House Painting with NCS in the USA

Giordano Beretta Computer Peripherals Laboratory HP Laboratories Palo Alto HPL-1999-79 June, 1999

color selection, Natural Color System, paint matching, design, architecture

We explore the use of the Natural Color System for designing color schemes in exterior and interior design. The availability of paints in the US market to realize such a color scheme is researched. Finally, we examine the performance of paint matching systems for one particular paint.

Presented at the Inter-Society Color Council's 68th Annual Meeting, Vancouver, B.C., 5-7 May, 1999. To be published in Atti della Fondazione Ronchi, Florence, Italy. Copyright Hewlett-Packard Company 1999

Introduction

1.0 Introduction
We recently had to repaint our house. Like many traditional houses, also ours was painted in a single off-white color inside-out. We asked several color scientists how they designed a color scheme for their house or apartment the last time they had to paint. We were surprised to invariably receive the same answer: I did not want to take any risk; I painted everything white. This prompted us to a challenge and try to design a bold color scheme. Based on previous work, we decided to design the color scheme as a set of symmetries in a perceptually dened color space. Although we had the software tools for designing a scheme in CIELAB, it turned out that using a computer was inconvenient and it was difcult to reliably predict in an uncontrolled environment the color appearance for small color nuances. The color chips available from paint stores were not suitable because the colors were not ordered in three dimensions. Since the second edition of the Natural Color System (NCS) atlas had recently become available and is widely used by designers in some European countries, we decided to design our color scheme using this system. Contrary to some other countries, here in the USA paint manufacturers do not use NCS to specify colors. Although most paint stores can computer match any color, this was not a viable solution in our case. The interior decoration of our house has a strong Japanese trait. One of the cornerstones in Japanese architecture is that color is not used to create varied light, but to intensify the natural effect of light. Thus we had to pay special attention to the spectral reectance curves of paints. This limited us to the use of pre-mixed paints, avoiding computer matches based on metamerism. In this paper we report on how we designed the color scheme, how we selected a paint brand, and how close we were able to match the colors in the NCS scheme with actual premixed paints available in the USA. For one of the colors in the scheme we also analyze how close a computer match is to a reference color in the real world of a paint store.1

2.0 Color schemes and color spaces

2.1 Color in architecture
In the past, the colors used to paint houses depended on the locally available pigments and their cost [5]. Today thousands of colors are available; many paint stores are equipped with colorimeters and ink formulation software that allow them to match any color within

1. The slides are available at http://www.hpl.hp.com/personal/Giordano_Beretta/Acrobat/Reports/ iscc2Slides.pdf and a set of photographs can be found in the directory http://www.hpl.hp.com/personal/ Giordano_Beretta/Bildli/PCD4004/

House painting with NCS in the USA

Color schemes and color spaces

a manufacturers gamut. This large choice makes it much harder to decide on a color scheme for a house. To help home owners in their decision process, most paint manufacturers compile standard color schemes among which people can choose the scheme they like best. However, especially in Europe there is a trend against strict methodologies and the systemization of color schemes for buildings [14], probably because the abstract choice of a scheme ignores the context in which the building is situated. This prompted us to ignore the standard color schemes provided by manufacturers. Most of the time architectural color is functional, as opposed to reference colors, which refer to natural objects. Although functional color can be arbitrary, it cannot be random, as random colors tend to clash and lead to coloristic garbage. To be sthetically pleasing, the colors in a scheme must be tied together by a fundamental principle. In earlier work related to the creation of business graphics [2], we noted how pleasing color schemes can be designed by selecting colors so that they form symmetries in perceptually uniform color spaces. Several authors, such as Nemcsics [13], Spillman [17], Tonnquist [18], Sivik and Hrd [16], Tosca [19], and Prieto [14], have written about the positive qualities of designing architectural color schemes following rules in a perceptually dened color space. Authors like Sivik and Hrd [16] have expressed strong opinions against notions like harmony and complementarity. From a semiotical point of view [3], in these papers the authors have looked at color mostly from a syntactical point of view, when color is purely concrete.2 In our case we are concerned with a specic architectural application and our considerations are about semantic, i.e., urbanistic and structural, functions of color. We hope this distinction will spare us from Sivik and Hrds criticism.

2.2 Selecting colors in CIELAB

In our earlier work [2], we had chosen CIELAB as the perceptually dened color space. The main reasons were: 1. it is based on CIE colorimetry 2. it is perceptually uniform with relatively good perceptual correlates 3. it is straightforward to implement in computer programs The tool we had developed was computer screen based, and allowed graphic designers to interactively develop a color palette for use in desktop publishing. Architectural color can be simulated on a computer screen, but in practice this is useful only to give a very rough idea in a sketch. Color appearance models are available to map the color reproduction on the screen so it appears the same as when viewed under an arbitrary illuminant, but this is
2. In Max Bills sense of color that exists by itself without any iconic relation to an object.

HPL199979

Color schemes and color spaces

cumbersome. Finally, a computer screen is tiny in comparison to a building, so it is impossible to judge the size effect, the interaction with the ambient, and the total appearance. In summary, while CIELAB is a very good color space for computer applications, an interactive software tool based on CIELAB is not a good choice for practical color selection in architecture.

2.3 Selecting colors with paint chips

Consumers can bring items such as fabric to the paint store and appreciate the combinations with paint chips (color swatches) under controlled illumination, or they can take chips home and appreciate them in situ when the color scheme can be designed from a clean sheet. The colors in a collection are often selected by professional color consultants, trend forecasters, and color stylists. The chips are arranged on strips of bristol or index grade paper and have words like system or calibrated in them. Paint stores offer large free collections of these paint chip strips. Most manufacturers order the strips loosely by hue and saturation. The problem with the paint chips is that the colors on each of them are often ordered in small groups of similar pigment formulation. Such an arrangement is not perceptual and there is no order nor scale. This makes it difcult to change a color in just one perceptual dimension; while a coarse color selection can be performed, it is hard to nd a perceptually close color. Furthermore, the imaginative color naming schemes popular with paint manufacturers are not helpful to describe the relation among colors in a palette how do Monet Moonrise and Dusty Mink interact? Why is a given color called Monet Moonrise by one manufacturer and Mesa Sunrise by another manufacturer? To successfully develop a scheme, we need a system that allows us to express the perceived relation among colors.

2.4 The natural color system NCS

The NCS natural color system [7] is a Hering-type color opponent system. Because it has the concept of color nuances colors of the same whiteness and blackness but different hue it is believed to be more suitable for designing color schemes for architectural uses [8, 17]. In several of the references in this paper, colors are specied in NCS. The Scandinavian Color Institute publishes an NCS color atlas that is produced to strict colorimetric tolerances. This atlas can be used in a light booth to select colors, on location to appreciate a color in context, and to match an existing color visually. If the atlas cannot be afforded, the Scandinavian Color Institute also sells an inexpensive color index, which is what we used, because the second edition atlas was not yet available when we realized our paint project. Ronnier Luos LLAB color appearance model allows the prediction of color in NCS notation from colorimetry.

House painting with NCS in the USA

Color schemes and color spaces

NCS is a Hering-type color space based on the three opponent color pairs black-white, red-green and blue-yellow. Within a hue, the colors are ordered by their white and black contents, from which saturation can easily be calculated. These three pairs form the axes for the three-dimensional NCS color solid, whose geometry is a double cone. In practice, a pair of two-dimensional representations (projections) is used: the NCS color circle and the NCS color triangle. The hue circle is dened by the red-green and blue-yellow axes. Each quadrant is partitioned in 10 slices of equal angle, as shown in Figure 1. The 400 indicated hues determine the NCS hue .
Y10R

Y20 R

Y3

0R
Y4 0R
Y 50 R

Y6

0R

Y7

0R

Y80

Y90R

HUE

B
FIGURE 1. The NCS color circle is a horizontal cut through the middle of the NCS color solid, where the four elementary colors are placed as the four cardinal points. Every quadrant between two elementary colors is divided into 100 steps. In this color circle, the hue Y70R is read as yellow with 70% redness.

Colors are described as relative amounts of the elementary colors so that the sum of the attributes is 100 [7]. Hence, units are percentages, unrelated to colorimetric quantities and unrelated to pigment mixtures. The hue notation consists of two letters which designate a
HPL199979

Color schemes and color spaces

quadrant and a number which indicates the relation between them in percent. The hue Y32R is yellow with 32% redness and 68% yellowness, while Y70R is a red hue with 70% redness and 30% yellowness. The NCS color triangle is a vertical cut through the NCS color solid at each hue, as shown in Figure 2. On the left side of the triangle is the gray scale from white (W) to black (S) and at the right of the triangle is the maximum chromatic color (C) of the hue. The color triangle enables the nuance of the color to be indicated, i.e., its degree of resemblance to the elementary colors white and black and to the full chromatic color of the same hue.

W
10 20 30 BLACKNESS 40 50 60 w 70 80 90 20 30 10 70 S 60 ES N 50 TIC MA 40 O R CH 80 90

FIGURE 2. The NCS color triangle is a vertical cut through the color solid at each hue. At the left side of the triangle is the gray scale from white (W) to black (S) and at the right of the triangle is the maximum chromatic color (C) of the hue considered. The black dot indicates a nuance with a blackness of 40 and a chromaticness of 30. The dotted line indicates the whiteness w = 30.

The scale on the vertical edge in Figure 2 indicates the NCS blackness s, and the scale on the lower edge indicates NCS chromaticness c (dened as the sum of the chromatic attributes). NCS whiteness w is the distance of a nuance from the edge SC of black colors. The sum of the values of the attributes is always 100, hence the whiteness must not be noted [7]. Also here, the scales for whiteness, blackness, and chromaticness are divided in 100 parts, so the numbers can be interpreted as percentages. The nuance notation indicates rst the blackness with two digits and then the chromaticness with two digits. The nuance notation
House painting with NCS in the USA

Color schemes and color spaces

4030 thus indicates a color with a blackness of 40 and a chromaticness of 30. The whiteness of this nuance is 100 40 30 = 30. The nuance and hue notation together give the complete NCS notation. In this notation S 4030Y70R, S stands for Second Edition, 4030 stands for the nuance, i.e., the degree of resemblance with black S (40% blackness s) and the maximum chromaticness C (30% chromaticness c). Finally, Y70R stands for the hue. The NCS chromaticness is intermediate between CIE 1976 chroma and CIE 1976 saturation [8]. The NCS saturation is dened as [7] c m = -----------c+w or c m = ---------------100 s (EQ 1)

The atlas indicates also the NCS lightness () and the photometric luminous reectance factor (Yl), which are not used in this paper. The atlas contains a total of 1750 color samples; a table at the end contains the NCS notations for the primary NCS standards after measurements. As an opponent hue system, NCS is not perceptually uniform; for example, the hue circle is not equi-spaced [18]. Therefore, for evaluating color differences we must convert the colors to the CIELAB color space, where we can use the CIE94 formula [4] E * 94 = with SL = 1 S C = 1 + 0.045 C * ab S H = 1 + 0.015 C * ab and kL = kC = kH = 1 (EQ 4) (EQ 3) L * 2 C * ab 2 H * ab 2 --------------- + --------------- + --------------- k L S L k C S C k H S H (EQ 2)

The computation of the lightness and chroma differences is straightforward. We compute the hue difference with the formula proposed by Sve [15]
2 h H * ab = 4 C * 1 C * 2 sin ------- 2 2

(EQ 5)

HPL199979

Color schemes and color spaces

2.5 Selecting colors with NCS

Sivik and Hrd [16] have explained in detail how to design color schemes with the NCS system. With the atlas it is very easy to determine the locus of colors with a constant attribute as described by Hrd [7]; this is illustrated in Figures 38. In Figure 3 the invariant is the whiteness (as indicated in Figure 2, the whiteness is the distance of a nuance from the chromaticness scale edge). The locus of points in such a scheme is a straight line parallel to the edge SC. The colors on the edge SC itself are called deep. Figure 4 shows the case in which the invariant is blackness. The loci in this case are straight lines parallel to the edge WC. The colors on the edge WC are called bright.

W
10 20 30 40 50 60 80 70 60 80 40 90 30 20 10 50 70 90

W
10 20 30 40 50 60 80 70 60 80 40 90 30 20 10 50 70 90

bright colors

deep colors

FIGURE 3. Color samples with the same NCS whiteness w.

FIGURE 4. Color samples with the same NCS blackness s.

This ability to form different color combinations based on whiteness and blackness is one of the strong points for architects and designers in the NCS system. Figure 5 illustrates that color samples which have the same chromaticness c are found on a line segment parallel to the gray edge WS. Figure 6 indicates that color samples having the same nuance are found in the same position in the color triangle, regardless of hue. The nuance is specied with the four digits comprising the blackness and the chromaticness, in our example 2030. The last two invariants for designing color scales in NCS are less common, at least in the literature. Figure 7 shows that color samples which have the same saturation are found on a straight line through the black point S. Note that contrary to the other attributes, which are percentages, the NCS saturation is a fraction. This is also the case for the NCS lightness. Figure 8 shows how in the atlas color samples having the same lightness are found on a straight line, the location of which depends on the NCS hue .
House painting with NCS in the USA

10

Architectural principles for a color scheme

W
10 20 30 40 50 60 80 70 60 80 40 90 30 20 10 50 70 90

W
10 20 30 40 50 60 80 70 60 80 40 90 30 20 10 50 70 90

FIGURE 5. Color samples with the same NCS chromaticness c.

FIGURE 6. Color samples with the same NCS nuance 2030 (regardless of hue).

The NCS atlas 96 has a page with the hue circle, a page with the neutrals and off-whites (c = 00 and c = 02), and all pages with the hue leaves in steps of ten units; nally there is a table with the colorimetrically precise NCS-notations. The NCS index is divided into four groups, one for each hue quadrant. In each groups there is an index page for each nuance. The various hues in the quadrant are always on the same location on the index page. In our project the priority was for the semantics of the color scheme. Therefore, we did not use Figures 38 to build a color scheme, because that would only by syntactic. Instead, we used the invariants as starting points from which we could change the color for a certain function by changing only one attribute at a time. We used the NCS atlas for guidance, not as a rigid set of principles or rules. The atlas allowed us to limit the total attributes in the color scheme, by identifying hues and nuances, although in practice we had to compromise on the available paints.

3.0 Architectural principles for a color scheme

3.1 Exterior
When a dwelling is painted, often two different color schemes are used. The rst scheme is designed by the architect and applies to the faade, trim, and doors. The second scheme is compiled by an interior designer or decorator for the paints to be used inside the dwelling. In the project described in this paper, we were able to design a single color scheme for both exterior and interior.
HPL199979

Architectural principles for a color scheme

11

W
10 20 30 40 50 60

m = 0.25 m = 0.5 m = 0.75

W
0.9 0.8 0.7 0.6 0.5 0.4

Y70R

C
90 80

C
90 80 70 60 50 40 30 20 10

LIGHTNESS

70 60 80 40 90 30 20 10 50

70

0.3 0.2 0.1

FIGURE 7. Color samples with the same NCS saturation m. Colors on such a line are called a shadow-series.

FIGURE 8. Color samples of hue Y70R with the same NCS lightness . Note the ordinate here indicates the lightness .

In architectural color design, not only are the interrelations among color elements important, but more fundamentally the exterior relations to building surroundings (contextual aspect) and building structure (structural aspect) [17]. To harmonize with the existing neighboring houses, the design started with the selection of a color for the faade. 3.1.1 Urban context The house is located in an area of a suburban town that at development time was at the towns periphery, where apricot orchards gave way to marshland. It is part of a subdivision created shortly after World War II to provide inexpensive housing to young families formed by soldiers who had returned from the war and were at the beginning of their civilian career. The modest lots and houses are relatively small and have a very simple oor plan. Our house consists of two parallelepipeds with a square base: one for the living space and one for the garage. With the creation of the Silicon Valley, the town grew, engulng and extending beyond the subdivision. More recently, in the roaring period of quick economic wealth of the 90s, the area between the city center and this peripheral area has undergone quite dramatic changes. Block after block properties have changed hands and the new owners have replaced the old houses with large volumes that Tosca labels modern mish-mash buildings [19].

House painting with NCS in the USA

12

Architectural principles for a color scheme

Our goal was to build a semiotic bridge between our modest ranch style house and the new swanky residences. This suburban city is embedded in a lush articial landscape known as urban forest, consisting of streets lined by tall trees3 and properties with numerous trees and shrubs. Under this dense canopy and green environment, many houses are painted in earth colors. 3.1.2 Faade color Driven by this contextual consideration and the temptation to make a semiotic allusion, we decided to choose an earth color for the faade. Due to the difculty to judge the effect of a color from a small patch in a color atlas, we walked the neighborhood streets until we found a house with a color we liked. This color was burnt umber4, a red-brown color from iron and manganese compounds that traditionally depended on the fact that iron oxide had been formed during the burning. Our house is on a street lined with mature camphors; the property line in the back is marked with tall redwoods (Sequoia sempervirens). The house on the left from the street is painted in a vivid light blue. This combination of surrounding colors prompted us to opt for a post-modern color that is just on the outer chromaticness border in the locus of burnt umbers [5]. We measured the faades color by placing samples from the NCS color index directly on the surface, as described by Fridell Anter for measuring inherent color [6]. On a sunny March day at 10 am we used two observers with normal color vision, as described in op. cit. The NCS specication for this color is S 4010Y70R; for illuminant D65 and the 10 standard observer the CIELAB coordinates are (L*, C*, h) = (60, 12, 50). Color is only one aspect of appearance. The faades of the new swanky houses are typically nished in a high granularity stucco. This texture of the surface modulates the light depending on the sun position, generally darkening the appearance. By contrast, the faade in our ranch style house is covered with the characteristic wood siding, which is a smooth surface. This put us in a difcult quandary. We could compensate for the stuccos granularity by increasing the blackness in the color specication. However, the house is situated in a hot meridional5 climate and has no insulation in the walls, so we wanted to keep the color as light as possible. With two tall redwoods in the back; several mature privets (Ligustrum lucidum), camellias (Camellia japonica), and various Rhododendra around the house, there is always some shade on the walls. We thought that if we chose a paint that has a spectrum with inec3. According to the Citys Comprehensive Plan, Goal N3, the urban forest consists mostly of southern magnolia, American sweetgum, London plane, Modesto ash, camphor, Chinese elm, and holly oak. 4. Umbers are brown, beige, and grey-pink colors of low chromaticness. 5. Approximately 3725N.

HPL199979

Architectural principles for a color scheme

13

tions, illuminating changing spectral composition from the vegetation and the suns path would provide sufcient color modulation to compensate for the at wall surfaces. 3.1.3 Structural detail Architecture is the creative work with volume to design space for living purpose. Colors have a spacial effect; warm colors tend to bring objects nearer, while cold colors appear to remove objects; lightness contrast makes this effect more pronounced [13, Section 5.1.1]. This means that the colors in a faade have to be chosen carefully because they interact, a fact already described systematically by Leonardo da Vinci. Nemcsics discusses in detail how to select the colors for structural detail depending on context and structure. The new houses we used for inspiration have more or less elaborate features in their faades, such as gutter, cornice, course, casement, sash, and footing. The most architecturally prominent detail is the course. Since the course can be extensive, in the new houses a neutral color such as black or a light gray is often chosen. This concords with Toscas observation that wealthy people tend to select schemes with few dark and muted colors [19]. In our modest ranch house these features are mostly missing, which confers the construction a boxy look. This appearance is countered by a relatively tall fascia under the Dutch gutter. In the living space box, this fascia and the casements are the only features. The garage box has more features; when a room was added on top, the overall architectonic balance was broken, and architect Tony A. Carrasco added some features to break the tall box into two stacked at boxes. On the side forming a concave angle with the main part the architect added a conspicuous pergola (trellis) continuing the fascia, while in the front he added a redundant gutter optically continuing the fascia. On the third side he added a stair with a conspicuous bannister. With this measure the tall box was broken into two stacked at boxes, maintaining the ranch style look of the house. Because of this modest nature of the features, we decided to add bold colors to our color schemes. For the trim (fascia, casement, sash, bannister, and pergola) we decided for a hue diametrically opposed to the faade color, i.e., B70G. We would have preferred a 0505 offwhite nuance, but it is not available in the NCS atlas for this hue. Our selection therefore fell on S 0510B70G. This increased chromaticness intensied the overall post-modern character of the color scheme. For the main door we decided to select a hue half-way between faade and trim: R20B. This hue is 10% more reddish than pure symmetry would dictate. To avoid too garish an appearance, we decided to select the hue one page towards the faade color (like a typographic ligature), since it occupies a larger surface (the door is in framed with a casement of trim color). For the door nuance we decided for the same chromaticness as the dominant interior walls, namely 50 (see below). This invariant prepares the visitor at the threshold for the chro-

House painting with NCS in the USA

14

Architectural principles for a color scheme

matic atmosphere inside. For the blackness, we decided for a deep color, matching the exterior wall in which the door is set. The full color specication for the main door is thus S 4050R20B, i.e., the same blackness as the exterior wall, the same chromaticness as the interior wall, a hue at balance between the exterior wall and trim.

3.2 Interior colors

In the previous section we have seen how the color scheme for the exterior was chosen strictly in function of the urban context of the house. For the interior such a constraint is no longer necessary; however, the interior colors were not chosen in a new scheme, rather the scheme of exterior colors was extended inside the house, as mentioned in the case of the door color. As a guideline to understand how to extend the scheme from the exterior to the interior, it is necessary to understand the two authors cultural baggage. Since city colors are experienced predominantly from the streets rather that the interior of residences, we rst explore the memory images ingrained in the authors visual system. This discussion will allow the reader to understand how we were able to extend the exterior color scheme to the interior. Lenclos, who has systematically analyzed and catalogued the colors in various geographic regions, has described how each nation, town, and village has its own particular chromatic identity. In his own professional formation it was important to discover the difference between the Western color of his native French Normandy and the Japanese color of Kyoto where he later studied [11]. 3.2.1 East versus West One of the authors grew up in Lugano (Switzerland), while the other grew up near Kyoto (Japan). Both cities have been architecturally inuential and have important buildings. Lugano and it environments have been the home of the architects of Rome and St. Petersburg, Francesco Borromini being the most famous. When these architects came home in retirement, they applied their knowledge and experience to the buildings of the local potentates. Over the centuries, this region has produced a number of inuential architects, the most well know contemporary being Mario Botta, who has changed Luganos color by bringing back terra-cotta as the main exterior color for large commercial buildings. Botta is also well known for his analytic functional placement of windows and tight design of interior lighting. With this background, it is easy to understand how Luganesi have a strong opinion on architecture and color. Kyoto as the capital of Japan for more than Thousand years bears a strong role in dening Japanese architecture and its colors. As noted by Akita [1], the colors of Kyoto are the green with landscape, brown with old wood buildings, and vermilion as an accent color in structures. Akita found that three colors are appropriate for the color of a wall in Kyoto: brown (ju-raku), reddish brown (ben-gara), and vermilion (shu-nuri). Paint as a protective coating of houses came to Japan only recently, explaining the importance of
HPL199979

Architectural principles for a color scheme

15

wood color; the rst major painted building has been Kyotos 1895 Heian Jingu (Shrine), in which the paints color is vermilion. We have seen in Section 3.1.3 on page 13 how Westeners use color to bring extra life to things. The Japanese use color to make things fade towards nothingness. As Kengo [9] exposes, Japanese use color not to create visual light but to intensify the natural effect of light. In an apparent contradiction when the natural effect of light is weak or non-existent in Japan architectural structures can be painted with vivid colors, hiding the original material and its latent effect, making even shape secondary to color. We compounded the tension brought by this apparent contradiction with a violent dialogue between Japanese and Western color schemes. 3.2.2 White and off-white interiors Before the development of modern hygiene in the last century and climate control devices such as central heating and air conditioning in this century, pests and molds were a major problem in living spaces, especially kitchens and bathrooms where steam is formed. For ages the remedy had been to regularly limewash walls, because slaked lime (water reacted with calcium oxide) is a caustic alkaline substance. Slaked lime (quicklime) is white, which brightens rooms that receive limited light though small windows (insulation is also a modern conquest) or from dim articial light sources. Other rooms were often painted in muted colors. In the 30s and 40s, Fillacier rst introduced the concept of interior color schemes where vivid colors are in a dynamic equilibrium [14]. Unfortunately, around 1955 white walls began to conquer interiors, reaching a peak in 1975 and only receding very slowly [12]. White is a convenient color for those insecure in their approach to color, because it does not clash with any color. Although white is a safe color, it is also psychologically sterile; Mahnke and Mahnke have devoted an entire chapter of their book [12] to a case against white in interiors. For our scheme we decided to nd the most dynamic equilibrium. Yet, a living space must also be a peaceful environment where one can nd calm and serenity after a day in todays frantic ofce world. We established a basic static equilibrium rule by following the recommendation of Nemcsics [13] that the ceiling should be lightest and the oor darkest, leaving the wall colors of a lightness between those of the oor and ceiling for a pleasant feeling. Since the oors are made of a light natural oak, to maximize the dynamic range in which we could choose wall colors, we decided to use a white designer base for the ceilings, if possible with titanium brighteners.

House painting with NCS in the USA

16

Architectural principles for a color scheme

3.2.3 Color for a study First we worked on the scheme for a study. We decided to keep the exterior trim color S 0510B70G inside the room. Trim includes all casements, sashes, doors, baseboard, and built-in book shelves. To further strengthen the interior-exterior relationship, we decided to keep the walls in the same hue as on the exterior, i.e., Y70R. Browsing a Patricia Guild book on interior design, the photograph of a room with hardwood oor, a bright white ceiling, and a vermilion wall was an immediate attention getter, because it was so reminiscent of Heian Jingus shunuri. We were quickly able to settle on a vivid S 0550Y70R, which is on the locus of the brightest colors for this hue (see Figure 4 on page 9); this brightness was constrained by the oors luminance. 3.2.4 Living room color The living room is decorated with Japanese furniture and prints, combined with neutrally colored Italian furniture. The room receives light from three cardinal directions plus a skylight in the ceiling. The strong presence of light as a modulating element calls for a more subdued color scheme that can intensify the natural effect of light as commanded by the Japanese tradition. The rst step to diminish the impact was to approximately switch the proportions of yellow and red in the wall color, keeping the nuance from the rst room to maximize the continuity in the color scheme. Hence, for the NCS color specication for the living room walls we settled on S 0550Y20R. To increase the dynamism, we decided to add an alternate wall color for some walls in the living room. The idea was to choose a bright and light color at the entrance, to form a virtual atrium creating a sequence for the person entering the house from the doors chromaticness to a bright but muted intermediate color when the house is entered, leading to a room of the same chromaticness as the door. To minimize the parameters in the color scheme, while keeping a hue contrast to the dominant wall color, the hue for the alternate wall color was to remain the Y70R of the exterior wall and the studys wall. For the nuance, the brightest and least chromatic sample in the NCS color atlas xes this alternate color in the scheme to S 0505Y70R. We found 0510 would have been too strong. The rooms dynamism was kept in bounds by using the alternate color for all trim, instead of the contrasting color used in the study. Also the bathrooms, bedrooms, and their trim was painted in this alternate color, except for the white ceilings. The molding was also painted in white to optically elevate the ceilings. Table 1 summarizes the selected color scheme. The cylindrical CIELAB coordinates have been obtained by measuring the samples in the NCS color atlas in 5 locations each (correHPL199979

Architectural principles for a color scheme

17

sponding to the spots of a dice showing the number ve) and averaging the results. The viewing conditions are the 10 CIE standard observer, illuminant D65, and the perfect reecting diffuser for the white reference.
TABLE 1. Color scheme for a house Color function Exterior wall Exterior trim Interior wall study Interior wall living Interior wall alternate Interior trim Main door Ceilings NCS specication S 4010Y70R S 0510B70G S 0550Y70R S 0550Y20R S 0505Y70R S 0505Y70R S 4050R20B S 0000N CIELAB Lightness 60 93 71 82 92 92 28 100 Chroma 12 9 50 57 8 8 39 0 Hue 50 166 43 73 67 67 2 0

The color scheme is represented graphically in Figure 9. The left side indicates the hues and the chromaticness, while the slice on the right indicates the nuances. The nuance 0510 was chosen because the nuance 0505 is not available in the color atlas for the particular hue of the exterior trim and 0510 would have been too chromatic for the wall alternate and interior trim; with this exception, there are four hues and four nuances in this scheme of seven structural colors plus white. These nuances are at only two blackness values and two chromaticness values.
Y
100 Y20R

W
10 20

50

Y70R

30 40

10

50 60 80 90 70 60 80 40 90 30 20 10 50

R20B B70G

70

FIGURE 9. Graphical representation of the data in table 1.

House painting with NCS in the USA

18

Architectural principles for a color scheme

3.3 Color interpretation

3.3.1 Japanese view The scheme as a whole is not Japanese, as this would have to be dominated by a dark brown and white. All we can do is look at colors individually; we have already seen how the color for the study interior wall is an immediate reference to Heian Jingus shu-nuri. For the other colors we can refer to Kobayashis color image scale [10], which is a popular tool in Japan. We could look up close matches between NCS atlas patches and patches in the color image scale, but the pocket booklet sold here in the USA is printed with ordinary CMYK process colors, which is not sufcient for colorimetry. The exterior wall color is a less chromatic version of rose gray (R/Gr in Kobayashis notation), which he describes as follows. This is a gray color, with a slight tinge of red. It is pleasant and calm, with a warm feel to it. [] Combined with light grayish or light tones, it is gentle and elegant, while in combination with dull or deep tones, it gives a Japanese effect, pleasant and sedate. In this sense, the dynamic balance of our vivid color scheme is very non-Japanese. The other colors in our scheme do not have close matches in the color image scale. The color that has some similarity to our interior wall living color is Gold (Y/S). Kobayashi explains that the Japanese name for this color evokes ripening rice and is associated with nature, contrary to the Western association with luxury. Used with vermilion, it has a luxurious and substantial appearance that used to be popular with the merchants of Osaka. When suggesting color schemes, Kobayashi combines this color with hues of similar nuance. 3.3.2 Western view We have discussed at length the exterior wall color and how the color scheme was built. Mahnke and Mahnke describe the character of colors when applied in interiors on ceilings, walls, and oors. The main interior colors in our scheme would be characterized as follows: Interior wall study. Warm, luminous. Has livable charm. Reection on the skin may enhance some skin tones. Interior wall living. Warm, exciting to irritating. Interior wall alternate. Aggression-inhibiting, weak, sweet. Ceilings. Empty. Helps to diffuse light sources and reduce shadows. Statements on single colors are of very limited usefulness. Their foundation can also be subjective; in fact, the living room has a more sedate feeling than the study, in part because
HPL199979

Paint selection

19

of the color itself (although Table 1 reveals that despite a constant nuance, the living room is 9 jnds lighter and 7 jnds more vivid), in part because of the opposite choice in the trim color. Maybe the dialogue between irritating and aggression-inhibiting walls and their correct geometry created just that dynamic balance we were looking for. Readers who wish to study in depth the analysis and synthesis of color schemes in architectural applications are urged to consult Nemcsics book on color dynamics [13]. In our project we did not consider this work because it is based on the Coloroid Color System, while our intent was to exercise the Natural Color System.

4.0 Paint selection

When the color scheme was complete, the next task was to nd actual paints. This task required the solution of two problems: 1. In the last paragraph of Section 3.1.2 on page 12 we mentioned that we wanted a paint with a spectrum that has inection points, so that the illumination modulates the exterior walls. In Section 3.2.1 on page 14 we saw how in Japanese interiors color is used to intensify the natural effect of light. Consequently we wanted a similar spectral characteristic of numerous inection points, and possibly steep slopes, also for the interior paints. 2. In the USA paints are not specied with the NCS notation, so for each color in our scheme we had to nd an actual paint. The rst decision was that even before painting, on the walls that required plastering we would specify the most textured application that the plasterer was offering. This texture would create local shadows, which would be tinged with the complementary color of the illumination during daylight. We further carefully controlled the geometric appearance parameters by consciously specifying the sheen for the various paints: ceilings at, walls eggshell (velvet), trim semi-gloss (satin), and main door high gloss. The second decision was to use only colors from the manufacturers standard formulary as demonstrated in paint chips. We anticipated several advantages: 1. The paint chips can be used as proxies for the actual paints at a considerable saving of time and money. 2. It is not necessary to store spare paint for subsequent patching and touch-up jobs, since a color would be available for many years. 3. Standard recipes are likely to be robust because they are based on entire units of the pigment dispenser. Thus two cans of paints bought on different days would produce colors that match with a high probability. 4. We speculate that the manufacturer has carefully selected recipes that use a low number of compatible pigments and therefore maintain a high chromatic purity.

House painting with NCS in the USA

20

Paint selection

We assumed that paint manufacturers apply due diligence in the manufacture of the paint chips, i.e., that they would use pigments closely matching those in the actual paint. This assumption allowed us to take the free chips to our laboratory and measure them right away, without having to buy a pint of paint, apply multiple coats and letting each dry, before being able to perform measurements in the eld. We also assumed that the manufacturer would be able to apply a more uniform and consistent paint coat than we could. We evaluated a large number of paints because we could easily feed the colorimetric data directly in a spreadsheet for evaluation.

4.1 White paint

For the white paint we used our nger to smear some paint on a piece of bristol grade paper. Under the light source of the stores paint chip display we could easily determine the brand with the brightest white. For verication, we applied two coats of two white paints on two adjacent areas of a ceiling, each several square meters in size. The difference in lightness was even more pronounced than on the bristol paper. This difference was so striking that it was not necessary to use any instrumentation. The colorimetry for the sample on bristol is L* = 98, C* = 3.

4.2 Spectral reectance

For convenience, we decided to limit ourselves to the local neighborhood paint store instead of attempting a systematic survey of paint manufacturers. Our store carried four paint brands: a store brand, two national brands, and a famous designer brand. We ruled out the designer brand because of its high cost and the probability that it was just a repackaged variant of one of the other brands. Standing in front of the paint chip displays, the difference in the size of the formularies was striking. The two national brands had both large selections, ordered roughly by hue and saturation. The store brand had a much smaller selection, covering a good size color gamut; there was no close match for some of the colors in our scheme. The designer brand was instead very small and the colors were grouped by emotional criteria, with no apparent order inside a group. To compare the brands, we chose one color and compared the chip spectra for that color. For the color we chose the trim S0510B70G because it is a light color; it would require small quantities of pigments, hence be harder to mix. To select the chips, we matched visually the sample from the NCS index with the store display. For the spectral quality, we wanted many inection points and steep slopes so the paint is modulated by the ambient light. Figure 10 shows the result. The designer brand has a relatively at spectrum, so it would simply follow the incident lights spectrum. Like the store brand, it has a relatively large hue difference as computed with (EQ 5) to the target color, which is due to the lim-

HPL199979

Paint selection

21

ited number of colors available in the palette. In the blue-green portion of the spectrum, national brand 2 had the steepest slope, so we chose that brand for all of our paints.
reflectance 1.0

0.8

0.6

0.4 Designer brand Store brand 0.2 National brand 1 National brand 2 wavelength 0.0 400 450 500 550 600 650 700 nm

FIGURE 10. Spectral reectance curves of four paint brands approximately matching the color S0510B70G for the trim. The fourth paint has the steepest slope in the dominant range for hue B70G.

After selecting a paint brand, we had to determine the closest standard color for each color in our scheme. We accomplished this task in three steps. In a rst step we used the NCS index to visually nd the paint colors approximately matching each color in our scheme. In a second step we measured the chips for these colors and determined the closest match in the sense of CIE94, equation (EQ 2) on page 8. The third step was iterative. We bought a pint of the paint, painted an area of about 1 m2, surrounded it by a trim line, and lived with it for a few days. If not satised, we then used the NCS index to nd a closest alternate color. To nd this alternate color, we used not only the paint chips in the display, but also a booklet with color cards representing standard colors not in the store display. Figure 11 shows the chips spectral reectance curves for the selected colors. Because of this third step, the color nally selected was not always the one closest to the color in our scheme in a colorimetric sense, but the color we subjectively felt looked the best. Therefore, the color difference between each color in our scheme does not truly indicate how reliably USA paints can be used to implement an NCS specication. Instead, this difference indicates how useful the NCS system is to build a color scheme from the stan-

House painting with NCS in the USA

22

Paint selection

1.0 reflectance 2022

interior alt.

0.8 1401
trim

0.6

1804
interior living

1854
interior study

0.4

2038
exterior

0.2 1033
door

wavelength [nm] 0.0 400 450 500 550 600 650 700

FIGURE 11. Spectral reectance curves of all selected paint chips. The curves are interesting, with abundant inection points and steep slopes, which together help illumination modulating the colors. See Figure 12 on page 25 for the ceiling.

dard palettes of colors offered by USA manufacturers. Table 2 lists the paints chosen and the color difference to the NCS specication.
TABLE 2. Selected paints form those available from national brand 2 Color function Exterior wall Exterior trim Wall study NCS specicn. S 4010Y70R S 0510B70G S 0550Y70R Paint chip L* 2038 1401 1854 58 91 71 CIELAB C* 10 7 48 h 51 139 43 CIE94 2 4 1 L* 2 1 1 Chip to NCS C* 3 2 1 h 0 -4 0

HPL199979

Paint selection

23

TABLE 2. Selected paints form those available from national brand 2 Color function Wall living Wall alternate Interior trim Main door Ceilings NCS specicn. S 0550Y20R S 0505Y70R S 0505Y70R S 4050R20B S 0000N Paint chip L* 1804 2022 2022 1033 designer white base 81 93 93 37 98 CIELAB C* 52 8 8 33 3 h 70 67 67 0 102 CIE94 2 1 1 9 3 L* 1 -1 -1 -9 -2 Chip to NCS C* 5 0 0 6 3 h -3 0 0 -1 0

The color differences in Table 2 were computed using (EQ 2) on the full precision6 data, not the values listed in this Table and in Table 1 on page 17. Therefore the differences in the perceptual correlates do not add up to the total color difference due to rounding. The viewing conditions are always illuminant D65 and 10 observer.

4.3 Chips and paints

The largest difference was in the color for the main door, for which we were able to nd only a darker color in the manufacturers palette. We believed this would be a smaller problem than might appear at rst, because we specied a high gloss paint. High gloss paints are very transparent, and we used a pure white undercoat to hopefully obtain a total color appearance more closely matching the specied color. The other colors matched quite accurately the corresponding NCS colors in the designed scheme. The color difference between the perfect reecting diffuser and a real physical paint chip is of course somewhat large. Since we nally judged painted colors and not chips, we need to evaluate the color differences between painted colors and the specied NCS colors. To gauge the validity of using paint chips as proxies of the paints, we also have to compare the painted colors to the chips colors, especially since we have effectively exploited the transparency of some paints.7 The colorimetry for the painted colors was somewhat tricky because of the textured plaster. The best solution would have been to use a uniform light source to illuminate a sufciently large portion of the wall and measure it with a spectroradiometer. We judged this to be an overkill for a house painting project and resorted to a simpler approximate method. We used a spectrophotometer with a small aperture (4 mm) and tried to nd possibly at portions in the plaster. For each paint we made 10 measurements in different locations and averaged them, with exception of the living room and the main door, which were outside the reach of a spectrophotometer tethered to a computer. Table 3 shows for each color the

6. The accuracy of the instrument used was 0.3 CIELAB units, determined in conformance to ISO 9001. 7. For the exterior walls we used two undercoats, tinging the second one with the nal color. For the interior walls we applied several coats to obtain complete coverage.

House painting with NCS in the USA

24

Paint selection

average paint color and the standard deviation. The standard deviation of the designer white base is the best indication for reliability of the measurements.
TABLE 3. Paints applied to surface Color function Exterior wall Exterior trim Wall study Wall living Wall alternate Main door Ceilings Applied paint L* 2038 1401 1854 1804 2022 1033 designer white base 63 90 71 80 94 28 96 C* 9 8 48 54 9 37 2 CIELAB h 49 125 43 72 72 9 98 L* 0.14 0.34 0.02 n/a 0.44 n/a 0.37 Standard deviation C* 0.11 1.26 0.13 n/a 0.82 n/a 0.10 h 0.33 8.40 0.06 n/a 1.32 n/a 2.54

Finally, Table 4 shows for each paint the color difference to the NCS specication and to the paint chip. The exterior wall turned out lighter than anticipated, the exterior trim somewhat greener, and the door much darker. Deep high gloss paints appear to be particularly difficult to predict, but fortunately the final color was pleasing and integrated in the actual scheme. In conclusion, the NCS color system is a useful tool to design a color scheme for painting a house.
TABLE 4. CIE94 color difference from paint to NCS and to chip NCS specication S 4010Y70R S 0510B70G S 0550Y70R S 0550Y20R S 0505Y70R S 4050R20B S 0000N Paint to NCS E 4 6 2 2 2 15 3 L* -3 2 1 2 -2 14 -2 C* 3 1 0 3 0 -5 3 h 0 -6 3 -1 1 6 0 2038 1401 1854 1804 2022 1033 designer white base Paint chip E 5 2 2 2 1 10 3 L* -4 1 0 1 0 9 3 Paint to chip C* 0 -1 -1 -3 -1 -4 1 h 0 -2 3 2 1 5 0

The big disappointment was the exterior wall color. With L* = 60 and C* = 12, the NCS sample chosen was at the light side of our preference. With L* = 58 and C* = 10 the paint chip was slightly darker and less vivid, so were able to (almost) compromise on the color choice. However, with L* = 62 (and C* = 9) the exterior wall resulted too light. The next darker color from that paint brand would have been 2039, with L* = 40 and C* = 8, which would have been too dark. In retrospect, we should have resorted to a different paint brand, since the store offered a choice.8

HPL199979

Paint selection

25

In fact, Figure 12 illustrates that the spectral reectance curve for this paint is not particularly interesting. The top curve is the reectance of the designer white base before any pigment is added. The chips reectance is translated down at a more or less constant displacement with respect to the applied paint. Adding black pigment would have corrected the paints color, however this would no longer be a standard color and make it harder to buy a matching paint later. We have to admit that for this paint we have failed our goal of selecting a paint with a spectral reectance curve that get strongly modulated by available reected light.
reflectance 1.0

0.8

0.6

Designer white base 2038 paint 2038 chip

0.4

0.2

0.0 400 450 500 550 600 650

wavelength [nm] 700

FIGURE 12. Spectral reectance curves for the designer white base paint used on ceilings and for the exterior wall, both chip and paint.

4.4 Matching by computer

Unintentionally, we were able to test also the usefulness of the computerized color matching offered by most paint stores. When we ordered new book cases for the study, we specied the exact brand and color number to the cabinet maker, explaining why we did not want a substitute brand or paint type. Unfortunately the cabinet maker had too much faith in computers. Although he swore he used the specied paint, the match is not good. Colorimetry reveals that although the lightness and the chroma are matched quite accurately,
8. Since this color choice was critical, we might also have fallen victim to the effect noted by Fridell Anter that the perceived color tends to have less blackness than the inherent color [6].

House painting with NCS in the USA

26

Paint selection

there is a hue error of 6 CIELAB units in the direction of blue and away from green. While compared to the NCS sample the original paint is slightly too greenish, the matched paint is decidedly too bluish. The graph in Figure 13 reveals exactly what might have happened. The paint appears to have been matched from a chip of the specied paint. The algorithm commonly used in these systems is to measure the specimen under a number of illuminants. The computer then uses a database with the standard recipes of the paint manufacturer providing the matching system to nd a recipe that closely matches the sample under the illuminants. Only if no close match can be found, the system computes a new recipe.
reflectance 1.0

0.8

0.6

chip original paint matched paint

0.4

0.2

0.0 400 450 500 550 600 650

wavelength [nm] 700

FIGURE 13. Spectral reectance curves of a paint, its computer match, and a chip of the rst paint.

The original paint had a spectrum different from its corresponding chip, and the color matching computer was fed an incorrect input. By chance, the system had a standard recipe close to that used for the original manufacturers paint chip and it inferred a match. Indeed, the spectral distribution of the original paint is even better than the chip, enhancing the light modulation. Since the study has four doors painted in the trim color, this creates a very annoying clash of colors. While the trim color is a delicate light blue with a hint of green that is modulated by the reections from the vermilion walls and the light from windows on two sides, the color of the book cases is just a plain and boring baby blue. The tungsten light in the evening accentuates this total appearance.

HPL199979

Conclusions

27

5.0 Conclusions
We have been able to successfully use the NCS color atlas and index to design the color scheme for a house, encompassing both the exterior and the interior. Using an atlas has allowed us to design a scheme representing a strong dynamic equilibrium. USA manufacturers offer a sufciently good selection of standard paint colors that a color scheme specied with NCS can be accurately realized with standard paints. However there are differences in the spectral compositions of the paint chips and the paints themselves, so that colors should not be computer-matched from paint chips. It is still necessary to test a paint color by applying it to an area at least a square meter in size, because this is the only way to appreciate the total appearance of a color. However, the atlas simplies the choice of an alternate color if the rst color is not satisfactory. Color is modulated by incident light, which in the real world is highly variegate. An interesting spectral distribution in the paint will lead to an interesting wall, that is modulated by the available light. It might be interesting to create a tool that can predict this effect. We tried to compare the colorimetry at illuminant A with the colorimetry at illuminant D65 for the paints in our scheme, but that was not a good predictor of the nal effect. It is probably necessary to convolve the reectance curves with a number of incident spectral power distributions and study the change of the reected spectral power distributions. The difculty in this is that the incident spectral power depends upon the ambient.

6.0 References
1. M. Akita, Colors in Public Usage: Preliminary Studies in the Historic City of Kyoto, Proc. 8th Congress of the International Colour Association, Kyoto (Japan), May 2530 1997, II, 925928. 2. G.B. Beretta, Color Palette Selection Tools, Proceedings SPSEs 43rd Annual Conference, 9496, May 2025, 1990, Rochester, New York. 3. J.L. Caivano, Color and Semiotics: A Two-way Street, Color Research and Applications, 23, 6, 390401, December 1998. 4. CIE 116, Industrial colour-difference evaluation, International Commission on Illumination, 1995. 5. K. Fridell Anter and . Svedmyr, Colour Scales of traditional pigments for external painting, Edition 2, Scandinavian Colour Institute AB, Stockholm, 1996. 6. K. Fridell Anter, Inherent and Perceived Colour in Exterior Architecture, Proc. 8th Congress of the International Colour Association, Kyoto (Japan), May 2530 1997, II, 897900. 7. A. Hrd, NCS: A descriptive colour order and scaling system with application for environmental design, Man-Environment Systems, 5, 3, 161167, 1975.

House painting with NCS in the USA

28

References

8. R.W.G. Hunt, Measuring Colour, 2nd edition, Ellis Horwood, Chichester (UK), 1991. 9. K. Kengo, Color to Make Architectural Forms Stand Out or Fade Away, Nipponia, 4, 68, 1998, http://jin.jcic.or.jp/nipponia/nipponia4/pdf/p10-1.pdf and *-2.pdf 10. S. Kobayashi, Color Image Scale, Kodansha International, Tokyo, 1991. 11. J.-P. Lenclos, The Geography of Colour, Proc. 8th Congress of the International Colour Association, Kyoto (Japan), May 2530 1997, II, 887888. 12. F.H. Mahnke and R.H. Mahnke, Color and Light in Man-made Environments, Van Nostrand Reinhold, New York, 1987. 13. A. Nemcsics, Colour Dynamics, Ellis Horwood, Chichester (UK), 1993. 14. S. Prieto, The Color Consultant: A New Professional Serving the Architecture Today in France, Color Research and Applications, 20, 1, 417, February 1995. 15. R. Sve, New Formula for the Computation of CIE 1976 Hue Difference, in Color Research and Application, 16, 3, 217218, June 1991. 16. L. Sivik and A. Hrd, Some Reections on Studying Colour Combinations, Color Research and Applications, 19, 4, 286295, August 1994. 17. W. Spillmann, Color Order Systems and Architectural Color Design, Color Research and Applications, 10, 1, 511, Spring 1985. 18. G. Tonnquist, Philosophy of Perceptive Color Order Systems, Color Research and Applications, 11, 1, 5155, Spring 1986. 19. Th.F. Tosca, Dreams of Light for the City, Color Research and Applications, 19, 3, 155170, June 1994.

HPL199979