Colour Quality Evaluation of Bluish-Green Serpentinite Based on the CIECAM16 Model

: The CIECAM16 colour appearance model is currently a model with high prediction accuracy. It can solve the problem of predicting the inﬂuence of different observation conditions on the colour of gemstones. In this study, a computer vision system (CVS) was used to measure the colour of 59 bluish-green serpentinite samples, and the tristimulus values were input into the CIECAM16 forward model to calculate the colour appearance parameters of serpentinite under different surrounds, illuminances, and light sources. It was found that the darkening of the surround causes the lightness and brightness to increase. Pearson’s r of brightness and colourfulness with illuminance is 0.885 and 0.332, respectively, which predicts the Stevens and Hunt effects. When the light source changes from D65 to A, the calculated hue angle shifts to the complementary area of the A light source, which is contrary to the CVS measurement result. The D65 light source is more suitable for the colour presentation and classiﬁcation of bluish-green serpentinite.


Introduction
Serpentinites are mainly composed of the serpentine mineral group, which includes hydrous phyllosilicates with an ideal chemical formula A 3 Si 2 O 5 (OH) 4 , where A is Mg, Fe 2+ , and Ni [1]. Serpentine mainly consists of a magnesium oxide octahedral sheet and a silicon-oxygen tetrahedral sheet, forming a closely connected layered structure [2]. According to the different arrangements of these layers, serpentine can be divided into three main serpentine-type minerals: lizardite, antigorite, and chrysotile [3]. Many scholars have used infrared spectroscopy, Raman spectroscopy, X-ray powder diffractometry, and other analytical methods to study the mineralogical and petrological characteristics of serpentinites [4,5]. Their deposits are widely distributed around the world. Among the most important varieties, we have Williamsite in America, Bowenite in New Zealand, and Xiuyan jade in China. According to the different formation mechanisms of serpentinite, it can be divided into two deposit types. One is formed in the hydrothermal metasomatic deposit, where Si-rich fluids react with Mg-rich dolomitic marble rocks at a medium temperature, such as Xiuyan jade [6][7][8] and Luchuan jade [9]. The other is mainly formed by the metasomatism of ultramafic rocks (e.g., dunite, containing mainly olivine) at a low-medium temperature, such as Taishan jade. The first type often contains high contents of SiO 2 , medium contents of MgO, and low contents of Al 2 O 3 and total Fe, while the second type is characterised by a higher amount of Fe, Cr, and Ni [6,10].
In ancient times, serpentinite was widely used in jewellery, decoration, and carving materials in the Mediterranean area, Asia, Oceania, and America [11]. It is valued for its typical yellowish-green and olive-green appearance, waxy lustre, and fine aphanitic structure. In recent years, a new type of gem-quality serpentinite called Tianqing Dong has appeared in the jade markets of China, and it was used to imitate jadeite when it first appeared. Unlike the other serpentinites that show a yellowish-green colour, it has a charming bluish-green hue, and there are few studies evaluating its quality.
For this study, a total of 59 natural bluish-green polished serpentinite samples were bought from the jewellery market in Xiuyan, Liaoning Province in China ( Figure 1). The samples were cut into well-polished rectangular blocks with an average size of 11 mm × 11 mm × 8 mm to avoid the influence of size and thickness on its colour. The refractive index is 1.56-1.57, and the density is 2.570-2.596. The transition of charge from Fe 2+ to Fe 3+ generates the bluish-green colour, and the d-d electron transition of Fe 3+ leads to the yellow colour, while Mn 2+ has a certain impact on it [44,45]. The colour of bluishgreen serpentinite is mainly caused by the charge transition from Fe 2+ to Fe 3+ , resulting in a broad absorption band centred at 630 nm, generating a weak, bluish-green hue [46].
The optical microstructure shows that the samples are mainly composed of serpentinetype minerals. It is characterised by colourless, parallel, or near-parallel extinction, and the interference colour is first-grade greyish white to yellow. It has interpenetrating blades ( Figure 2a) [47], and a few leaf-shaped serpentine-type minerals can be seen locally in directional arrangement (Figure 2b). The sample with finer particle size has a better texture and transparency than the others. The scanning electron microscope (SEM) images are shown in Figure 3. They have a curly and folded shape, and some of them have a lamellar crystallisation structure. The optical microstructure shows that the samples are mainly compo tine-type minerals. It is characterised by colourless, parallel, or near-para and the interference colour is first-grade greyish white to yellow. It has in blades ( Figure 2a) [47], and a few leaf-shaped serpentine-type minerals cally in directional arrangement ( Figure 2b). The sample with finer par better texture and transparency than the others. The scanning electron mic images are shown in Figure 3. They have a curly and folded shape, and have a lamellar crystallisation structure.
The main absorption peaks of the infrared spectra (FTIR) of the sam cm −1 , 1078 cm −1 , 1050-960 cm −1 , 632 cm −1 , 563 cm −1 , and 462 cm −1 (Figure 4 typical spectral of antigorite. It splits into two weak absorption peaks i 1050-960 cm −1 . Lizardite has a flat crystal structure, making the Si-O tet pseudohexagonal symmetry. So, the two in-plane vibrations are degenera only a broad band in the region of 1050-960 cm −1 . Chrysotile forms a co octahedron sheets outside and tetrahedron sheets inside. It splits into tw sorption peaks in this region because its symmetry is the lowest. Antigori of octahedral and tetrahedral sheets alternately and reversely, forming a structure. Therefore, it only has two weak absorption peaks in this re bending vibration of Mg-O at 563 cm −1 shows a sharp peak in antigori broad absorption band or a shoulder peak in chrysotile and lizardite [49]. The main absorption peaks of the infrared spectra (FTIR) of the samples are 3674 cm −1 , 1078 cm −1 , 1050-960 cm −1 , 632 cm −1 , 563 cm −1 , and 462 cm −1 (Figure 4), which is the typical spectral of antigorite. It splits into two weak absorption peaks in the range of 1050-960 cm −1 . Lizardite has a flat crystal structure, making the Si-O tetrahedron have pseudohexagonal symmetry. So, the two in-plane vibrations are degenerate, and there is only a broad band in the region of 1050-960 cm −1 . Chrysotile forms a coiled layer with octahedron sheets outside and tetrahedron sheets inside. It splits into two obvious absorption peaks in this region because its symmetry is the lowest. Antigorite is composed of octahedral and tetrahedral sheets alternately and reversely, forming a wavy curved structure. Therefore, it only has two weak absorption peaks in this region [48]. The bending vibration of Mg-O at 563 cm −1 shows a sharp peak in antigorite instead of a broad absorption band or a shoulder peak in chrysotile and lizardite [49]. The optical microstructure shows that the samples are mainly composed of serpentine-type minerals. It is characterised by colourless, parallel, or near-parallel extinction, and the interference colour is first-grade greyish white to yellow. It has interpenetrating blades ( Figure 2a) [47], and a few leaf-shaped serpentine-type minerals can be seen locally in directional arrangement ( Figure 2b). The sample with finer particle size has a better texture and transparency than the others. The scanning electron microscope (SEM) images are shown in Figure 3. They have a curly and folded shape, and some of them have a lamellar crystallisation structure.
The main absorption peaks of the infrared spectra (FTIR) of the samples are 3674 cm −1 , 1078 cm −1 , 1050-960 cm −1 , 632 cm −1 , 563 cm −1 , and 462 cm −1 (Figure 4), which is the typical spectral of antigorite. It splits into two weak absorption peaks in the range of 1050-960 cm −1 . Lizardite has a flat crystal structure, making the Si-O tetrahedron have pseudohexagonal symmetry. So, the two in-plane vibrations are degenerate, and there is only a broad band in the region of 1050-960 cm −1 . Chrysotile forms a coiled layer with octahedron sheets outside and tetrahedron sheets inside. It splits into two obvious absorption peaks in this region because its symmetry is the lowest. Antigorite is composed of octahedral and tetrahedral sheets alternately and reversely, forming a wavy curved structure. Therefore, it only has two weak absorption peaks in this region [48]. The bending vibration of Mg-O at 563 cm −1 shows a sharp peak in antigorite instead of a broad absorption band or a shoulder peak in chrysotile and lizardite [49].     The characteristic Raman peaks of antigorite occur at 230 cm −1 , 373 cm −1 , 46 685 cm −1 , and 1045 cm −1 [50], and the Raman spectra of the samples are shown in 5a. The peak at 230 cm −1 is the vibrations of O-H-O groups. O is the nonbridging atom of the tetrahedron SiO4, and H is the hydrogen atom of the OH group tilted an octahedral cation vacancy [51]. The peak at 373 cm -1 is the symmetric bending rahedra SiO4. The peak at 685 cm −1 in antigorite is produced by the Si-Ob-Si stre mode [52], while in chrysotile and lizardite, it shifts to about 690 cm −1 . The antisy ric Si-Ob-Si stretching modes cause a peak at 1045 cm −1 , which can distinguish an from chrysotile and lizardite [53]. Raman peaks of apatite at 964 cm −1 (Figure 5b) a    The characteristic Raman peaks of antigorite occur at 230 cm −1 , 373 cm −1 , 461 685 cm −1 , and 1045 cm −1 [50], and the Raman spectra of the samples are shown in F 5a. The peak at 230 cm −1 is the vibrations of O-H-O groups. O is the nonbridging ox atom of the tetrahedron SiO4, and H is the hydrogen atom of the OH group tilted to an octahedral cation vacancy [51]. The peak at 373 cm -1 is the symmetric bending o rahedra SiO4. The peak at 685 cm −1 in antigorite is produced by the Si-Ob-Si stret mode [52], while in chrysotile and lizardite, it shifts to about 690 cm −1 . The antisym ric Si-Ob-Si stretching modes cause a peak at 1045 cm −1 , which can distinguish anti from chrysotile and lizardite [53]. Raman peaks of apatite at 964 cm −1 (Figure 5b) an at 195 cm −1 , 362 cm −1 , and 677 cm −1 (Figure 5c) are also found in some samples. In The characteristic Raman peaks of antigorite occur at 230 cm −1 , 373 cm −1 , 461 cm −1 , 685 cm −1 , and 1045 cm −1 [50], and the Raman spectra of the samples are shown in Figure 5a. The peak at 230 cm −1 is the vibrations of O-H-O groups. O is the nonbridging oxygen atom of the tetrahedron SiO 4 , and H is the hydrogen atom of the OH group tilted toward an octahedral cation vacancy [51]. The peak at 373 cm −1 is the symmetric bending of tetrahedra SiO 4 . The peak at 685 cm −1 in antigorite is produced by the Si-O b -Si stretching mode [52], while in chrysotile and lizardite, it shifts to about 690 cm −1 . The antisymmetric Si-O b -Si stretching modes cause a peak at 1045 cm −1 , which can distinguish antigorite from chrysotile and lizardite [53]. Raman peaks of apatite at 964 cm −1 (Figure 5b) and talc at  (Figure 5c) are also found in some samples. In addition, other minerals such as dolomite and calcite are found in previous studies [46]. Therefore, according to the FTIR and Raman spectra, the bluish-green serpentinite samples are mainly composed of antigorite, including minor amounts of other minerals such as apatite, talc, dolomite, calcite, etc.

Background and Light Source
Munsell neutral grey colour chips contain 37 neutral grey colours wit lightnesses, ranging from N0.5 to N9.5. In this experiment, Munsell N9 was the test background in a neutral environment.
A shadowless dome light source with a colour temperature of 6500k was this experiment. Its interior was plated with a special diffuse reflective mater the light could be evenly irradiated to the surface of samples. The light source w directly above the sample 9.5 cm away.

Colour Measurements
The computer vision system was adopted to measure the colour of gemst ( Figure 6) [54]. The lens of the industrial camera capturing images is Mako g-5 resolution of 2464 (horizontal) by 2056 (vertical) and a frame rate of 23.7. It sensor of CMOS-type Sony IMX264 and a global shutter. The camera was co IPC through the GigE interface, placed vertically above the sample, and then th between them was adjusted to 25 cm to obtain the most realistic colour.
We selected the areas of serpentinite with uniform colour and no inclusio age capture, and the values of L*, a*, and b* were automatically converted and by using CKVisionBuilder V3.0 software (developed by Shenzhen CkVisio

Background and Light Source
Munsell neutral grey colour chips contain 37 neutral grey colours with different lightnesses, ranging from N0.5 to N9.5. In this experiment, Munsell N9 was selected as the test background in a neutral environment.
A shadowless dome light source with a colour temperature of 6500 k was adopted in this experiment. Its interior was plated with a special diffuse reflective material so that the light could be evenly irradiated to the surface of samples. The light source was placed directly above the sample 9.5 cm away.

Colour Measurements
The computer vision system was adopted to measure the colour of gemstones ( Figure 6) [54]. The lens of the industrial camera capturing images is Mako g-507c, with a resolution of 2464 (horizontal) by 2056 (vertical) and a frame rate of 23.7. It also has a sensor of CMOS-type Sony IMX264 and a global shutter. The camera was connected to IPC through the GigE interface, placed vertically above the sample, and then the distance between them was adjusted to 25 cm to obtain the most realistic colour.
We selected the areas of serpentinite with uniform colour and no inclusions for image capture, and the values of L*, a*, and b* were automatically converted and measured by using CKVisionBuilder V3.0 software (developed by Shenzhen CkVision Machine Vision Technology Co., Ltd., Shenzhen, China). Each sample was tested three times at different areas to obtain the average value and then converted into the colour tristimulus value X, Y, and Z using Colourtell software.

Colour Appearance Model
CIECAM16 forward model was used in this experiment ( Figure 7) by input three tristimulus values X, Y, and Z and other variables, including the adaptive white points Xw, Yw, and Zw, relative background brightness Yb, adaptive field lu nance LA, and the surround parameters c (the impact of surround), Nc (a chromati duction factor), and F (a factor determining the degree of chromatic adaptation). T after the steps of cone cell response transformation, colour adaptation transforma and nonlinear compression processing, the colour appearance attribute parameters w calculated, including the brightness (Q), lightness (J), colourfulness (M), chroma saturation (s), hue angle (h), and hue composition (H). The colour difference △E base CAM16-UCS is given by the following equations [41]: The colour difference between the two samples can be computed as the Euclid distance: where  ',  ′, and  ′ are the J', a', and b' differences between two samples.

Colour Appearance Model
CIECAM16 forward model was used in this experiment ( Figure 7) by inputting three tristimulus values X, Y, and Z and other variables, including the adaptive field white points Xw, Yw, and Zw, relative background brightness Yb, adaptive field luminance L A , and the surround parameters c (the impact of surround), Nc (a chromatic induction factor), and F (a factor determining the degree of chromatic adaptation). Then, after the steps of cone cell response transformation, colour adaptation transformation, and nonlinear compression processing, the colour appearance attribute parameters were calculated, including the brightness (Q), lightness (J), colourfulness (M), chroma (C), saturation (s), hue angle (h), and hue composition (H). The colour difference ∆E based on CAM16-UCS is given by the following equations [41]: The colour difference between the two samples can be computed as the Euclidean distance: where ∆J , ∆a , and ∆b are the J , a , and b differences between two samples.  The colour appearance properties of serpentinite samples were calculated under three different surrounds (average, dim, dark), five different illuminances (500 lx, 1000 lx, 1500 lx, 2000 lx, 2500 lx), and three different light sources (D65, D50, A). In the calculation, it was necessary to convert the illuminance E into the adaptive field luminance L A , and the formula is as follows:

Statistical Analysis
The CIECAM16 model was written in Microsoft Office Excel. One-way ANOVA and Pearson's correlation were used to analyse the influence of the transformation among different conditions on colour appearance attributes. All the data analysis was performed by using IBM SPSS statistics software (version 25, IBM, Armonk, NY, USA).

Surround Changes
The colour appearance observed by the human visual system is not only related to the tristimulus itself but also to the field of view, which has four components: stimulus, proximal field, background, and surround. Surround refers to all areas outside the background, and the colour of gemstones we observe is also affected by its changes. According to the ratio of surround brightness to the average brightness of the field of view, there are three types of surrounds in the CIECAM16 model: average, dim, and dark. When the ratio is greater than 20%, the surround is average. When the ratio is zero, the surround is dark. For the same serpentinite sample, the colour observed in a bright and dark room will be slightly different.
One-way ANOVA was used to analyse the impact of surround change on colour appearance attributes, and the results are shown in Table 2. When the significance coefficient p is < 0.05, it is considered to have a significant impact. We observe that the surround change exerts a significant influence on lightness (J), brightness (Q), saturation (s), colourfulness (M), and chroma (C) but has no significant impact on hue composition (H) and hue angle (h). F represents the between-group variance, and the F of brightness is the largest (775.37), indicating that the surround change has the greatest impact on brightness Q. The increase rate of different colour appearance attributes is shown in Table 3, and the maximum brightness was 40.73%, while the lightness was 14.65%.

Illuminance Changes
Illuminance refers to the luminous flux of visible light received by an object per unit area. In this part, we calculated the colour appearance attributes under five illuminances: 500 lx, 1000 lx, 1500 lx, 2000 lx, and 2500 lx. One-way ANOVA was used to analyse the influence of illuminance change on the prediction of serpentinite colour appearance attributes, and the results are shown in Table 4. The change of illuminance had a significant impact on brightness (Q), colourfulness (M), and saturation (s). The between-group variance F value of brightness was the largest, indicating that illuminance change had the most significant impact on it. Pearson; correlation coefficients of Q and M were 0.885 and 0.332, while there was a negative correlation between illuminance and saturation, and its Pearson correlation coefficient was −0.670 (Table 5). Then, nonlinear fitting was performed for the relationship between illuminance and average saturation, and a function was obtained:

Light Source Changes
Li et al. formed a new uniform colour space CAM16-UCS based on the CIECAM16 model and proposed to use lightness (J), colourfulness (M), and hue angle (h) as the threedimensional coordinates in this colour space [41]. When the light source was transformed from D65 to D50 and then to A, hue angle (h) and colourfulness (M) both increased significantly, while brightness (Q) decreased slightly. The colour difference ∆E of serpentinite samples under three light sources was calculated by using the CAM16-UCS colour difference formula (Equation (7)), and the results are shown in Table 6. The colour difference between D65 and A was 13.10. When the colour difference is greater than 6, the human eye can perceive it [55]. So, the colour of serpentinite samples changed apparently under different light sources. The coefficient of variation (CV) can be used to measure the dispersion of the data distribution, and it was calculated according to Equation (10). The CV of hue angle and the colourfulness decreased gradually as the light source changed ( Table 7), indicating that the colour distribution of serpentinite was more discrete under the D65 light source.
where σ is the standard deviation of samples, and µ represents the sample mean. The chromaticity values of samples based on CIE1976 L*a*b* under two different light sources were measured by the computer vision system (CVS). It was found that the a* value decreased and the b* value increased under the A light source, and the chroma value decreased gradually. The hue angle also decreased, shifting to the yellow area, contrary to the result calculated by the CIECAM16 model. Meanwhile, the colour distribution area of the samples was also more concentrated under the A light source (Figure 8).

Influence of Surround Change on Colour Appearance
When the surround changes from average to dim and then to dark, the lightness and brightness of serpentinite both increase gradually (Figure 9a), while the chroma, colourfulness, and saturation decrease with the gradual darkening of the environment,

Influence of Surround Change on Colour Appearance
When the surround changes from average to dim and then to dark, the lightness and brightness of serpentinite both increase gradually (Figure 9a), while the chroma, colourfulness, and saturation decrease with the gradual darkening of the environment, showing a negative correlation (Figure 9b). Compared with the darker light around, the sample will appear brighter to a certain extent. This is because human vision has the automatic colour balance function, which leads to lightness compensation. The growth rate of brightness is much higher than that of lightness. That may be due to the large field of view of the human eye, and it will be more affected by the surround, so the perception of brightness changes more clearly. Conversely, the chroma and colourfulness gradually decrease with the darkening of the environment. This is caused by the action of the cone cells and rod cells on the retina of the human eye. In the dark surround, rod cells play a major role and have only the perception of grey, causing the chroma and colourfulness of the sample in the dark surround to decrease accordingly. The sensitivity of rod cells tends to be saturated more easily in bright light. So, in the dark surround, they will be more sensitive to light changes, resulting in an obvious downward trend from dim to dark. However, bluish-green serpentinite belongs to low chroma (15.59-20.62). So, the overall trend of surround darkening on the decline of colourfulness and chroma is not obvious.
When the surround changes from average to dim and then to dark, the lightness and brightness of serpentinite both increase gradually (Figure 9a), while the chroma, colourfulness, and saturation decrease with the gradual darkening of the environment, showing a negative correlation (Figure 9b). Compared with the darker light around, the sample will appear brighter to a certain extent. This is because human vision has the automatic colour balance function, which leads to lightness compensation. The growth rate of brightness is much higher than that of lightness. That may be due to the large field of view of the human eye, and it will be more affected by the surround, so the perception of brightness changes more clearly. Conversely, the chroma and colourfulness gradually decrease with the darkening of the environment. This is caused by the action of the cone cells and rod cells on the retina of the human eye. In the dark surround, rod cells play a major role and have only the perception of grey, causing the chroma and colourfulness of the sample in the dark surround to decrease accordingly. The sensitivity of rod cells tends to be saturated more easily in bright light. So, in the dark surround, they will be more sensitive to light changes, resulting in an obvious downward trend from dim to dark. However, bluish-green serpentinite belongs to low chroma (15.59-20.62). So, the overall trend of surround darkening on the decline of colourfulness and chroma is not obvious.  In average surround, the lightness perceived contrast among the 59 serpentinite samples is the largest. As the surround changes from average to dim to dark, the dark area will appear lighter while the change of light area is not obvious, resulting in a decrease in the lightness perceived contrast, which confirms the Bartleson-Breneman phenomenon. Bluish-green serpentinite belongs to medium and high lightness (42.29-71.94) and low chroma (15.59-20.62). So, it can present a charming bluish-green hue with higher saturation in average surround.

Influence of Illuminance Change on Colour Appearance
As illuminance increases, brightness increases and so does the difference between the highest and lowest brightness, resulting in a horn-like distribution of the shadowed portion. These results are consistent with the Stevens effect, i.e., that brightness increases with increasing illuminance. At the same time, under high illumination, the serpentinite colour contrast will be more pronounced because the dark colour will be darker, and the light colour will be brighter, increasing the perceived contrast. The colourfulness M also increases, which confirms the Hunt effect, whereby the colourfulness increases with increasing illuminance. However, when the illuminance rises to 2500 lx (L A = 159.16), the average colourfulness decreases slightly. The colour of samples measured by the CVS under three different illuminances of 500 lx, 1500 lx, and 2500 lx (L A = 31.83, 95.49, and 159.16) are consistent with the results predicted by the colour appearance model, indicating that the CIECAM16 model can predict the effect of illuminance change on the colour appearance of serpentinite ( Figure 10). stones with low lightness, such as dark blue sapphire, increasing illuminance can ma them more attractive. However, bluish-green serpentinite has a high lightness and lo chroma, and consequently, too high a illuminance level dilutes its colour. Therefore, it not conducive to the presentation of bluish-green colour for serpentinite under high luminance. s = 100 × ( ) .
(1 Figure 10. Influence of illuminance change on colour appearance attributes M, Q, ands. Both c ourfulness and brightness rise with the increase of illuminance, but the growth rate of Q is mu greater than that of M, causing a significant decrease in average saturation. Figure 10. Influence of illuminance change on colour appearance attributes M, Q, and s Both colourfulness and brightness rise with the increase of illuminance, but the growth rate of Q is much greater than that of M, causing a significant decrease in average saturation. Saturation is the colourfulness of a stimulus relative to its brightness, and it can be calculated in the CIECAM16 model with Equation (11). The increased range of colourfulness (Pearson's r = 0.332) is much smaller than that of the brightness (Pearson's r = 0.885), so the overall saturation of the sample shows a downward trend. For some gemstones with low lightness, such as dark blue sapphire, increasing illuminance can make them more attractive. However, bluish-green serpentinite has a high lightness and low chroma, and consequently, too high a illuminance level dilutes its colour. Therefore, it is not conducive to the presentation of bluish-green colour for serpentinite under high illuminance.

Influence of Light Source Change on Colour Appearance
The colour of the gemstone perceived by the naked eye depends not only on the tristimulus value of the sample itself but also on the colour of the light source [56]. The CIE standard defines three standard illuminants, D65, D50, and A [57]. D65 and D50 are used to simulate the sunlight, and D65 (colour temperature of 6500 K) is often used as a standard light source for international colour evaluation, while D50 (colour temperature of 5000 K) is recognised as the standard colour temperature in the world printing industry. Light Source A (colour temperature of 2856 K) is often used in family rooms and jewellery stores. The energy peaks of D65 and D50 are mainly concentrated in the bluish-green area, which is around 430-470 nm, and the peak value of D50 shifts more towards the red region than D65. In comparison, the spectral energy distribution of Light Source A is mainly concentrated in the red area, which is about 560-780 nm. Therefore, the colour of serpentinite is more bluish green under the D65 light source and more yellowish green under the A light source ( Figure 11). The same results were obtained using the computer vision system (CVS) under different light sources (Figure 8).
However, the colour data calculated by the CIECAM16 model under different light sources are exactly opposite to the CVS measurement results (Figure 12). This is because human eyes can automatically correct the white balance of the colour. Our brain will perceive the colour of the surrounding light source and subtract it from the colour of the object. Therefore, when we observe the serpentinite samples under Light Source A, the calculated colour is more biased towards the complementary colour of this source, which is blue. However, serpentinite is an aggregate made up of tiny crystal particles that can reflect, refract, and scatter light, and it also exhibits a certain degree of transparency. Therefore, in real observation, the influence of the colour appearance phenomenon is very faint, and it is much less than the influence of the light source on the gem itself. That is why the hue of colour measured by the CVS is more inclined to the yellow area.

Influence of Light Source Change on Colour Appearance
The colour of the gemstone perceived by the naked eye depends not only on the tristimulus value of the sample itself but also on the colour of the light source [56]. The CIE standard defines three standard illuminants, D65, D50, and A [57]. D65 and D50 are used to simulate the sunlight, and D65 (colour temperature of 6500 K) is often used as a standard light source for international colour evaluation, while D50 (colour temperature of 5000 K) is recognised as the standard colour temperature in the world printing industry. Light Source A (colour temperature of 2856 K) is often used in family rooms and jewellery stores. The energy peaks of D65 and D50 are mainly concentrated in the bluish-green area, which is around 430-470 nm, and the peak value of D50 shifts more towards the red region than D65. In comparison, the spectral energy distribution of Light Source A is mainly concentrated in the red area, which is about 560-780 nm. Therefore, the colour of serpentinite is more bluish green under the D65 light source and more yellowish green under the A light source (Figure 11). The same results were obtained using the computer vision system (CVS) under different light sources (Figure 8). However, the colour data calculated by the CIECAM16 model under different light sources are exactly opposite to the CVS measurement results ( Figure 12). This is because human eyes can automatically correct the white balance of the colour. Our brain will perceive the colour of the surrounding light source and subtract it from the colour of the object. Therefore, when we observe the serpentinite samples under Light Source A, the calculated colour is more biased towards the complementary colour of this source, which is blue. However, serpentinite is an aggregate made up of tiny crystal particles that can reflect, refract, and scatter light, and it also exhibits a certain degree of transparency. Therefore, in real observation, the influence of the colour appearance phenomenon is very faint, and it is much less than the influence of the light source on the gem itself. That is why the hue of colour measured by the CVS is more inclined to the yellow area. Consequently, the CIECAM16 model is not suitable for predicting the hue of serpentinite under different light sources. Under the light source D65, the colour of bluish-green serpentinite measured by CVS is closer to the sample, and the distribution is more dispersed, so it is more suitable for colour presentation and mineral quality classification. Consequently, the CIECAM16 model is not suitable for predicting the hue of serpentinite under different light sources. Under the light source D65, the colour of bluish-green serpentinite measured by CVS is closer to the sample, and the distribution is more dispersed, so it is more suitable for colour presentation and mineral quality classification.

Conclusions
Gem-quality bluish-green serpentinite is mainly composed of antigorite, including minor amounts of other nonserpentine minerals such as apatite, talc, dolomite, calcite, etc. The CIECAM16 model can be used to predict its colour appearance properties. Changes in surrounds, illuminances, and light sources significantly affect the colour appearance of the samples. As the surround changes from average to dim and then to dark, the lightness (J) and brightness (Q) gradually increase, while the chroma (C) and colourfulness (M) decrease. The lightness contrast becomes more evident, which confirms the Bartleson-Breneman phenomenon. With average surrounds, serpentinite can present a charming bluish-green hue with higher saturation.
Brightness and colourfulness both increase with illuminance enhancement, indicating that the CIECAM16 model can predict the Stevens and Hunt effects. However, since the growth rate of colourfulness is much lower than that of brightness, a downward trend in saturation is observed. Consequently, too high a illuminance level does not favour the presentation of the colour of serpentinite.
According to the CAM16-UCS colour space, the change of the light source affects the hue angle (h) and colourfulness (M) of serpentinite. As the colour temperature of the light source decreases, the hue angle and colourfulness both increase, and the colour shifts towards the blue rather than yellow area. The CIECAM16 model may not be suitable for predicting the hue of serpentinite under different light sources. The D65 light source is more suitable for the colour presentation and classification of bluish-green serpentinite than the other ones (D50 and A).