Materials & Design 209 (2021) 109943

Contents lists available at ScienceDirect

Materials & Design

journal homepage: www.elsevier.com/locate/matdes

Accurate and Computational: A review of color reproduction in

Full-color 3D printing
Jiangping Yuan a,b, Guangxue Chen a,⇑, Hua Li a, Hartmut Prautzsch b, Kaida Xiao c
a
State Key Laboratory of Pulp and Paper Engineering, South China University of Technology, Guangzhou, China
b
Institute for Visualization and Data Analysis, Karlsruhe Institute of Technology, Germany
c
School of Design, University of Leeds, Leeds, UK

h i g h l i g h t s g r a p h i c a l a b s t r a c t

 Shows the state of the art of color

reproduction methods in full-color 3D
printing.
 Judges the suitability of current
methods expanded to a universal
color standard.
 Cases analysis of three key issues for
achieving accurate full-color 3D
printing.
 States a potential theoretical
framework suitable for various 3D
printing materials.

a r t i c l e i n f o a b s t r a c t

Article history: As functional 3D printing becomes more popular with industrial manufacturing applications, it is time to
Received 1 March 2021 start discussing high-fidelity appearance reproduction of 3D objects, particularly in faithful colors. To
Revised 12 May 2021 date, there is only limited research on accurate color reproduction and on universal color reproduction
Accepted 27 June 2021
method for different color 3D printing materials. To systematically understand colorization principles
and color transmission in color 3D printing, an exhaustive literature review is stated to show the state
of the art of color reproduction methods for full-color 3D printing, such as optical parameter modeling,
Keywords:
colorimetric difference evaluation, computer aided colorization and voxel droplet jetting. Meanwhile, the
3D printing
Color reproduction
challenges in developing an accurate color reproduction framework suitable for different printing mate-
Material modeling rials are fully analyzed in this literature review. In full-color 3D printing, coloring, rendering and acqui-
Quality evaluation sition constitute the core issues for accurate color reproduction, and their specific concepts are explained
Digital Fabrication in concrete examples. Finally, the future perspectives of a universal color reproduction framework for
accurate full-color 3D printing are discussed, which can overcome the limitations of printing materials,
combined with computational boundary contoning.
Ó 2021 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2. Overview of color reproduction methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

⇑ Corresponding author.
E-mail address: [email protected] (G. Chen).

https://doi.org/10.1016/j.matdes.2021.109943
0264-1275/Ó 2021 The Author(s). Published by Elsevier Ltd.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
J. Yuan, G. Chen, H. Li et al. Materials & Design 209 (2021) 109943

2.1. Optical parameter modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2.2. Colorimetric difference evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3. Computer-aided colorization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.4. Voxel droplet jetting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3. Color accuracy issues of full-color 3D printing techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.1. Coloring issues of full-color 3D printing processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2. Rendering issues of full-color 3D printing processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.3. Acquisition issues for full-color 3D printing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4. Outlook on an accurate color reproduction framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Funding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Declaration of Competing Interest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

1. Introduction ing processes, as shown in Fig. 1 [11]. Those color 3D printing tech-
niques mentioned above realize the binder jetting (BJ), sheet
3D printing is regarded as a revolutionary technology for cus- lamination (SL), and material jetting (MJ) principles, respectively.
tomized fabrications. With the increasing functionality of printed While vat photopolymerization (VPPM) based color 3D printers
parts, 3D printing techniques have been widely applied in, e.g., are also capable of reproducing a wide range of colors, their color
aerospace engineering, automotive electronics, biomedicine, cul- reproduction performance is not as good as material jetting based
tural creativity, and geographic information industries to manufac- printing [12]. Color reproduction on the VPPM type is achieved by
ture complex components and special parts with competitive curing the entire surface of each light-curing resin layer using mul-
performance [1–5]. Besides material formulation, microstructure tiple linear curing arrays, which limits color mixing and pigment
optimization and mechanical properties, which are in the fore- path distribution [13,14]. Color 3D printing devices based on the
ground in these applications, the surface colors of 3D printed three remaining principles in Fig. 1, e.g., material extrusion (ME),
objects are important and do no longer meet modern aesthetic powder bed fusion (PBF), and direct energy deposition (DED) have
and practical needs [6,7]. Since 2014, color 3D printing techniques been developed first with material chambers and then with print-
emerged mainly for multicolor 3D visualizations with powder- ing or extrusion heads, but their color reproduction capabilities are
based 3D printers produced by 3D-systems [8], paper-based 3D also far from satisfactory [15–17]. There are benchmarks for 3D
printers produced by MCOR-technology [9], and plastic-based 3D printing quality evaluation [18–20], but benchmarks such as stan-
printers produced by Stratasys [10]. According to the ISO/ ASTM dardized test sets, measurement and viewing conditions for evalu-
52900, there are seven general principles for additive manufactur- ating color 3D printing setups are ignored. Combined with the

Fig. 1. Comparison of kernel processes used in color 3D printers.

2
J. Yuan, G. Chen, H. Li et al. Materials & Design 209 (2021) 109943

above literature analysis, the fundamental reason for the slow pro- differences are demonstrated: The Fig. 3 (a)-(c) demonstrate the
gress in accurate color reproduction of current 3D printing devices light path transmission for conventional print, irregular print,
can be summarized as a lack of a universally accepted color repro- and color 3D print with a standard measurement combination of
duction theory and an evaluation system [21]. a 0° illumination angle and a 45° viewing angle (Short for
Moreover, the diversity of principles, technologies and materi- 0°as45°). In Fig. 3 (b), a measuring instrument would not correctly
als used in existing color 3D printing techniques has led to the sit- measure the angular and spectral distribution of the reflected light
uation where further developments happen scattered across at the measurement point with this viewing angle. Moreover, in
different fields. Mainly, there are 3D printing device developers Fig. 3 (c), the angular and spectral distribution of the reflected light
in manufacturing and computer science, 3D printing materials sci- changes even more dramatically than that in Fig. 3 (b). Practically,
entists, and color scientists in the fields of printing and optics a spectrophotometer with an integrated sphere can improve mea-
working on what is commonly called color or multicolor 3D print- surement accuracy, but the problems above cannot be overcome
ing [22–24]. In order to stress that an accurate and realistic appear- when the geometry of the 3D printed sample is more complex with
ance reproduction of 3D objects is more than just multicolor a large number of tiny colored curved surface with spatial irregu-
printing, a terminology full-color 3D printing is suggested for this larities and uneven optical features. Meanwhile, the X-Rite
advanced area of 3D printing that provides more accurate repro- research center performed a measurement survey for color pig-
duction [25]. In full-color 3D printing, the color of each point on ments stacked on flat prints and reported changes under different
a printed object should be accurately reproduced, but current pro- viewing and illumination angles as shown in Fig. 3 (d)-(f) [33].
cesses don’t really achieve this goal. This is geared towards person- Fig. 3 (d) & (e) vividly illustrate all standard combinations available
alized fabrication that requires fine color gradations in the target for current measurement instruments while Fig. 3 (f) shows the
area. However, the full-color 3D printing also suffers from printing results obtained by four different color measurement instruments
speed and build size for mass production [26]. with these angle combinations for the same sample point. To meet
Moreover, consistent with different printing substrates, there the demand for better measurements, instruments with a wider
are paper-, plastic-, powder-, metal-, food- and organism-based range of standard angle combinations are constantly being devel-
full-color 3D printing processes [27]. The color of full-color 3D oped in color science.
printed objects is perceived from the absorption and reflection In addition, a precise measurement mechanism to develop full-
characteristics of the selected coloring substrates. For non- color 3D printing with accurate color reproduction is still
scattering materials (e.g., transparent inks), this follows the princi- unknown. So far, color evaluation is mostly based on chromaticity
ple of subtractive color mixing [28,29]. differences in flat areas. However, under the current limitations on
Color appearance of subtractive color mixing of 2D printer can illumination and viewing angles, measurement deficiencies are
be accurately predicted by colorimetric measurement standardized still prevalent. Material appearance reproduction is discussed in
by the International Commission on Illumination (CIE) with LUT- the proceedings of the 27th Color and Imaging Conference pub-
based color characterization model recommended by the Interna- lished by the Society for Imaging Science and Technology in 2015
tional Color Consortium (ICC) [30]. However, it is not actually [34]. These discussions show that material appearance can be
effective for full-color 3D printing, although both 3D printing and regarded as a visual sensation reproduction, which indicates that
2D printing utilize digital discretization for color transmission the accuracy of color reproduction is affected by other aspects,
[31]. The key discrepancies are illustrated in detail in Fig. 2. The such as size, texture, gloss, transparency, and opacity. A typical
arrows on the left side in the upper and lower row, respectively, example for current efforts is the European Union’s Horizon 2020
symbolize the digital discretization process in 2D printing. Mean- research and innovation program led by the Norwegian Color
while, the digital discretization process for 3D printing can be and Visual Computing Laboratory with the goal to explore repro-
divided into two steps: the first step is similar with that in 2D ducing and measuring material appearance [35]. Achieving and
printing; the second step is illustrated by the arrow on the right controlling accurate color reproduction of full-color 3D printing
side in the upper row. It should be noted that these two steps is a tricky interdisciplinary issue although some practical models
are directly integrated into the slicing process of color 3D printing and specific theories have been developed in industrial manufac-
systems. turing, artistic aesthetics, color science and computer science
Obviously, the lack of a unified and accepted accurate measure- [36,37].
ment theory allowing to evaluate color reproduction quality for In this review, the discrepancies in current color reproduction
full-color 3D printed products is a challenging problem [32]. The as well as in color measurement on the one hand and evaluation
principles followed by color measurement tools in 2D printing on the other are both elaborated. In Section 2, the state of the art
are not applicable to color 3D entities. In Fig. 3, the essential of color reproduction methods in full-color 3D printing based on

Fig. 2. Digital discretization in 2D printing and 3D printing.

3
J. Yuan, G. Chen, H. Li et al. Materials & Design 209 (2021) 109943

Fig. 3. Schematic diagram of measurement variances and influencing factors for opaque objects: (a) smooth or flat substrate measurement with single illumination and
viewing angle; (b) uneven or irregular substrate measurement with single illumination and viewing angle; (c) color 3D object measurement with single illumination and
viewing angle; (d) color patch measurement with 45° illumination and multi-viewing angles; (e) color patch measurement with 15° illumination and multi-viewing angles;
(f) color comparison of different measurement angles under two types of illumination.

specific printing substrates is presented in detail. In Section 3, core refraction accumulation. For a translucent 3D object, light diffusion
accuracy issues in color transmission for full-color 3D printing and is used when there are many changes and few separations aniso-
2D printing are compared with theoretical examples. Finally, the tropic media, while light scattering is used a small number of
enlightening perspectives are also given for exploring the research changes in the medium and more distinct separations [38,39].
highlights with the potential for a universal accurate color repro- However, the light diffraction predictions are rarely studied for
duction in cross-substrate full-color 3D printing. printing materials.
Importantly, it is the accurate modeling of light scattering from
2. Overview of color reproduction methods translucent objects that is the central challenge of color rendering
of materials yet to be solved in computer graphics, especially for
In this section, an overview of existing color reproduction meth- anisotropic and heterogeneous translucent materials [40]. Light
ods used in full-color 3D printing is presented. In terms of color scattering from translucent materials can be divided into planar
transmission, these methods can fall into the following four sub- scattering and volumetric scattering in terms of the number of
categories ‘‘optical parameter modeling”, ‘‘colorimetric difference shape dimensions [41]. Since the fabricated objects usually belong
evaluation”, ‘‘computer-aided colorization”, and ‘‘voxel droplet jet- to 2.5- and 3-dimensional material rendering, volumetric render-
ting”. In principle, optical parameter modeling and colorimetric ing prediction and its optical density material approximation dis-
difference evaluation are synthetic tools based on theoretical tribution have become a research hotspot in the field of
results in optics while computer-aided colorization and voxel dro- computer graphics [42]. The works by Théo Phan Van Song on
plet jetting are developed by analyzing printing results and four-flux volume scattering models allows for color prediction of
behavior. 2.5D printed objects for different thicknesses and color inks using
colorimetric and spectral data [43–46]. In addition, with the devel-
2.1. Optical parameter modeling opment of texture mapping functions for translucent materials,
high-resolution appearance description models for translucent
Optical parameter modeling used physical measurements and is 3D objects with isotropic scattering parameters is simulated by
based on multivariate functions describing radiative light transfer bidirectional scattering distribution functions (BSDF) [47,48]. To
performances from arbitrary 3D objects illuminated by a given combat the edge-loss in reflectance measurement on translucent
light source. In general, these radiative transfer performances materials including colored and opalescent materials, an improved
mainly consist of light reflection, light refraction, light diffusion, bidirectional scattering-surface distribution functions (BSSDFs) is
and light diffraction [38]. In terms of transmission properties of proposed for comprehensive numerical simulation of translucent
printing materials, any 3D object can be divided into transparent, appearance in [49]. The BSSDFs, which consist of bidirectional
translucent, and opaque geometries [39]. Light reflection of opaque scattering-surface reflectance distribution functions (BSSRDF) and
3D objects and light transmission of transparent 3D objects can be bidirectional scattering-surface transmittance distribution func-
easily measured by current standard optical devices whereas light tions (BSSTDF), are difficult to measure, to store, and to compute
transmission of transparent materials is characterized by total light [50]. Since the effective scattering distance of most commercial

4
J. Yuan, G. Chen, H. Li et al. Materials & Design 209 (2021) 109943

printing materials for 3D printing is substantial, the strategies used combinations with a 5° resolution, than the CIE recommend for
with the models above mainly rely on local pure scattering control 2D printing [33]. It greatly enriches and improves acquisition accu-
with the result that visible inaccuracies become more apparent on racy for surface appearance of complex color 3D objects including
color 3D printed parts [51]. Subsequently, scattering-aware texture color, texture, and gloss. Worldwide, this huge and costly device is
reproduction for planar 3D prints had been developed by an itera- unparalleled in its quality measurements by other machines. How-
tive optimization scheme controlling a fully volumetric material ever, the measurement process is time-consuming and requires a
distribution and a commercial five-tone 3D printing process. Two large memory space. For example, the appearance acquisition time
years later, this method has been extended to arbitrarily shaped of a common basketball surface takes about 8 h, and its raw data
objects by a full-fledged optimization on a small domain to effec- reaches 1.5 TB when stored in the full bidirectional texture func-
tively compensate for internal light scattering [52]. These models tion (BTF) format (e.g., chunked data).
are well studied in 3D object rendering, but are poorly investigated Subsequently, the X-Rite, a leading United States-based color
in high-fidelity full-color 3D printing. measurement equipment company developed a portable commer-
Undoubtedly, color reproduction models adopting light reflec- cial hemispherical 3D object appearance acquisition prototype pro-
tions for opaque 3D objects have always been mainstream tech- viding multiple illumination angles and viewing angles in 2016,
niques in full-color 3D printing. Classical models include the naming TAC7 3D scanner. Apart from utilizing color measurement
Blinn-Phong model, Ward or Cook-Torrance model, and simple principles of 2D printing, the output data of this 3D scanner can
bidirectional reflectance distribution function (BRDF) models either be saved as a sparse BTF or a SVBRDF normal map. To better
[53–55]. With the development of texture mapping functions for understand the critical advances in optical parameter modeling,
opaque 3D objects, high-resolution appearance description models the corresponding data acquisition and transformation is illus-
for even more complicated 3D geometries were formed by the spa- trated in Fig. 4. Compared to the BRDF, the SVBRDF has an addi-
tially varying BRDF (SVBRDF) [56,57]. Moreover, microfacet model- tional bivariate mathematical function, which enables the
ing is widely applied to realistic surface appearance measurement representation of planar textures on opaque materials. The BRDF
applications [58]. is sufficient for acquiring functional material appearances for
Typically, the State Key Laboratory of Modern Optical Instru- plastic- and metal-based 3D prints, while the appearance repro-
ments at Zhejiang University in 2013 reviewed recent BRDF mod- duction of multi-material 3D printed products with color complex-
els applied to appearance measurements of colored 3D freeform ity must use the SVBRDF. Fig. 4 (d) stands for combining dense
models and proposed a realistic color rendering method for 3D illumination and viewing angles in a color 3D scanning instrument.
objects with a quantifiable appearance reproduction [59]. In the Of course, combining illumination and viewing angles in an instru-
French National Metrological Institute for radiometric and photo- ment, their numbers need to be balanced, meanwhile, larger num-
metric quantities (LABORATOIRE NATIONAL DE METRILOGIE ET bers mean more accuracy but less efficiency. In Fig. 4 (g) & (h), the
D́ESSAIS (LNE), an independent research institute), a program mea- color perception of a sample appearance is significantly different
suring the BRDF specular peak was started in 2008 to explore the from each other, for instance, the material appearance of the color
precise measurement of gloss attributes, and launched a lab facility 3D object captured by the 3D-DEMO is more saturated and
named Conoscopic Device for Optic Reflectometry (ConDOR) brighter. Conversion between full BTF, sparse BTF and conventional
around 2015 [60]. As its core member, Gaël Obein formed a new digital format is possible, but each conversion has a loss of color
optical measurement team at the Conservatoire National des Arts information.
et Métiers (CANM) to further develop a new and large 3D scanning For the TAC7 3D scanner, the X-Rite color research group devel-
device (3D-DEMO) to capture the material appearance of arbitrary oped a new appearance exchange format (AxF) based on the planar
3D objects with complex surface properties. The 3D-DEMO enables color exchange format (CxF) used in the ISO117972 standard,
to measure a wider range of illumination and viewing angle which consists of a diffuse albedo map, a specular roughness

Fig. 4. The data acquisition and transformation in optical parameter modeling: (a) BRDF model; (b) SVBRDF model; (c) conventional digital format; (d) acquisition with
multi-viewing and illuminating angles; (e) 3D-DEMO scanner; (f) TAC7 scanner; (g) full BTF captured by 3D-DEMO scanner; (h) sparse BTF captured by TAC7 scanner.
Modified examples used with permission from [33], copyright (2016) X-Rite.

5
J. Yuan, G. Chen, H. Li et al. Materials & Design 209 (2021) 109943

map, a specular color map, a normal map (replace bitmaps as when it comes to reproduce colored 3D objects with high gloss
needed), a surface orientation map (to capture anisotropy), a materials. Because the combined effects of shape and material
height map (also known as displacement map), a Fresnel map, an properties is poorly understood, the optical parameter modeling
index of refraction map, an alpha map and car paint (CPA) [33]. is still difficult to set up an accurate color reproduction method
The first version of the TAC7 3D scanner and its AxF format was for all types of 3D printing materials.
formally released at the ICC 3D Display and Color 3D Printing sym-
posium held at Taipei City in 2016, but was not recommended by
2.2. Colorimetric difference evaluation
most of ICC technical committees due to its high price and subop-
timal accuracy for measuring colored 3D objects with complex
Colorimetric measurements are used to evaluate full-color 3D
gloss and transparency. Meanwhile, like the AxF structure, a sur-
printing. They are based mainly on numerical minimization of
face appearance description tag is defined and embedded into
color differences to match the visual perception of the human
the iccMAX profile to construct the chromatic Ward BRDF model
eye. So far, they are not used to automatically adjust the printing
consisting of multiple illumination angles and viewing angles
parameters or to pretreat the input for current full-color 3D print-
[61]. Later, an internal ICC technical committee for appearance
ers, yet the advantages of a colorimetric measurement are its sim-
acquisition and measurement of colored 3D objects was estab-
plicity, speed and low-cost, but it does not work well with
lished with the X-Rite research center.
transparent or glossy multi-jetting materials [65,66].
Further, the 3D manufacturing format (3MF) is a 3D print file
Significantly, the color quality of 3D printed objects has been
format with color attributes has been developed by a key industry
improved continuously by minimizing color differences in full-
consortium led by Microsoft. It is widely used in additive manufac-
color powder-, plastic- and paper-based 3D printing. In 2012, Maja
turing and physical visualizations of digital material. The main-
Stanic at the University of Zagreb in Croatia printed a highly satu-
stream manufacturers of color 3D printers have been promoting
rated artwork with the Z510 powder-based 3D printer and mea-
this format since 2015 [62]. For instance, the HP 3D printing divi-
sured CIE L*a*b* values on selected smooth surface segments of
sion implemented this format to all its color 3D printer series in
the printed model with the Gretag-Macbeth XTH spherical spec-
2016. Likewise, MCOR-technology Inc. used it for their relative
trophotometer [67,68]. The printed surfaces presented vivid colors
color 3D printing software in 2018.
in the reproduced artwork. Maja Stanic developed a metric using
Optical parameter modeling provides more accurate optical
the CIEDE2000 color difference formula and further proposed a
measurements and color appearance description formats for 3D
color compensation lookup table for powder-based 3D printing
printing, which affects the surface color output. Based on the tex-
tests to provide direct color selection for general 3D printing users.
ture image gamut mapping principle, algorithms and frameworks
In 2013, Kaida Xiao et al. from the University of Liverpool
for color gamut mapping and gloss mapping are used to explore
printed soft prostheses by the SLA 3D printer and took the Minolta
accurate surface reproduction of 3D printed objects [63,64]. This
CM-2600D spectrophotometer to measure the CIE DE*ab values of
technique helps to improve the color reproduction accuracy of
surface color points on the facial prosthesis shown in Fig. 5 (a). Fur-
full-color 3D printing given pure translucent materials but fails
ther, they optimized the color reproduction consistency for biolog-

Fig. 5. Examples for color reproduction optimization based on colorimetric differences:(a) Kaida Xiao group; (b) Guangxue Chen group; (c) Pei-li Sun group. Examples used
with permission from [71], copyright (2014) Springer Nature; [72], copyright (2014) the authors; [77], copyright (2018) Society for Imaging Science and Technology; [78],
copyright (2016) the author; [80], copyright (2018) the author; [81], copyright (2015) SPIE; [82], copyright (2016) Society for Imaging Science and Technology; [83], copyright
(2016) the authors.

6
J. Yuan, G. Chen, H. Li et al. Materials & Design 209 (2021) 109943

ical tissues by a multi-parameter feedback framework [69–72]. In difference evaluation methods for color 3D objects in 2017 for a six
the upper part of Fig. 5 (a), their 3D prosthetic printing system is years period [85].
illustrated. It consists of the following six steps: 3D image acquisi- Besides, colorimetric difference evaluation methods have also
tion, 3D image design, color management, color texture mapping, been developed for other substrates. For example, a color research
3D color printing, post processing. The lower part of Fig. 5 (a) group at Wuhan University presented a complete dye matching
shows the color test chart used in their multi-parameter feedback reproduction framework for digital printing on textiles using the
framework and the effect of the imported objects on facial Kubelka–Munk theory, which can be extended to the 3D printing
replication. of textiles [86]. Further, the 2.5D appearance reproduction evalua-
Since 2013, the color 3D printing group led by Prof. Guangxue tion method and the relative ink droplet characterization of ele-
Chen at South China University of Technology focused on the color vated printing was analyzed by the Océ-Canon company in the
reproduction theory and the process optimization for paper-based Netherlands [87,88]. The core idea behind colorimetric difference
full-color 3D printing with the MCOR IRIS 3D printer and the evaluation is to match points on the smooth regions of printed
MIMAKI UV-photocurable 3D printer for applications in the cul- objects to minimize their color differences. However, sampling
tural creativity industry [73–77]. Based on a colorimetric differ- color points only on smooth areas does not always cover key or
ence evaluation, this group explored accurate color reproduction complex color features, because colorants are more difficult to
strategies of A3 paper-based 3D printing (The latest MCOR print precisely on rough than on smooth regions. Optical properties
paper-based 3D printer uses A4 paper) for precise visualization such as gloss, transparency and texture are challenging in integrat-
of surgical models (originate from MRI data), military 3D maps ing a color difference evaluation model for different coloring mate-
(originate from DEM data) and improved high-fidelity reproduc- rials [89]. Although the principle of colorimetric difference
tion of expensive 2.5D oil paintings (originate from 3D scanning evaluation is well-known in 2D printing to optimize color repro-
data) [78–80], see Fig. 5 (b). Compared with current sequential duction, applying this principle to the full-color 3D printing to
printing of color sliced layers using four primary color inks, this achieve accurate color measurements for complex 3D curved sur-
group proposed a new computational framework to adjust the faces with existing color measuring instruments remains a big
order and amount of sliced color layers and transparent layers to challenge.
attenuate the jagged boundary effect of UV-photocurable 3D
printed objects, as shown in the lower part of Fig. 5 (b). This frame- 2.3. Computer-aided colorization
work implemented staircase effect mitigation with a fixed sliced
layer thickness by a material jetting 3D printing process with a Computer-aided colorization is an indirect technique for accu-
printing accuracy of 10 mm providing new ideas for optimizing sur- rate color reproduction of 3D printed objects based on the combi-
face reproduction in other 3D printing processes on a bigger print- nation of computer geometry simulation and digital colorization. It
ing scale. can be categorized into computational transferred printing, com-
In the period 2015–2016, Pei-Li Sun et al. at the Institute of putational block coloring and computational contour inkjet print-
Color and Vision of the Taiwan University of Science and Technol- ing. In the following, these sub-categories are explored in more
ogy attempted to minimize color differences to improve the accu- detail. There are two computational transferred printing tech-
racy of color reproduction for plastic-based color 3D printing [81– niques: the computational hydrographic printing (CHP) and com-
83]. The Fig. 5 (c) depicts 3 different specific studies with a BJ-type putational thermoforming printing (CTP).
full-color 3D printer: (1) A puzzle-like omni-directional color test Generally, hydrographic printing refers to a unique color repro-
target for estimating color variations across different surface direc- duction method controlled by computational texture registration
tions. The measuring results can be used to reduce the color vari- algorithms that transfer accurately printed high-resolution color
ations (See the upper part of Fig. 5 (c)). (2) Gamut mapping from patterns on a thin film to the designated surface of the 3D objects
sRGB display to 3D prints shown in the middle part of Fig. 5 (c). to be manufactured. The principle of traditional hydrographic
The result shows SGCK (chroma-dependent sigmoidal lightness printing is that printed polyvinyl alcohol (PVA) film tiles on a water
mapping and cusp knee scaling) model is preferred to Clipping tank with room temperature wrap around the desired object until
(minimum DE clipping) and HPminDE (hue preserved minimum it is fully covered. Subsequently, the object together with the film
DE clipping). (3) Minimize color differences across 5 important is rotated several times in the water in only one direction [90]. The
illuminants shown in the lower part of Fig. 5 (c). The result is eval- traditional hydrographic printing is a simple and economical
uated by the perceptual differences and shows D65 illuminant method that enables the color features of complex geometries pro-
shows better preference. These test targets can provide an efficient duced with a wide range of materials in high quality. Nevertheless,
reference for regular colored 3D prints, but don’t take into account it is still difficult to accurately apply color textures to printed 3D
color variations in the build orientation. In addition, SGCK is not objects with doubly curved surfaces.
necessarily the best algorithm for color reproduction among the Noteworthy, the low-cost and efficient computational frame-
color gamut mapping algorithms for texture images. work shown in Fig. 6 (a) to calculate the precise stretch and distor-
In 2016, Gui-Hua Cui et al. at Wenzhou University proposed a tion of the printed color pattern on transfer films has been
color difference prediction framework by calculating surface chro- proposed primarily by the State Key Lab of CAD&CG at Zhejiang
matic values of 3D printed objects with psychological scale exper- University [91]. Accurate surface color texture alignment is com-
iments to evaluate and optimize the color finally reproduced. This puted by simulating the transfer film and executing it by a precise
prediction framework originated from inherently imperfect color calibration system controlled by a semiautomatic mechanical
difference models (CIE DE*ab, CIE DE*00) in 2D printing, whose apparatus with a 3D vision capturing device. In computational
number of flat color blocks printed by paper-based full-color 3D hydrographic printing, they achieve vivid color reproduction in
printing is too small [84]. Currently, no numerical computational three steps: (1) First, one prints an achromatic 3D model; (2) sec-
model has been constructed for this prediction framework. ond, a suitably distorted texture is printed on a PVA film by 2D
In order to accelerate the progress in global standardization of inkjet printing; (3) third, the film is oriented, dipped into water
accurate color reproduction, the CIE 8th Division appointed the with room temperature and finally transferred to the printed
8–17 technical committee (CIE TC 8–17: Methods for Evaluating object. On the tiger mask in the lower part of Fig. 6 (a), two red
Color Difference between 3D Color Objects) to develop formal color marked areas are compared to show how the film thickness

7
J. Yuan, G. Chen, H. Li et al. Materials & Design 209 (2021) 109943

Fig. 6. Computational transferred printing workflow and it output comparison: (a) hydrographic printing; (b) thermoforming printing. Examples used with permission from
[91], copyright (2015) Association for Computing Machinery (ACM); [92], copyright (2016) ACM.

influences the color reproduction of 3D objects by two simulated block coloring is also ideal for complex 3D objects with microstruc-
hydrographic films. tures. Joint research efforts at the University of Cagliari and New
On the other hand, computational thermoforming utilizes the York University led to several decomposition techniques: a disjoint
melting decomposition of a printed plastic sheet to transfer the decomposition into height-field blocks, volumetric decomposition,
attached ink to the designated surface. The computational thermo- surface segmentation, and axis-aligned height-field segmentation
forming method was developed and improved jointly by a research [93]. These techniques have been applied to the FDM (Fused Depo-
group at the ETH Zurich and a group at the Disney Research Centre sition Modeling) 3D printers for large-scale pure-color rather than
[92]. This method provides a simulation algorithm to convert a tex- gradient-color 3D models. The axis-aligned height-field segmenta-
tured digital 3D model into a 2D image printed on the plastic sheet tion provides a potential way to achieve cheaper color 3D prints
prior to pyrolysis. The upper part of Fig. 6 (b) illustrates the three and its key steps are illustrated in Fig. 7 (a). This segmentation
phases of thermoforming simulation and the finally formed result. scheme may result in a decrease in the final forming efficiency
The color of the dark blue sheet visualizes the thickness of the sim- and overall strength due to too many segmented sub-blocks. Previ-
ulated plastic sheet, changing from 0.5 mm to 1.0 mm. On the bot- ously, Jiangping Yuan et al. from the South China University of
tom of Fig. 6 (b), this method is compared to the powder-based 3D Technology proposed a 3D cutting & bonding frame (3D-CBF)
printing technique and hydrographic transfer technique for a Plu- method for full-color paper-based 3D printing, which can be inte-
tus cat. The computational thermoforming replica showed a higher grated into the slicing phase [94,95]. Fig. 7 (b) shows cutting angle
resolution than powder-based 3D print, but fewer artifacts in its settings and layout options for a cuboid model with specific color
flat regions than that reproduced by the hydrographic technique. features to test the adaptability of the proposed 3D-CBF for com-
However, this fabrication technique is prone to color deterioration plex cases. In Fig. 7 (c), the effect of three cutting angles and five
issues due to high temperature and vacuum suction. In summary, adhesive brands on each 3D printed banana model is demonstrated
it should be stated that the computational transfer printing is not in detail. In addition, the effect of five bonding temperatures on the
available for 3D printed entities that are sensitive to water or heat. physical visualization of banana with Garefu-brand adhesive and
It must be said that computational hydrographic printing and ther- 90°cutting angle was evaluated by its bonding strength and surface
moforming printing are good for color reproduction of oversized smoothness. Based on all the above cases, strategies for subblocks
3D objects formed by general 3D printers and ordinary inkjet print- decompositions can help to reduce the difficulty of accurate color
ers at very low cost. reproduction.
With respect to computational block coloring, it is a new way to Likewise, computational contour inkjet printing is adapted
meet accurate color reproduction for large-sized 3D entities with inkjet printing to transfer the color of a filament to the contour
concave and convex surface areas, based on surface segmentation of any layers by an additional melting process. The original pur-
algorithms and splicing strategies. The principle of computational pose of this method was to develop practical paths by unitizing

8
J. Yuan, G. Chen, H. Li et al. Materials & Design 209 (2021) 109943

Fig. 7. Strategies for color 3D object decomposition: (a) axis-aligned height-field segmentation; (b)~(c) 3D cutting & bonding frame method. Examples used with permission
from [93], copyright (2018) ACM; [94], copyright (2017) Emerald; [95], copyright (2017) Springer Nature.

customized color filaments and a single spray nozzle to improve seem to promise more accurate color reproduction. How to simu-
color reproduction accuracy of the FDM 3D printer. The prepared late and compute the colorization flow in each bonding paper sheet
color filaments were colored by certain conventional 2D printing is another important issue. Taking binder jetting as an example, to
processes. The color position on the filament is computed to match improve the color reproduction accuracy of powder-based full-
the target color position on the contour of each printed color layer. color 3D printing by contour coloring, one still needs to fully
At Jinan University, Yang et al. presented a new method for the understand how powder particles behave under their formation
multi-color FDM 3D printing process with color adherences pro- and how color pigments penetrate them. Then together with
duced by silk-screen printing, based on the principle of low- advanced algorithms, accurate full-color 3D printing for various
dimensional colorization [96]. Its limitation is that the surface printing materials may become possible with this method.
color of the next layer is susceptible to inevitable pollution by
residual colorants in the operating nozzle. Moreover, Jennifer 2.4. Voxel droplet jetting
Lewis’s group from the Harvard John A. Paulson School of Engineer-
ing and Applied Sciences proposed a spray regulator for a single Since 2015, a new data-driven method which can be defined as
nozzle to quickly distribute voxelized droplets of multiple coloring the voxel droplet jetting, has been developed for higher accuracy in
materials. This device has specific microfluidic structures installed full-color 3D printing by computing discrete ink droplets at voxel-
in the print nozzle to switch and combine up to eight basic coloring level. This has been applied rapidly in the field of customized col-
materials while avoiding residues at the nozzle [97]. One device orful geometries due to its precision close to conventional 2D
can integrate 128 nozzle-based arrays to print 128 gradient colors printing with advanced color separation methods and printing
at the same time, which provides better color control for the com- heads. For example, commercial 2D printing heads from world-
putational contour in using multi-liquid coloring materials. wide companies such as Epson, Fuji, Konica Minolta, Kyocera,
In general, computer-aided colorization allows better color Panasonic, Ricoh, Toshiba, Xaar etc., can be also customized for
reproduction in plastic-based full-color 3D printing. For paper- color 3D printers. Fig. 8 shows respective examples and specific
based full-color 3D printing, the computational block strategies principles for the voxel unit generation of full-color 3D printing.
9
J. Yuan, G. Chen, H. Li et al. Materials & Design 209 (2021) 109943

Fig. 8. Voxel unit generation: (a) halftoning printing; (b) contoning printing; Examples used with permission from [98], copyright (2015) ACM; [102], copyright (2017) ACM.

However, before describing voxel unit printing, three major chal- pling or unacceptable dot patterns in the contoning image. The rea-
lenges need to be addressed: the accurate generation of voxel dro- son for this behavior is that in a contoning image every pixel is
plets, the precise prediction of the droplet jetting behavior and the printed with its individual tone whereas in an halftoning image,
arrangement of color concentrations. different color tones are generated by areas of different sizes cov-
Initially, based on scalable error diffusion algorithms in 2D ered by the same single color. Then in 2018, the Wyss Institute for
printing, the Fraunhofer IGD (Institute for Computer Graphics Biologically Inspired Engineering at Harvard University and the
Research in Darmstadt) developed a novel traversal algorithm for MIT Media Lab’s Mediated Matter Group developed a simple
solid voxelization to achieve full-color visualization with multi- multi-material transformed voxel printing method by converting
jetting 3D printers [98,99]. The left part of Fig. 8 (a) shows a discontinuous data sets into dithered deposition marks for more
halftoning traversal simulation and details of the slice-wise vox- accurate color reproduction with four cases [103]. This method
elized surface, and its right part compares facial texture reproduc- focuses on preventing data alteration and boundary information
tion from their software simulation and color 3D print. The loss for the color transition from a digital representation to the
reproduction of each color on the texture is preserved, but differ- physical material compositions in a multi-material jetting process.
ences in visual perception of the current overall appearance are Theoretically, given concise and computational spectral absorp-
still perceptible. Subsequently, this adaptive error diffusion tion, a weighted regression prediction model can be set up for arbi-
halftoning approach has been further standardized for the RGBA trarily many resin-based colorants and can further be extended to
(Red Green Blue Alpha) data format to control color layers of 3D a color reproduction accuracy evaluation method for current all
color objects [100]. With general multi-material jetting 3D print- plastic-based color 3D printing processes.
ers, color is represented discretely by ink drops at voxel positions. However, one of the biggest limitations is that voxel coloriza-
For instance, the GrabCAD Voxel 3D Printers launched by Stratasys tion in plastic-based full-color 3D printing is achieved by the direct
in 2017 showed the trans-scale 3D printing at voxel level and accumulation of only liquid droplets without further active mate-
image-based texture reproduction for transparent 3D entities rials, while a second material with a different physical form is used
[101]. Soon afterwards, the MIT CSAIL Computational Fabrication in paper- and powder-based full-color 3D printing. For this reason,
Group proposed a new contoning printing method based on spec- the voxelized droplet coloring suffers the unpredictable hydrody-
tral vector error diffusion algorithm to solve the layout discretiza- namics in the penetration or immersion of solid substrates (paper
tion and color quantization issues in the resin-based 3D printing sheets, powder particles). More experimental studies are needed to
process [102]. understand and manage the voxel droplet jetting.
Vividly, the differences between halftoning and contoning (con-
tinuous toning) for a voxel unit generation are illustrated in Fig. 8 3. Color accuracy issues of full-color 3D printing techniques
(b). The left part of Fig. 8 (b) exhibits an image printed at 180 dpi
printer resolution, showing the outcomes of the halftoning (top) To judge whether a universal accurate color reproduction
and contoning (bottom) methods, with no signs of spatial subsam- framework can be derived for the full-color 3D printing methods
10
J. Yuan, G. Chen, H. Li et al. Materials & Design 209 (2021) 109943

Table 1
A summary of color reproduction methods and their main features.

Studies Color reproduction categories Process principles Materials Specimen Geometry

Théo et al. (2017, 2018) Optical parameter modeling MJ Plastic-based Color patches
Elek et al. (2017) Optical parameter modeling MJ Plastic-based Color blocks
Sumin et al. (2019) Optical parameter modeling MJ Plastic-based Earth model
Song et al. (2013) Optical parameter modeling ME Plastic-based Color sphere
Rouiller et al. (2013) Optical parameter modeling VPPM, BJ Plastic-based, powder-based Cylinder, spot domes
Cheydleur. (2016) Optical parameter modeling VPPM Plastic-based Spot basketball
James. (2016) Optical parameter modeling MJ Plastic-based Cubes
Vu et al. (2016) Optical parameter modeling MJ Plastic-based Texture-less samples
Stopp et al. (2008) Colorimetric difference evaluating BJ Powder-based Cubes
Page et al. (2017) Colorimetric difference evaluating MJ Plastic-based Colored bands
Stanic et al. (2008) Colorimetric difference evaluating BJ Powder-based Color patches
Xiao et al. (2013a, 2013b, 2014) Colorimetric difference evaluating VPPM Plastic-based Soft tissue prostheses
Sohaib et al. (2018) Colorimetric difference evaluating BJ Powder-based Facial prostheses
Yuan et al. (2016a) Colorimetric difference evaluating SL Paper-based Hemispheres, cone
Yuan et al. (2016b, 2017a, 2017b) Colorimetric difference evaluating SL Paper-based Banana, container, medical models
Yuan et al. (2019, 2020) Colorimetric difference evaluating MJ Plastic-based Oil painting, 3D chart
Yan et al. (2018) Colorimetric difference evaluating SL Paper-based Geographical model
Chen et al. (2016) Colorimetric difference evaluating MJ Plastic-based Oil paintings
Sun et al. (2015, 2016) Colorimetric difference evaluating BJ Powder-based Color charts
Pi et al. (2016) Colorimetric difference evaluating BJ Powder-based Santa model
Li et al. (2016) Colorimetric difference evaluating SL Paper-based Color patches
Wei et al. (2018) Colorimetric difference evaluating ME Plastic-based Colored cloth
Michael et al. (2016) Colorimetric difference evaluating MJ Plastic-based Wall ornaments
Parraman et al. (2018) Colorimetric difference evaluating MJ Plastic-based Personalized map
Christoph et al. (2016) Colorimetric difference evaluating MJ Plastic-based Medical models
Zhang et al. (2015) Computer-aided colorization CHP Plastic-based Stanford rabbit
Schuller et al. (2016) Computer-aided colorization CTP Plastic-based Plutus cat
Muntoni et al. (2018) Computer-aided colorization ME Plastic-based Fertility, Kitten, Chinese Lion
Yuan et al. (2017c) Computer-aided colorization SL Paper-based Hollowing structures, cylinder
Wang et al. (2019) Computer-aided colorization BJ Powder-based Cuboid, cubes
Yang et al. (2018) Computer-aided colorization ME Plastic-based Colorful cupcake, bottle
Skylar-Scott et al. (2019) Computer-aided colorization ME Plastic-based Colorful cubes
Brunton et al. (2015) Voxel droplet jetting MJ Plastic-based Bald head, Ruslan, Nefertiti, Apple
Alan et al. (2018) Voxel droplet jetting MJ Plastic-based Anatomy model, lion, lizard
Urban et al. (2019) Voxel droplet jetting MJ Plastic-based Buddha, Stanford’s dragon, Lucy model
Vahid et al. (2017) Voxel droplet jetting MJ Plastic-based Colorful mask, fish
Christoph et al. (2018) Voxel droplet jetting MJ Plastic-based Statue, patient hand

Table 2
A summary of pivotal accuracy issues and details from the studies.

Color reproduction Studies Printing material categories Accuracy issues Suitability

categories Index
Optical parameter modeling [43–46,51,52,59,60,61,62,63,64] Plastic-based, powder-based Acquisition issues wwwq
Colorimetric difference [67,68,69–72,73–77,78–80,81– Plastic-based, powder-based, paper- Acquisition issues, coloring wwq
evaluating 83,86,87,88] based issues
Computer-aided colorization [91,92,93,94,95,96,97] Plastic-based, paper-based, powder- Coloring issues, rendering wwww
based issues
Voxel droplet jetting [98,99,100,101,102,103] Plastic-based Rendering issues www

mentioned above, the literature on various color 3D printing tech- characteristics are also very complex. Overall, MJ, BJ, and ME have
niques is categorized in Table 1. This table contains most of the multiple types of color reproduction methods discussed above, but
principles seen in full-color 3D printing together with commonly SL and VPPM have fewer types of color reproduction methods for
used materials, and represents the state of the art in color repro- current full-color 3D printing processes.
duction. For each color reproduction category, Table 2 exhibits rel- In Table 2, the three key issues affecting the accuracy of color
ative subjective scale on the suitability for the universal color reproduction in full-color 3D printing can be found in the research
accuracy reproduction ranging from low (q) to high (wwwww). literature of colorimetric difference evaluating and computer-
Herein, one hollow marker has the same weight as half a solid mar- aided colorization, which both focus on two major types of color
ker. For each category, the core accuracy issues are also correlated accuracy issues, while the other methods are limited to one type
with current respective experimental studies. of color accuracy study. In terms of the printing material categories
Table 1 shows the principle statistical characteristics of the on which the color reproduction methods are based, the research
color 3D printing processes used in the four types of color repro- literature of colorimetric difference evaluating and computer-
duction methods, where the highest frequency is explored in MJ, aided colorization both covers the mainstream coloring materials,
followed by BJ, SL, ME and VPPM in that order. From the distribu- but there are minor differences in other color reproduction meth-
tion of printing materials, plastic-based types are explored in the ods. The research literature of optical parameter modeling method
first place, followed by powder-based types, and finally by paper- focuses on plastic-based and powder-based color 3D printing pro-
based types. From the geometric characteristics of the prints, all cesses, but its research on the optimization of color reproduction in
kinds of geometric shapes are involved, and their surface color paper-based full-color 3D printing is currently in a gap. The suit-

11
J. Yuan, G. Chen, H. Li et al. Materials & Design 209 (2021) 109943

3.2. Rendering issues of full-color 3D printing processes

Printing substrates and coloring materials in full-color 3D print-

ing differ from conventional 2D printing where paper or plastic
sheets are used as printing substrates. There is specific research
on how paper properties affect the color accuracy in 2D printing
[104]. Currently, printed paper can easily be measured and ana-
lyzed. For example, 2D printing service providers can purchase
tons of customized printing papers with consistent properties
and almost no deviations in their thickness. However, in color 3D
printing, every slice has a different slice underneath. Therefore,
whiteness, transparency, smoothness, and glossiness vary from
one layer to the next and moreover depend on all slices’ under-
neath. This dependency on the layer’s underneath is a main prob-
lem in accurate color reproduction and has not been addressed so
Fig. 9. The two coloring modes of the coloring phase: global coloring, left; boundary far. Furthermore, it is crucial to quantify the effect of substrate
coloring, right. thickness on surface characteristics by the measured density or
spectral values [105]. In addition, properties of layers need to be
ability index statistics show that current four categories of color
matched with the underlying layer substrates. The way in which
reproduction methods vary in the development of a generic color
the colored printing substrate is deposited and attaches to the
accurate reproduction framework for full-color 3D printing and
upside of the slice underneath is markedly different from 2D print-
that there is the greatest potential to achieve this goal using
ing. The rendering principles of 2D printing are depicted in Fig. 10.
computer-aided colorization.
In 2D printing on paper or film, the paper or film is regarded as
Color reproduction is standardized in 2D printing and has sim-
the printing substrates, and color ink consisting of specific discrete
ilarities with the full-color 3D printing. To obtain accurate color
ink droplets is applied with a fixed thickness. The color reproduc-
reproductions for printed color images, one needs to understand
tion accuracy is determined by the ink dot area coverage on the
the color transmission process. The color reproduction theory, a
paper surface. In the example shown in Fig. 10 (a) & (b), the
color quality evaluation framework and a color management sys-
magenta droplets with a thickness r are evenly distributed on
tem can contribute to the numerical models for faithful color
the paper covering half the area. This is fundamentally different
reproduction. This has been standardized for 2D printing tech-
in full-color 3D printing. In an analogous example shown in
niques over the past years. The academic community is aware that
Fig. 10 (c) & (d), the magenta dots cover a boundary layer of width
color reproduction behaves in 3D printing is distinct from in 2D
L of each slice with thickness H and lie on a regular 3D grid. How-
printing, and its intrinsic differences in the color transmission flow
ever, they are irregularly positioned and spaced on the boundary.
are not studied and understood in depth. The ‘‘standardization” in
Therefore, halftone screening techniques perform poorly in 3D
this manuscript is to express how to enhance the color reproduc-
printing. The area without magenta droplets is covered by achro-
tion consistency of full-color 3D printing using different printing
matic material such as UV cured white adhesive, which also influ-
materials. Therefore, the coloring, rendering and acquisition issues
ences the final color, its brightness and saturation.
for accurate full-color 3D printing are analyzed in detail.

3.3. Acquisition issues for full-color 3D printing

3.1. Coloring issues of full-color 3D printing processes
In full-color 3D printing, the input is a 3D data file while the
In current full-color 3D printing, surface color information of output is a colored 3D entity. Color measuring devices used in
printed object can be determined while its interior color is hard 2D printing do not work for arbitrary surfaces in general. Typically,
to detect and usually ignored. There are two ways to generate different reflectance values can be perceived at different points on
the outer surface color slice by slice: The first one is global coloring a general surface with the same viewing and illumination angle. In
where each slice or layer is completely colored by the surface color. addition, it is also illustrated that the measured colorimetric values
The second one is boundary coloring where each slice is only col- are not constant for a fixed point on the curved surface in Fig. 3 (c),
ored within some boundary layer of a certain thickness. These when changing either the number of illuminates used or their
two modes are shown in Fig. 9. angles relative to the surface. Worldwide, the color 3D printing
In 2D printing, the color thickness is determined by the primary industry is investigating these theoretical and practical issues with
ink. In 3D printing, however, the thickness of the colored boundary color 3D scanners as discussed in Section 2.1. In an example shown
region can vary. Essentially, the printing substrate and the thick- in Fig. 11 (a), a customized wine package with a green brand logo is
ness of the colored boundary layer determine the color reproduc- completely printed by a powder-based color 3D printer. The red
tion accuracy. The full-color 3D printing substrate usually dashed circle marks a tiny complex curved surface on the green
consists of the coloring material and the filling material in bound- logo that easily leads to poor color reproduction and acquisition.
ary coloring mode. The choice of an opaque and white filling sub- Besides the specialized optical measuring instruments
strate is more concise and efficient for the assignment of the ratio described together with Fig. 4, there are other color 3D scanners
between the coloring material and the filling material in each color employed that use high-definition cameras for shape and surface
layer. However, the filling material in UV curable 3D printing can texture acquisition [78,106]. To explain why poor acquisition per-
be either transparent or white UV ink, although this makes color formance occurs with such camera-based scanners, a half-cylinder
prediction more difficult. In addition, optical properties of these model that approximates the geometry sampled in the red circle is
achromatic substrates also affect the boundary colorization range. considered as an example shown in Fig. 11 (b). In a camera image
However, the effects of layer thicknesses and slicing strategies on of it, distances are not preserved, which leads to a mismatch of
the colored boundary region are not addressed completely. Fur- points on the color 3D object. Further, non-orthogonal projections
ther, when fixing the layer thickness, stacking the primary inks (as by a camera) are non-area preserving and lead to inaccurate
at the desirable boundary is also a new area to explore. color acquisitions. To reduce these accuracy problems, it appears
12
J. Yuan, G. Chen, H. Li et al. Materials & Design 209 (2021) 109943

Fig. 10. The color rendering principles in 2D printing and powered-based full-color 3D printing: (a) a colorization area in 2D printing; (b) a vertical section of (a); (c) a
colorization area in full-color 3D printing; (d) a vertical section in full-color 3D printing; (e) a horizontal section in full-color 3D printing.

Fig. 11. An example illustrating the accuracy challenge of color acquisition for color 3D prints: (a) a color 3D printed entity with a specific green brand logo; (b) a 2.5D
approximation of the area circled in red on the printed logo; (c) a conventional global photography in vertical direction of (b); (d) an area preserving projection of (b).

therefore possible to measure key color features with an area pre- arrangement. Full-color 3D printing is where the importance of
serving transformation based on photography in normal direction color is more emphasized in this workflow.
and match image points to points on the 3D printed entities, rather To begin with a 3D model with a simple texture, the raw mate-
than on some planes that are commonly used in the additive man- rial for the color is just roughly selected, rather than laboriously
ufacturing community. Furthermore, efficient segmentations into determined. For full-color 3D printing, 3D models with complex
regions with key features and their projections are essential to textures must be converted into the OBJ or VRML format. These
accurately capture curved surfaces. For example, Fig. 11 (c) gives two formats contain RGB color information for texture visualiza-
a photographic image where the colors match exactly but the posi- tion, which sometimes leads to color deviations or tone value com-
tions do not match one by one, while the map used in Fig. 11 (d) is pressions. Since the color range of a 3D model in the usual sRGB
area preserving and one to one, but the colors do not match color space is often larger than that of the coloring materials in
exactly. Combing two such maps to solve the twofold matching full-color 3D printing, an additional color management is essential
is a problem in itself. It should be explained that the Fig. 11 (c) is for accurate color reproduction. In 2D printing, the color manage-
a stitching of multiple vertical shots and therefore has no shadows. ment has been standardized using ICC profiles, which provides
the desired gamut mapping strategies and algorithms in the stan-
dard conversion module [107].
4. Outlook on an accurate color reproduction framework However, for accurate full-color 3D printing, color management
is still new. In Reference [98], Brunton et al. explore a halftoning
The workflow of 3D printing generally consists of three parts: framework for faithfully 3D printed color objects. Although
the 3D model input, the layer construction and the material designed for polymer-based color 3D printing with translucent

13
J. Yuan, G. Chen, H. Li et al. Materials & Design 209 (2021) 109943

Fig. 12. A color 3D printing workflow for halftoning framework. Examples used with permission from [98], copyright (2015) ACM.

printing materials, their framework can potentially be extended to 3D printing are cyan (C), magenta (M), yellow (Y), black (K), and
other material types with the following elements and optimiza- white (W), even some devices are available in spot colors such as
tions of the four core steps illustrated in Fig. 12: voxelization, color orange. In principle, there are only two options for color separa-
management, layer construction, and material arrangement. tion: halftoning (discrete color printing) and contoning (continu-
In the voxelization step: the color texture has to be mapped to ous color printing). Current halftoning techniques consist of
the surface voxels and slices with voxels of the same z-coordinates dithering, error diffusion, and iterative methods [118]. As
have to be generated from. Voxelization is the conversion of 3D explained in Section 2.4, the contoning method can provide more
model data into voxel data, either from a mesh model or a point faithful reproduction, depending on an accurate spectral function
cloud model [108,109]. The voxel data can also be divided into sur- match of each color layer. This is closely related to the computa-
face voxel and internal voxel depending on its position in the 3D tional assignment of the coloring materials (inks, binders or curing
model. The voxel implementation algorithms can be divided into resins) and is further discussed below with the material arrange-
two categories: CPU (central processing unit) method and GPU ment step.
(graphics processing unit) method [110,111]. Then, surface voxels For the layer construction step, most full-color 3D printing sys-
associated with RGBA values are to generate slices with specific tems are equipped with a dedicated slicing tool to better accom-
strategies. However, the thickness limitations of printing sub- modate the color representation of specific printing materials.
strates can also affect voxel unit output. For example, the thickness Ordinary 3D printing systems only need to consider the number
of paper sheets used in paper-based full-color 3D printing is nor- and thickness of the layers, but full-color 3D printing systems still
mally 104 mm, and the thickness of powder substrates in need to determine the boundary coloring width of each slice. How-
powder-based full-color 3D printing is generally 100 mm, which ever, the current color width of each layer is directly fixed in the
is far from the fine resolution used in plastic-based full-color 3D slicing software with a default value that cannot be modified by
printing. Further experiments are needed before color halftoning the user during the printing process. The boundary coloring width
can be extended to coarser resolutions. The approach taken in is easily understood and implemented for the plastic-based 3D
the color similarity perception analysis for customized 3D tablets printing techniques using the same physical state of coloring resins
can, for example, serve as a reference for such experiments [112]. and filling resins, such as stereo lithography (SLA), digital lithogra-
The color management step includes a gamut mapping and phy projection (DLP), continuous liquid interface production (CLIP),
color separation. Gamut mapping means to map color data to the continuous inkjet printing (CIJ) and drop-on-demand inkjet print-
printer’s tonal values based on the ICC profiles or iccMax profiles ing (DOD) [119]. However, beyond that, few systematic studies
[30,61,107]. Since there are more than 100 different 3D printing have been conducted to investigate the effect of the boundary col-
equipment types, the gamut mapping, which depends on the print- oring width on color reproduction accuracy in the layer construc-
ing material, is complex [113]. The ICC provides just four rendering tion. In fact, in our previous studies on accurate reproduction of
intentions, but there are many more color gamut mapping algo- 2.5D oil paintings, it was also found that the printing sequence of
rithms ranging from a pointwise gamut mapping to a spatial gamut color and non-color layers significantly affects the color reproduc-
mapping, as well as from small to large color gamuts [114–116]. tion accuracy by the reduction of nasty jagged effect [78,120].
However, all these traditional gamut mapping algorithms are not Recently, a unique 3D color test chart was purposefully designed
perfect for full-color 3D printing. As convolutional neural networks to examine the effect of colored layer features on color reproduc-
(CNN) have already been applied to the color gamut mapping of tion of plastic- and paper-based 3D printed parts and the results
ultra-high-definition television image [117], it is conceivable that also revealed a significant correlation based on image quality met-
machine learning could help to find suitable gamut mappings for rics from the HD orthographic imaging, but these didn’t show a lin-
full-color 3D printing systems. ear quantization trend [121,122]. To determine the slicing strategy,
Currently, despite the iccMax profile’s superiority in expressing the thickness, and number of sliced layers, it should be clearly
material appearance parameters, its encapsulated BRDF parametric understood how interior color tones influence the ones on the
module is not promoted by the ICC [61]. It is difficult to develop an outer surface. In future, an intelligent regulation of the boundary
advanced material appearance model because they need more data coloring width based on the adaptive feedback algorithms can be
space, lead to longer conversion times, and because the conversion embedded into non-polymer based full-color 3D printing systems.
of appearance and color parameters between different iccMax pro- Finally, the material arrangement step outputs slice files listing
files is not standardized. the specific components per voxel in a printer format. When the
Accordingly, color separation then means to map tonal values to boundary color and computational width of each layer is
the primary colors represented by printing materials (coloring determined, each digital color can be physically visualized by the
materials and filling materials). The primary colors for full-color assigned coloring and filling materials. The color separation

14
J. Yuan, G. Chen, H. Li et al. Materials & Design 209 (2021) 109943

determines the primary material choice, but specific implementa- Combining the boundary contoning theory and the surface coloring
tion methods depend on the boundary coloring width. For exam- efficiency prediction model can inspire interested scholars to
ple, Brunton et al. [98] and Babaei et al. [102] discussed two rethink and develop powerful coloring materials to precisely mod-
spectral function predictions for the boundary coloring width to ulate computational material assignment for a generic accurate
optimize the material arrangement within their halftoning and color reproduction framework. Besides, test benchmark design,
contoning framework, respectively. experimental guidelines and numerical analysis models for evalu-
In 3D halftoning, any layer is filled at discrete spots with the ating the color reproduction performance of full-color 3D printing
coloring material by stacking a mixture of primary color droplets are also key to implementing accurate material-aware full-color
with variable concentrations and ratios [123]. In this framework, 3D printing. Furthermore, the accurate full-color 3D printing pro-
the number of primary colors and inks concentrations is quickly cesses can also be extended to full-color 4D printing with precise
obtained. With contoning however, different tones are achieved color response, which is also a new research area for scholars in
by stacking colors with variable depths. This means that color the field of functional material design.
spots end up with different thicknesses. It is tricky to determine
a smallest sufficient number of spot categories based on the spec-
tral reflectance in normal direction of the color layer rather than in Funding
its building direction. If the 3D contoning is performed only in the
boundary coloring area of the printed layer, a boundary contoning This work has been financially supported by the Natural Science
framework can be formed to overcome the current material vox- Foundation of China (Grant No. 61973127), Guangdong Provincial
elization challenges of paper-based and powder-based full-color Science and Technology Program (Grant No. 2017B090901064),
3D printing. Besides, collotype printing [124] is a classical and Chaozhou Science and Technology Program (Grant No.
continuous-tone printing process for high-fidelity reproduction of 2020ZX14).
2D or 2.5D artworks. It requires extreme color separation skills
to determine the spot color categories for an authentic reproduc- Declaration of Competing Interest
tion of treasured Chinese brush paintings [125]. At present, these
subjective color separation skills have not been fully implemented The authors declare that they have no known competing finan-
by objective digital techniques. If these challenges can be mastered cial interests or personal relationships that could have appeared
together, the contoning framework will also be applicable to a pre- to influence the work reported in this paper.
cise color control for powder- and paper-based full-color 3D
printing.
In summary, the evolution from boundary coloring mode to References
boundary coloring theory is an important reference for the imple-
mentation and optimization of the current color accurate repro- [1] J.N. Fullerton, G.C. Frodsham, R.M. Day, 3D printing for the many, not the few,
Nat. Biotech. 32 (11) (2014) 1086–1087.
duction framework for full-color 3D printing, accompanied by [2] Wohlers Report 2017, http://www.wohlersassociates.com/2017report.htm.
the construction of the boundary coloring width computational ISBN: 978-0-9913332-3-3.
model. When the boundary coloring theory and the boundary col- [3] Amit Bandyopadhyay, Kellen D. Traxel, Invited Review Article: Metal-Additive
Manufacturing — Modeling Strategies for Application-Optimized Designs,
oring width model are standardized, the boundary contoning Addit. Manuf. 22 (2018) 758–774.
framework can be coded for the voxel arrangement of various [4] L. Safai, J.S. Cuellar, G. Smit, A.A. Zadpoor, A review of the fatigue behavior of
printing substrates. Moreover, this will fundamentally change the 3D printed polymers, Addit. Manuf. 28 (2019) 87–97.
[5] W. Elkhuizen, T. Essers, Y. Song, J. Geraedts, C. Weijkamp, J. Dik, S. Pont, Gloss,
current theories for accurate color reproduction and quality evalu- Color, and Topography Scanning for Reproducing a Painting’s Appearance
ation in full-color 3D printing. Furthermore, the development of Using 3D Printing [J], ACM J. Comput. Cult. He. 12 (4) (2019) 1–22.
surface coloring efficiency prediction models based on mainstream [6] L. Ren, Z. Song, H. Liu, et al., 3D printing of materials with spatially non-
linearly varying properties [J], Mater. Design 156 (OCT.) (2018) 470–479.
color 3D printers and existing printing materials can improve the [7] Y.L. Cheng, C.H. Chang, C. Kuo, Experimental study on leveling mechanism for
suitability of generic boundary coloring width models for accurate material-jetting-type color 3D printing, Rapid Prototyp. J. 26 (1) (2020) 11–
color reproduction in various full-color 3D printing processes. 20.
[8] 3D-SYSTEMS. 2014. ProJet 860Pro. http://www.3dsystems.com/3dprinters/
professional/projet-860pro (accessed May 2021).
[9] MCOR TECHNOLOGIES. 2014. MCOCR IRIS. http://MCORtechnologies.com/3d-
5. Conclusion printers/iris/ (accessed May 2021).
[10] STRATASYS. 2014. Objet500 Connex3. http://www.stratasys.com/3dprinters/
production-series/connex3-systems (accessed May 2021).
This review paper gives an overview over the four predominant [11] ISO/ASTM 52900, Additive manufacturing—general principles—requirements
color reproduction methods in Section 2 and a summary of the for purchased AM parts 2015.
state of the art in full-color 3D printing. Current studies on accu- [12] M.J. Tsai, C.W. Mei, Y.L. Cheng, et al., in: International Conference on Machine
Learning and Cybernetics (ICMLC), 2017, pp. 664–669.
rate color reproduction for color 3D printing focus mainly on MJ, [13] M.R. León, M. Umorin, M. Özcan, et al. Color dimensions of additive
BJ, SL and ME, all of which are also familiar to the materials & manufactured interim restorative dental material [J]. J. Prosthetic. Dent, 2019.
design community. However, coloring, rendering, and acquisition [14] L. Zhou, J. Fu, Y. He, A review of 3D printing technologies for soft polymer
materials, Adv. Funct. Mater. 2000187 (1–38) (2020), https://doi.org/10.1002/
issues are not comprehensively investigated and are challenging adfm.202000187.
problems in developing a general color accuracy evaluation system [15] T. Tran, On the Bottleneck of Adopting 3D Printing in Manufacturing, Virtual
for full-color 3D printing. and Physical Prototyping 12 (4) (2017) 333–334.
[16] T. Kuipers, W. Elkhuizen, J. Verlinden, et al., Hatching for 3D prints: Line-
Color accuracy problems and potential solutions are discussed based halftoning for dual extrusion fused deposition modeling[J], Comput.
with vivid illustrations and easy-to-understand subtitles. Graph-UK 74 (2018) 23–32.
Computer-aided coloring methods show the greatest potential for [17] X.C. Wang, C. Chen, J.P. Yuan, G.X. Chen, Color Reproduction Accuracy
Promotion of 3D-Printed Surfaces Based on Microscopic Image Analysis, Int J
achieving accurate full-color 3D printing with many types of print-
Pattern. Recogn. 34 (01) (2020) 5593–5600.
ing materials. Based on the halftoning and contoning techniques to [18] H.C. Wu, T.C.T. Chen, Quality control issues in 3D-printing manufacturing: a
generate voxel units in plastic-based full-color 3D printing, a com- review[J], Rapid Prototyp. J. 24 (3) (2018) 607–614.
prehensive outlook on the implementation of an accurate color [19] M. Amini, S. Chang, Process monitoring of 3D metal printing in industrial
scale[C]//ASME 2018 13th International Manufacturing Science and
reproduction framework with four core steps is provided and Engineering Conference, American Society of Mechanical Engineers Digital
further extended to other physical states of printing materials. Collection (2018).

15
J. Yuan, G. Chen, H. Li et al. Materials & Design 209 (2021) 109943

[20] J.Y. Lee, A.P. Nagalingam, S.H. Yeo, A Review on the State-of-the-art of Surface [52] D. Sumin, T. Rittig, V. Babaei, T. Nindel, A. Wilkie, P. Didyk, B. Bickel, J.
Finishing Processes and Related ISO/ASTM Standards for Metal Additive Křivánek, K. Myszkowski, T. Weyrich, Geometry-Aware Scattering
Manufactured Components, Virtual and Physical Prototyping (2020) 1–29. Compensation for 3D Printing. ACM Trans. Graph. 38, 4, Article 111 (July
[21] G.X. Chen, C. Chen, Z.H. Yu, H. Yin, L.X. He, J.P. Yuan, Color 3D Printing: 2019), 14 pages.
Theory, Method, and Application. New Trends in 3D Printing, IntechOpen [53] B.T. Phong, Illumination for computer generated pictures, Commun. ACM 18
(2016) d), https://doi.org/10.5772/63944. (6) (1975) 311–317.
[22] Intechopen company (2016). https://www.intechopen.com/ books/new- [54] G.J. Ward, Measuring and modeling anisotropic reflection, ACM Computer
trends-in-3d-printing (accessed May 2021). Graphics (ACM) 26 (2) (1992) 265–272.
[23] ICC, Taipei, 2016. ‘‘Color Management in Displays and 3D Printing”. http:// [55] S.R. Marschner, S.H. Westin, E.P.F. Lafortune, K.E. Torrance, D.P. Greenberg,
www.color.org/events/taipei/1-ICC_overview.pdf (accessed May 2021). Image-based BRDF measurement including human skin. In: Lischinski D.,
[24] I. Tastl, M. A. Lopez-Alvarez, A. Ju, et al. A Soft-proofing Workflow for Color 3D Larson G.W. (eds) Rendering Techniques’ 99. Eurographics. Springer, Vienna.
Printing - Addressing Needs for the Future[J]. J. Electron. Imaging, 2019, 2019 (1999).
(6):479-1-479-6. [56] D.B. Goldman, B. Curless, A. Hertzmann, S.M. Seitz, in: Shape and spatially-
[25] J.P. Yuan, C. Chen, J.N. Tian, D.Y. Yao, H. Li, G.X. Chen, Color Reproduction varying BRDFs from photometric stereo, IEEE, Piscataway, NJ, 2005, pp. 341–
Evaluation of Paper-Based Full-Color 3D Printing Based on Image Quality 348.
Metrics, Digital Printing 5 (2020) 26–34. [57] O. Rouiller, B. Bickel, J. Kautz, W. Matusik and M. Alexa, 3D-Printing Spatially
[26] G.X. Chen, X.C. Wang, H.Z. Chen, C. Chen, Realization of Rapid Large-Size 3D Varying BRDFs. IEEE Comput. Graph & Appl. 33(6) (2013) 48-57. Nov.-Dec.
Printing Based on Full-Color Powder-Based 3DP Technique, Molecules 25 2013, doi: 10.1109/MCG.2013.82.
(2020) 2037. [58] M.M. Bagher, C. Soler, N. Holzschuch, Accurate fitting of measured reflectance
[27] J.P. Yuan, M. Zhu, B.H. Xu, G.X. Chen, Review on processes and color quality using a Shifted Gamma micro-facet distribution. Computer Graphics Forum.
evaluation of color 3D printing, Rapid Prototyp. J. 24 (2) (2018) 409–415. Oxford, UK: Blackwell Publishing Ltd, 31(4) (2012) 1509-1518.
[28] H.R. Kang, 1993, January. Comparisons of color mixing theories for use in [59] L.N. Song, Research on Appearance Reproduction of Object Surface Based on
electronic printing. In Color and Imaging Conference, Society for Imaging Bidirectional Reflectance Distribution Function, Electro-Optic Technol. Appl.
Science and Technology, 1993, (1), 78-82. 28 (4) (2013) 10–15.
[29] X. Lin, M. Wiertlewski, Sensing the frictional state of a robotic skin via [60] L. Simonot, G. Obein, B. Bringier, D. Meneveaux, Modeling, measuring, and
subtractive color mixing, IEEE Rob. Autom. Lett. 4 (3) (2019) 2386–2392. using BRDFs: significant French contributions, JOSA-A 36 (11) (2019) 40–50.
[30] ICC. 2010. File Format for Color Profiles, 4.3.0.0 edition. http://www.color.org. [61] V. James, 3D appearance management using iccMax, ICC Display & 3D Print
(accessed May 2021). Meeting proceedings, 2016.
[31] G.X. Chen, Advances in 3D Printing Research / 2016–2017 Sensorimaging [62] 3MF Consortium. 3D Manufacturing Format Specification and Reference
Disciplinary Development Report / CCSA Disciplinary Development Research Guide, 1 (2015) 1-48.
Series [M], China Science and Technology Press, Beijing, 2018, pp. 283–302. [63] B.M. Vu, P. Urban, T.M. Tanksale, S. Nakauchi, Visual perception of 3D printed
[32] CIE, CIE Tutorial on Colorimetry and Visual Appearance http://cie.co.at/ translucent objects, Color and Imaging Conference. Soci. for Imaging. Sci. &
news/cie-tutorials-colorimetry-and-visual-appearance (2017). (accessed Technol. 1 (2016) 94–99.
May 2021). [64] W.S. Elkhuizen, T.T. Essers, Y. Song, S.C. Pont, J.M. Geraedts, J. Dik, Gloss
[33] R. Cheydleur, Measurement For 3D Printing. ICC Display & 3D Print Meeting Calibration and Gloss Gamut Mapping for Material Appearance Reproduction
proceedings, Taipei (2016). of Paintings, EUROGRAPHICS Workshop on Graphics and Cultural Heritage
[34] Proc. SPIE 9398, Measuring, Modeling, and Reproducing Material Appearance (2018) 161–164.
2015, 939801 (15 April 2015), Soci. for Imaging. Sci. & Technol.; doi: 10.1117/ [65] S. Stopp, T. Wolff, F. Irlinger, T. Lueth, A new method for printer calibration
12.2193203. and contour accuracy manufacturing with 3D print technology, Rapid
[35] Appearance Printing European Advanced Research School, 2017. http://www. Prototyp. J. 14 (3) (2008) 167–172.
appears-itn.eu/ (accessed May 2021). [66] M. Page, G. Obein, C. Boust, A. Razet, Adapted Modulation Transfer Function
[36] H. Kim, Y. Lin, T.L.B. Tseng, A review on quality control in additive Method for Characterization and Improvement of 3D printed surfaces, Elect.
manufacturing, Rapid Prototyp. J. 24 (3) (2018) 645–669. Imaging. 8 (2017) 92–100.
[37] S. Tominaga, R. Schettini, A. Trémeau, T. Horiuchi, Computational Color [67] M. Stanić, B. Lozo, T. Muck, S. Jamnicki, R. Kulčar, Color measurements of three-
Imaging, 7th International Workshop, CCIW 2019, Chiba, Japan, March 27-29, dimensional ink-jet prints, Proceedings of NIP, Pittsburgh 24 (2008) 6–11.
2019, https://link.springer.com/book/10.1007/978-3-030-13940-7 (accessed [68] M. Stanic, B. Lozo, D.G. Svetec, Colorimetric properties and stability of 3D
May 2021). prints, Rapid Prototyp. J. 18 (2) (2012) 120–128.
[38] H.C. Hottel, A.F. Sarofim, Radiative transfer, Landolt-Börnstein -Group VI [69] K.D. Xiao, F. Zardawi, R. Noort, J.M. Yates, Color reproduction for advanced
Astronomy and Astrophysics 142 (9) (1967) 42–44. manufacture of soft tissue prostheses, J Dent. 41 (2013) 15–23.
[39] R.E Hummel. Optical Properties of Materials. In: Understanding Materials [70] K.D. Xiao, F. Zardawi, R. Noort, J.M. Yates, Developing a 3D Colour
Science. Springer, New York, USA, 2004. https://doi.org/10.1007/0-387- Reproduction System for Additive Manufacturing of Facial Prostheses, Color
26691-7_13. and Imaging Conference. Soci. for Imaging. Sci. & Technol. 1 (2013) 206–209.
[40] J. Wang, S. Zhao, X. Tong, et al., Modeling and Rendering of Heterogeneous [71] K.D. Xiao, F. Zardawi, R. Noort, J.M. Yates, Developing a 3D color image
Translucent Materials Using the Diffusion Equation, ACM Trans. Graph. 27 (1) reproduction system for additive manufacturing of facial prostheses, Int J of
(2008) 1–18. Adv. Manuf. Technol. 70 (2014) 2043–2049.
[41] D.R., Ney and E.K. Fishman, ‘‘Volumetric Rendering.” IEEE Computer Graphics [72] A. Sohaib, K. Amano, K.D. Xiao, J.M. Yates, C. Whitford, S. Wuerger, Colour
& Applications 10.2(1990):24-32. quality of facial prostheses in additive manufacturing, Int J of Adv. Manuf.
[42] A. Genty-Vincent, T.P.V. Song, C. Andraud, M. Menu, Four-flux model of the Technol. 96 (1–4) (2018) 881–894.
light scattering in porous varnish and paint layers: towards understanding [73] L.X. He, G.X. Chen, Effect of Aqueous Adhesives on Color of Paper Surface in
the visual appearance of altered blanched easel oil paintings, Appl. Phys. A Paper-based 3D Printing, Pack. Eng. 37 (3) (2016) 153–156.
123 (7) (2017) 473. [74] J.P. Yuan, G.X. Chen, Z.H. Yu, Speedup Method of paper-based 3D Color
[43] T.P.V. Song, C. Andraud, M.V.O. Segovia, Towards Spectral Prediction of 2.5D printing Based on Slicing strategies, Pack. J. 8 (1) (2016) 53–58.
Prints for Soft-Proofing Applications, International Conference on Image [75] J.P. Yuan, G.X. Chen, J.H. Liao, Z.H. Yu, Visualization of Large-Size Model Based
Processing Theory Tools & Applications, 2017. on Paper-Based 3D Printing, Advanced Graphic Communications, Pack. Eng. &
[44] T.P.V. Song, C. Andraud, M.V.O. Segovia, Spectral predictions of rough ink layers Mater. 369 (2016) 333–338.
using a four-flux model, Twenty Fifth Color and Imaging Conference (2017). [76] J.P. Yuan, X.Y. Yan, X.C. Wang, G.X. Chen, Paper-based 3D printing
[45] T.P.V. Song, C. Andraud, L.R. Sapaico, M.V.O. Segovia, Color prediction based industrialization for customized wine packaging applications, NIP & Digit.
on individual characterizations of ink layers and print support, Conference: Fabric. Conf. 1 (2017) 118–121.
IS&T International Symposium on Electronic Imaging 2018, Material [77] X.Y. Yan, J.P. Yuan, G.X. Chen, Applications Analysis of Paper-Based Color 3D
Appearance, 2018. Printing in the Map Industry. In: Zhao P., Ouyang Y., Xu M., Yang L., Ren Y.
[46] T.P.V. Song. Optical models for appearance management of printed materials. (eds) Applied Sciences in Graphic Communication and Packaging. Lecture
Centre de Recherche et Restauration des Musées de France. PhD thesis, 2018. Notes in Electrical Engineering, 477 (2018) 377-383. Springer, Singapore.
[47] T.F. Schiff, J.C. Stover, D.R. Cheever, D. R. Bjork, Maximum and minimum https://doi.org/10.1007/978-981-10-7629-9_46.
limitations imposed on BSDF measurements. Proceedings of SPIE - The [78] C. Chen, 3D Digitalization and 3D Printing Reproduction Method for Oil
International Society for Optical Engineering, 1989, 66. Paintings, South China University of Technology, PhD Report, 2016.
[48] C. Asmail, Bidirectional scattering distribution function (BSDF): a [79] L.X. He, Study on Paper-based 3D Printing Based on DEM Data, South China
systematized bibliography, J. Res. Nat. Inst. Stand. Technol. 96 (2) (1991). University of Technology, Master Report, 2016.
[49] G.V.G. Baranoski, A. Krishnaswamy, ‘‘Light, Optics, and Appearance”, in Light [80] J.P. Yuan, L. Cai, X.C. Wang, G.X. Chen, Visualization of Biomedical Products
& Skin, Interactions (2010). based on Paper-based Color 3D Printing, NIP & Digit. Fabric. Conf. 1 (2019)
[50] E. Eon, G.A. Irving, Quantized-Diffusion Model for Rendering Translucent 128–131.
Materials. ACM Trans. Graph., 30, 4, Article 56 (July 2011), 13 pages. DOI: [81] P.L. Sun, Y. Sie, Color dithering methods for LEGO-like 3D printing. Color
10.1145/1964921.1964951. Imaging XX: Displaying, Processing, Hardcopy, and Applications. Int Society
[51] O. Elek, D. Sumin, R. Zhang, T. Weyrich, K. Myszkowski, B. Bickel, A. Wilkie, J. for Opt. and Photo. 93950J (2015) 9395: 9399.
Křivánek. Scattering-aware Texture Reproduction for 3D Printing. ACM Trans. [82] P.L. Sun, Y.P. Sie, Color uniformity improvement for an inkjet color 3D
Graph. 36, 6, Article 241 (November 2017), 15 pages. printing system, Elect. Imaging 20 (2016) 1–6.

16
J. Yuan, G. Chen, H. Li et al. Materials & Design 209 (2021) 109943

[83] Y.P. Pi, P.L. Sun, Color gamut mapping for 3D printing, ICC Display & 3D Print [105] A. Nadir, Effect of Layer Thickness on Surface Properties of 3D Printed
Meeting proceedings, 2016. Materials Produced from Wood Flour/PLA Filament, Polym. Test. 71 (2018)
[84] Z.Y. Li, H. Min, G.H. Cui, H.X. Liu, Evaluation of the Color-difference Formulae 163–166.
for Neutral Colors. Advanced Graphic Communications, Pack. Tech. & Mater. [106] P.C. Zhang, T.J. Arre, A. Ide-Ektessabi, A line scan camera-based structure from
(2016). motion for high-resolution 3D reconstruction, J. Cult. Herit. 16 (5) (2015)
[85] CIE TC 8-17, Methods for Evaluating Color Difference between 3D Color 656–663.
Objects, http://www.cie.co.at/technical-work/jtcs. (2017). (accessed May [107] A. Sharma, Understanding color management, John Wiley & Sons, 2018.
2021). [108] F. Liu, Voxelization processing method and device based on triangular mesh
[86] C.A. Wei, X.X. Wan, J.F. Li, A modified single-constant Kubelka-Munk model model [P], 2018.
for color prediction of pre-colored fiber blends, Cellulose 25 (6) (2018) 1–12. [109] Y.S. Xu, T. Sebastian, L. Hoegner, et al., Voxel-based segmentation of 3D point
[87] D. Michael, Elevated printing: application use cases, ICC Display & 3D Print clouds from construction sites using a probabilistic connectivity model,
Meeting proceedings, 2016. Pattern Recogn. Lett. 102 (2018) 67–74.
[88] C. Parraman, M.V.O. Segovia, 2.5 D Printing: Bridging the gap between, 2D [110] D.J., Liu, Y. Yang, X.Y. Wang, & H.Z. Chen. Model voxelization efficiency
and 3D applications., John Wiley & Sons, 2018. optimization method and system based on CPU operation [P], 2020.
[89] B. Christoph, K. Dominik, J.C. Weaver, O. Neri, Data-Driven Material Modeling [111] G. Young, A. Krishnamurthy, GPU-accelerated generation and rendering of
with Functional Advection for 3D Printing of Materially Heterogeneous multi-level voxel representations of solid models. Comput. Graph-UK, 2018,
Objects, 3D Print. Addit. Manuf. 3 (2) (2016) 71–79. 75(OCT.):11-24.
[90] Y.B. Dong, P.J. Li, Research on the modification of water transfer PVA sheet [112] J.P. Yuan, H. Li, B.H. Xu, G.X. Chen, Impact of Geometric Features on Color
after adding starch, Appl. Chem. Industry 36 (4) (2007) 376–379. Similarity Perception of Displayed 3D Tablets, J. Imaging Sci. Technol. 64 (5)
[91] Y.Z. Zhang, C.J. Yin, C.X. Zheng, K. Zhou, Computational hydrographic printing, (2020) 1–12.
ACM Trans. Graph. 34 (4) (2015) 13–24. [113] P. Morovic, J. Morovic. Color Mapping in 3D Printing [P], 2019.
[92] C. Schuller, D. Panozzo, A. Grundhofer, H. Zimmer, E. Sorkine, S.H. Olga, [114] M.K. Cho, H.K. Choh, S.E. Kim, Y.T. Kim, Y. Bang, Gamut mapping method for
Computational thermoforming, ACM Trans. Graph. 35 (4) (2016) 43–52. ICC saturated intent. In Color Imaging XII: Processing, Hardcopy, and
[93] A. Muntoni, M. Livesu, R. Scateni, A. Sheffer, D. Panozzo, Axis-Aligned Height- Applications, International Society for Optics and Photonics. 6493 (2007)
Field Block Decomposition of 3D Shapes, ACM Trans. Graph. 1 (1) (2018) 1– 64930K.
15. [115] J.P. Yuan, J.Y. Hardeberg, G.X Chen, Development and evaluation of a hybrid
[94] J.P. Yuan, Z.H. Yu, G.X. Chen, M. Zhu, Y.F. Gao, Large-size color models point-wise gamut mapping framework. In Proceedings of the Colour and
visualization under 3D paper-based printing, Rapid Prototyp. J. 23 (5) (2017) Visual Computing Symposium, Gjovik, Norway, IEEE: Gjovik, Norway,
911–918. 2015:1–4.
[95] J.P. Yuan, Y.M. Liu, G.X. Chen, L.X. He, Process Analysis of Seamless Adhesion [116] L. Xu, B. Zhao, M.R. Luo, Colour gamut mapping between small and large
for Cutting Model Printed by Color Paper-Based 3D Printing. Advanced colour gamuts: Part I. gamut compression, Opt. Express 26 (9) (2018) 11481–
Graphic Communications and Media Technologies. Lecture Notes, Electr. Eng. 11495.
417 (2017) 483–488. [117] M. Takeuchi, Y. Sakamoto, R. Yokoyama, S.U.N. Heming, Y. Matsuo, J. Katto, A
[96] M.H. Yang, X.G. Lv, X.J. Liu, J.Q. Zhang, Research on color 3D printing based on Gamut-Extension Method Considering Color Information Restoration using
color adherence, Rapid Prototyp. J. 24 (1) (2018) 37–45. Convolutional Neural Networks, in: In 2019 IEEE International Conference on
[97] M.A. Skylar-Scott, J. Mueller, C.W. Visser, J.A. Lewis, Voxelated soft matter via Image Processing (ICIP) IEEE, 2019, pp. 774–778.
multimaterial multinozzle 3D printing, Nature 575 (7782) (2019) 330–335. [118] D.L. Lau, G. Arce, Modern Digital Halftoning. (2001).
[98] A. Brunton, C.A. Arikan, P. Urban, Pushing the Limits of 3D Color Printing: [119] X. Wang, M. Jiang, Z.W. Zhou, J.H. Gou, D. Hui, 3D printing of polymer matrix
Error Diffusion with Translucent Materials, ACM Trans. Graph. 35 (1) (2015) composites: A review and prospective, Compos. Part B-Eng 110 (2017) 442–
1–13. 458.
[99] B. Alan, A.A. Can, M.T. Tejas, P. Urban, 3D printing spatially varying color and [120] J.P. Yuan, C. Chen, D.Y. Yao, G.X. Chen, 3D Printing of Oil Paintings Based on
translucency, ACM Trans. Graph. 37 (4) (2018) 157–170, https://doi.org/ Material Jetting and Its Reduction of Staircase Effect, Polymers 12 (2020)
10.1145/3197517.3201349. 2536.
[100] P. Urban, T.M. Tanksale, A. Brunton, B.M. Vu, S. Nakauchi, Redefining A in [121] J.P. Yuan, J.N. Tian, C. Chen, G.X. Chen, Experimental Investigation of Color
RGBA: Towards a Standard for Graphical 3D Printing, ACM Trans. Graph. 1 (1) Reproduction Quality of Color 3D Printing Based on Colored Layer Features,
(2019) 1–15. Molecules 25 (2020) 2909.
[101] Stratasys, GrabCAD, https://www.stratasys.com/software, (2017). (accessed [122] J.N. Tian, J.P. Yuan, H. Li, D.Y. Yao, G.X. Chen, Advanced Surface Color Quality
May 2021). Assessment in Paper-based Full-color 3D printing, Materials 14 (4) (2021)
[102] V. Babaei, K. Vidimče, M. Foshey, A. Kaspar, P. Didyk, W. Matusik, Color 736.
Contoning for 3D Printing, ACM Trans. Graph. 36 (4) (2017) 124–139. [123] S. Gooran, F. Abedini, 3D Surface Structures and 3D Halftoning. Society for
[103] B. Christoph, K. Dominik, C.W. James, S. Sunanda, A. Hosny, J. Costa, N. Imaging Science and Technology, 2020. Printing for Fabrication 2020, 2020.
Oxman, Making data matter: Voxel printing for the digital fabrication of data DOI: https://doi.org/10.2352/ISSN.2169-4451.2020.36.75.
across scales and domains, Sci. Adv. 4 (5) (2018) eaas8652. [124] D. Defibaugh. 1997. The Collotype: History, process & photographic
[104] O. Norberg, M. Andersson, B. Kruse, The influence of paper properties on color documentation. Rochester Institute of Technology (1997).
reproduction and color management, NIP & Digit. Fabric. Conf. 2 (2003) 836– [125] F. Tang, W. Dong, Y. Meng, X. Mei, F. Huang, X. Zhang, O. Deussen, Animated
840. Construction of Chinese Brush Paintings, IEEE T. Vis. Comput. Gr. 24 (12)
(2018) 3019–3031.

17