; prediction made at wavelength λ • Combining multiple ink layers is then additive in K and S: • where B is substrate, l is number of ink layers, ci concentration and Ki absorption coefficient of the i-th layer Wang & Wang, 2007
- 6. Background: print spectral & color models • Prediction of integrated halftone color: Yule-Nielsen modified Neugebauer • Reflectance of entire halftone pattern predicted from constituent ink dot overlap colors: • where R(λ) is reflectance of optically- integrated halftone pattern neighborhood, wj is relative area coverage of the j-th Neugebauer Primary – P, and n is Yule-Nielsen non- linearity accounting for optical dot gain. Side view Colorants Substrate Appearance W C M Y Neugebauer primaries Materials Example halftone C M Y CMY + + = NP areas W=1/9 C=1/9 M=2/9 CM=2/9 CY=1/9 MY=1/9 CMY=1/9Colorant vector [C,M,Y]=[5/9, 6/9,3/9]
- 7. Background: print spectral & color models • Model summary: • specific ways for predicting specific phenomena • scattering, absorption, optical dot gain and optical integration • specific assumptions about properties of constituent parts • e.g., homogeneity of ink layers, interfaces, independence of layer properties, infinite substrate thickness, etc. • knowledge of certain component properties is a prerequisite • Reflectance of infinitely thick ink layers, scattering of substrate, concentrations of inks, optical dot gain, Neugebauer Primary area coverages. • Practical challenges: • difficult to know properties • fundamental properties often estimated from measurements • Opportunity: think of models merely as computational mechanisms
- 8. Background: ICC-based soft-proofing • What: print color simulations (soft-proofs) obtained by building ICC profiles of display and printing system and color managing print colorimetry to display device color space • treats printing system like a black box • at best, result is color accurate at LUT nodes • in-between, assumptions of linearity are made • process is blind both to spatial features (e.g., grain) and to actual nature of transitions between profile LUT nodes • incl., mapping from device color space to printing system’s colorant space (i.e., color separation) and consequences of specific halftoning used (e.g., when dots start overlapping). • Limitations: • no sense of print’s spatial features, • makes transitions have a characteristically display-like look • mismatch in artifacts (missing some from the print; introducing others not in print)
- 9. Background: ICC-based soft-proofing Original Soft-proof Print photo False artifacts Grain Transitions
- 10. An end-to-end print simulation approach 3 key elements: • Predicting print appearance from halftone data instead of continuous-tone data input to a printing system after color management. • Extending printer modeling by a mechanism that enforces spectral correlation. • Printer model based scaling, instead of device color space linear interpolation.
- 11. Element 1: Print-ready halftone data • Pass print job through same • color management, • color separation, • linearization and • halftoning • as used for printing. • End result: halftone that could be directly printed, but that will be used for on-screen simulation instead.
- 12. Element 2: RONT model • Starting point: Kubelka Munk + Yule-Nielsen modified Neugebauer, with key parameters being Reflectance, Opacity and the optical dot-gain N exponent. (RON) • BUT: RON does not represent inkjet systems well • Approach: add computational mechanisms to model • All of KM+YNN performed wavelength-by- wavelength • BUT: strong correlation between adjacent spectral bands (e.g., Singh et al., 2003; Morovic et al., 2014)
- 13. Element 2: RONT model • Concept: adjust reflectance predictions in a given band based on predictions made for its neighbors. • takes advantage of data from spectral neighborhood • allows for enforcing correlation found in natural spectra that a per-band model may break • Mechanism: matrix transformation of the reflectances predicted using RON parameters only • To compute such matrix T, L2 norm between intermediate RON reflectance predictions (RI) and measured reflectances (RM) needs to be minimized: that not all wavelengths have an equal spr axis. A wavelength-by-wavelength view (F ences between individual correlations in mo
- 14. Element 2: RONT model • where R matrices have m rows (for m samples) and s columns (for s spectral bands) and where f() is function specifying t terms to use in error minimization, resulting in dimensions of f(RI) being m×t and those of T being t×s. • f() may add constant term, cross terms and power terms • in simple, hypothetical case of s=2, result of f() could be the following for a second order case: • Having computed T, its result is applied to predictions based on RON parameters:
- 15. Element 3: Print optical integration-based scaling • simulate optical integration that takes place when print is viewed at different distances • each display pixel represents optical integration of print halftone pixel neighborhood • display pixel color needs to be computed using full RONT model for neighborhood, instead of as mean of per-halftone-pixel predictions.
- 16. Test setup Label Z6200-4 PWT-4 L310-4 L310-6 Printer (HP) Desigjet Z6200 PageWide Technology Latex 310 Latex 310 Substrate HP Heavyweight Coated HP Heavyweight Coated Self-Adhesive Vinyl HP Photorealistic Paper Inks CMYK aqueous CMYK aqueous CMYK latex CMYKcm latex Max. drops per ink per pixel 2 2 3 2 Neugebauer Primaries 34=81 34=81 44=256 36=729 Training samples 756 620 3257 4808 Test samples 459 620 3257 4808
- 17. Results (∆E2000) System Model Median 95th %tile Max. Z6200-4 RON 8.5 23.2 35.4 RONT 1.5 4.8 8.2 PWT-4 RON 3.2 8.1 15.6 RONT 0.9 2.2 3.4 L310-4 RON 4.1 11.2 29.9 RONT 1.3 2.8 6.5 L310-6 RON 1.7 4.9 10.6 RONT 0.9 2.2 7.7
- 19. Conclusions • Successful print simulation can make the difference between • wasting resources and frustrating (or even losing) customers, • delivering salable products efficiently & with good customer experience. • A new solution was presented here that: • bases simulations on same data that drives printing systems – i.e., colorant channel halftone data • ensures that simulated halftone data is scaled for display in a way that respects its local properties, • Resulting simulations that are both more color-accurate and represent print-specific features and limitations faithfully. • Next steps: test new framework on broader variety of printer-substrate-ink configurations and extended as necessary.