**5. Colour management**

Due to the different ways in which colour communication between 3D image capture and 3D image printing is utilised, there are often significant discrepancies in colour between the printed and original objects when using this type of manufacturing process. Conventional colour image reproduction techniques based on CIE colorimetry have been used for more than 80 years and perform very well in transforming colour images from one digital media to another under various viewing conditions [25]. However, the application of conventional colour image reproduction techniques for 3D printing technology is not as straightforward. With this protocol, 3D colour objects can be split into a monochromatic 3D objects and a 2D colour images. With this in mind, a conventional colour reproduction technique can be applied to transform the 2D colour image from camera RGB to the corresponding printer RGB through human colour appearance attributes. To achieve an accurate colour reproduction through 3D imaging devices, specific colour profiles need to be developed in order to link the colour system of a specific device to the human visual system [26].

#### **5.1. Development of colour profiles for 3D image devices**

used to identify the position, angulation and orientation of any attachments that need to be incorporated into the data incorporated in to the process. Their position can then be translated

Natural orifices such as nasal apertures and auditory canals must be designed to ensure that the prosthetic has correct anatomical form and be modelled in a way that will allow for appropriate function if needed. For the nose, it will allow for the passage of air during respiration whilst for the ear, although external form may be missing, it will allow non‐ impeded hearing via direct access to the external auditory meatus (canal). Similar considera‐ tions can be made when designing orbital prostheses as the housing of an artificial glass or acrylic globe will be required within the palpebral fissure (eye lids). In some instances, it should be remembered that the provision of attachments may impinge on these design features and

through the design and manufacturing process into the final prosthesis (**Figure 4**).

so optimal modelling and detailing must be undertaken to camouflage these issues.

calibration image is then also applied to the whole model prior to production.

Once the final model has been produced, additional details need to be added in order to produce a satisfactory prosthesis. Final details including the incorporation of peripheral extensions, feathered edges, any necessary strengthening buttresses and the smoothing of seams and joins on the model's inner surface can all be undertaken prior to printing. The colour

Due to the different ways in which colour communication between 3D image capture and 3D image printing is utilised, there are often significant discrepancies in colour between the printed and original objects when using this type of manufacturing process. Conventional colour image reproduction techniques based on CIE colorimetry have been used for more than 80 years and perform very well in transforming colour images from one digital media to another under various viewing conditions [25]. However, the application of conventional colour image reproduction techniques for 3D printing technology is not as straightforward. With this protocol, 3D colour objects can be split into a monochromatic 3D objects and a 2D colour images. With this in mind, a conventional colour reproduction technique can be applied to transform the 2D colour image from camera RGB to the corresponding printer RGB through human colour appearance attributes. To achieve an accurate colour reproduction through 3D imaging devices, specific colour profiles need to be developed in order to link the colour system

**4.4. Data detailing**

96 New Trends in 3D Printing

**4.5. Design finalisation**

**5. Colour management**

of a specific device to the human visual system [26].

To truly reproduce colour from a 3D camera to 3D printer, both a camera colour profile and a printer colour profile need to be developed to connect camera RGB and printer RGB with human eye response (CIE XYZ tristimulus values [27]). Variations in skin colour and tones are considerable [28], and to try and incorporate the whole spectrum of skin colours available for different human populations templates such as the digital Macbeth ColorCheckerDC charts (X-Rite Inc., Grand Rapids, MI, USA) can be used to provide numerous training colours (see **Figure 5a**). Utilising this method, a 2D colour chart can be converted to a 3D model with dimensions of 200 (l) × 150 (w) × 3 (h) with the colours desired and printed using a 3D printer such as a Z Corp Z510 colour printer. These can then be used as your testing colours (**Fig‐ ure 5b**). Subsequently, the CIE XYZ tristimulus values for each training colour produced in the printed colour chart can be obtained by taking colour measurements using a spectropho‐ tometer such as a Minolta CM‐2600d (Konica Minolta Inc., Tokyo, Japan). During measure‐ ments, standard viewing conditions should be applied such as a viewing geometry of d/8 (diffuse illumination, 8‐degree viewing), with specular component included and aperture sizes consistently set to a defined diameter—3 mm. The illuminant must also be consistent and ideally of an industry standard including setting to CIE standard D65 to simulate skin colour in daylight conditions and CIE 1931 standard observer [27].

**Figure 5.** 3D models of a) training and b) testing chart.

Based on printer RGB and CIE XYZ tristimulus values for the training colours, a printer colour profile can be developed using a third‐order polynomial regression model [28]. A camera colour profile can also be produced utilising the same chart. Colour images are captured by the 3dMD camera system and the camera RGB for the number of training colours used identified. Then, based on these camera RGB and CIE XYZ tristimulus values, a camera colour profile can be developed using a second‐order polynomial regression [29]. Using this method, for each pixel of image, camera RGB is first transformed to CIE XYZ tristimulus values and then transform back into printer RGB.

## **5.2. Evaluation of colour reproduction**

To evaluate colour reproduction for the human face, a colour test chart first needs to be designed. This can be undertaken using 14 predetermined human skin colours, including four Caucasian, two Chinese, two Asian, four African and two Caribbean skin shades (see **Figure 5b**) [30]. A 3D colour chart must then be generated using the colour printer to be used (Z Corp Z510), with defined and consistent dimensions. After post-processing in the same way any final prosthesis would undergo, this chart is referred to as the original colour chart.

Colour reproduction can then be evaluated using two reproduction charts that must be produced using two different 3D colour image reproduction systems. When the first chart is produced, the colour image for the original chart must be captured using the 3dMD camera system and then processed with only minor corrections in 3D geometry before it is sent to the Z Corp Z510 printer for 3D printing. This printed colour chart can then be referred to as the first reproduction (Chart 1). For the second processing, it can be conducted following the proposed 3D colour image reproduction process. This printed chart can then be referred to as the second reproduction chart (Chart 2).

To assess the performance of the colour reproduction, CIE XYZ tristimulus values for each colour patch within each colour chart must be measured using a spectrophotometer (Konica Minolta cm‐2600d). Colour difference between the original chart and each of the two repro‐ duction charts for each of the 14 testing colours under CIE illuminant D65 should be calculated using a CIELAB colour difference formula [31]. The mean, maximum, minimum and standard deviation for the colour difference can be recorded and tabulated. Examples of these values can be seen in **Table 1**. If undertaken correctly, this method will demonstrate that a significant improvement in colour reproduction can be achieved using 3D colour image reproduction systems. For its successful application in the production of facial soft tissue prostheses, an acceptable colour difference for the 3D printed objects is approximately 3–4 ΔE\*ab [32].


**Table 1.** Examples of colour difference between original (Org) colour chart and reproduction (Rep) colour charts.

## **6. Colour texture mapping**

Many 3D photogrammetry systems can give the illusion of texture by wrapping the 2D image over the 3D surface. This wrapped texture does not actually produce fine wrinkles, pores or blemishes on the surface of the skin and hence cannot be reproduced within the rapid prototyping process.

for each pixel of image, camera RGB is first transformed to CIE XYZ tristimulus values and

To evaluate colour reproduction for the human face, a colour test chart first needs to be designed. This can be undertaken using 14 predetermined human skin colours, including four Caucasian, two Chinese, two Asian, four African and two Caribbean skin shades (see **Figure 5b**) [30]. A 3D colour chart must then be generated using the colour printer to be used (Z Corp Z510), with defined and consistent dimensions. After post-processing in the same way any final prosthesis would undergo, this chart is referred to as the original colour chart.

Colour reproduction can then be evaluated using two reproduction charts that must be produced using two different 3D colour image reproduction systems. When the first chart is produced, the colour image for the original chart must be captured using the 3dMD camera system and then processed with only minor corrections in 3D geometry before it is sent to the Z Corp Z510 printer for 3D printing. This printed colour chart can then be referred to as the first reproduction (Chart 1). For the second processing, it can be conducted following the proposed 3D colour image reproduction process. This printed chart can then be referred to as

To assess the performance of the colour reproduction, CIE XYZ tristimulus values for each colour patch within each colour chart must be measured using a spectrophotometer (Konica Minolta cm‐2600d). Colour difference between the original chart and each of the two repro‐ duction charts for each of the 14 testing colours under CIE illuminant D65 should be calculated using a CIELAB colour difference formula [31]. The mean, maximum, minimum and standard deviation for the colour difference can be recorded and tabulated. Examples of these values can be seen in **Table 1**. If undertaken correctly, this method will demonstrate that a significant improvement in colour reproduction can be achieved using 3D colour image reproduction systems. For its successful application in the production of facial soft tissue prostheses, an acceptable colour difference for the 3D printed objects is approximately 3–4 ΔE\*ab [32].

**CIE ΔE\*ab Mean Max Min SD Org vs. Rep1** 20.8 27.8 8.0 5.5 **Org vs. Rep2** 4.5 11.1 2.5 2.3

**Table 1.** Examples of colour difference between original (Org) colour chart and reproduction (Rep) colour charts.

Many 3D photogrammetry systems can give the illusion of texture by wrapping the 2D image over the 3D surface. This wrapped texture does not actually produce fine wrinkles, pores or

then transform back into printer RGB.

98 New Trends in 3D Printing

the second reproduction chart (Chart 2).

**6. Colour texture mapping**

**5.2. Evaluation of colour reproduction**

**Figure 6a** demonstrates a side‐by‐side view of a 3D polygon mesh and the 2D bitmap image captured by a 3D camera system (3dMD System). It can be seen from the image that the 3D mesh does not actually contain fine details including pores and wrinkles, which are clearly visible in the 2D bitmap. Each polygon in the 3D mesh is linked to a particular region in the 2D bitmap. **Figure 6b** shows highlighted polygons in the 3D mesh and the region used from the 2D bitmap to overlay colour onto a monochromatic polygon.

**Figure 6.** (a, b). Images demonstrating a 3D polygon mesh and the 2D bitmap image captured by a 3D camera system.

To enhance realism and improve the characterisation of the patient's skin, various techniques have been developed. One such method includes incorporating surface details such as pores, wrinkles and fine lines into the 3D model. Height‐field mapping (also known as bump mapping) within the CAD design process can be used to translate the texture reference image onto a geometric pattern. Height‐field mapping is a method of translating fine 2D images onto 3D geometric virtual models. This is based on white (high) and black (low) greyscale texture reference mapping as detailed above, and can be applied to map representative texturing to the appropriate areas of the facial model. These techniques can not only use patient specific data from adjacent facial anatomy/topography but also, using pre‐treatment 2D photographs of the same area. The latter allows for manually adding realistic skin textures to the prosthetic model surface in cases where texture from other parts of the missing face are either inappro‐ priate or not consistent or representative of the area to be replaced. Based on the grey level intensity (from white to black), computer software can control the depth of imperfections over a skin surface. An original skin depth map is shown in **Figure 7 (a)** and can be used to add texture over a flat surface. Surface topography changes according to the grey level intensity and the positioning of individual pores and wrinkles is shown in (b). The resultant mesh can be 3D printed, producing a skin‐like texture over a flat surface (c).

**Figure 7.** (a) Depth map of skin showing pores and fine wrinkles as dark areas. (b) Software‐induced texture mapping over a flat surface using the depth map from (a). (c) 3D printed surface with pores and wrinkles clearly visible on the surface (Approx. part Dim: 38 × 38 × 3 mm).

Given the flexibility of such software, texture can be mapped not only onto flat surfaces but also over complex 3D shapes. An example of this texture mapping over a nose model with varying pore depths is show in **Figure 8**.

**Figure 8.** Illustrations detailing texture mapping. (Approx. part Dim: 70 × 61 × 29 mm).

However, the detail provided in the final prosthesis is dependent not only on the resolution of 3D data obtained but also the resolution of the 3D printer and characteristics of the powder and binder type used within the process. The use of course powder in the printing process will reduce the detail of the texture derived from the height‐field mapping even though such detail can be mapped within the CAD process. In contrast, using finer powder will enable the addition of very fine details over the printed prosthesis.
