Assessing Light in Architecture : a Numerical Procedure for a Qualitative and Quantitative Analysis

International Lighting Conference, Associazione Italiana di Illuminazione (AIDI),CIE (Commission internationale de l’Éclairage), Venise, 9-10 Octobre 2006.

ASSESSING LIGHT IN ARCHITECTURE:A NUMERICAL PROCEDURE FOR

A QUALITATIVE AND QUANTITATIVE ANALYSIS

Claude M.H. Demers (1)

1. INTRODUCTIONLight is perhaps one of the most decisive attributes of a successful building and has yet to be fullyunderstood by architects. There are interesting design tools available to architects and designersassisting them in the design process, but these need considerable time to master or understand,whereas the method of image analysis proposes an intuitive approach with considerable advantages.The aesthetic response of light in space depends on numerous factors that should be included in thequalitative analysis of a lighting system. Qualitative analysis normally involves a process thatincorporates a large part of subjectivity and therefore, results become difficult to comment or discussamongst professionals. In that view, it appears important to integrate a method that allows theintegration of a quantitative component in the qualitative assessment of light in space. Quantifiableaspects of light that relate to the definition of space can be identified in the analysis of a digital image.The digital method of analysis presented in this abstract has been developed since 1993 [Demers]and applied on research consultancy for the design of several buildings such as the New CanadianEmbassy in Berlin and more recently in the design of the CIFSS Pavilion at Laval University. Itproposes a qualitative interpretation of lighting patterns on surfaces, a fundamental relation betweenlight and architecture. Secondly, a quantitative interpretation of lighting patterns is introduced andbecomes most relevant as it allows architects to compare different design solutions quite effectivelywithout entering into intricate calculations at the early design stages. The present research showsapplications of the methodology to real case studies in naturally and artificially lit spaces,experimenting with several lighting conditions from high to low contrast to evaluate the limits of theimage as a tool of analysis. Benefits of using such a method reside in its readily availability to allpractitioners and students at a very low cost and ease of interpretation. Interest in the methodologyinvolving this digital analysis of images has also been shown in subsequent researches by Demers(2004), as well as by Siret (1997) and Bracarensi et. al. (2002). The method has also beensuccessively used amongst architectural design students as a mean to interpret design alternativesprimarily on the qualitative aspects (Dubois, 2006).

2. METHOD OF ANALYSISArchitecturally, contrasts of light and shade define surfaces and contribute to the character of a space,a question related not only to light and shade patterns, but also on the nature of the different sourcesof light and surrounding surfaces. This section demonstrates components of the digital image that canbe used to produce an assessment of light in space.

1 M.Arch. Ph.D. (Cambridge). Associate Professor, School of Architecture at Laval University, Groupe

de Recherches en Ambiances Physiques (GRAP), Faculté d’aménagement, d’architecture et des artsvisuels, Vieux Séminaire, Université Laval, Québec, G1K 7P4, Canada. [email protected]


An image, when properly exposed, is “a direct record of the amounts of light to which each area hasbeen exposed” [Evans, 1959, p. 77]. The image therefore constitutes the basis of the following methodof acquiring lighting data in architecture. Digital photography has been the subject of importantdevelopments and improvements during the past decade, offering support for recording lightingpatterns in space and making it an ideal tool for research and practice. Integrated pose-meters, arenow more versatile and offer more precision in the calculation of the right exposure of the subject inthe view field of the camera. New digital cameras integrate a multi-reading pose-meter that grantsmore precision and flexibility in the evaluation of the correct exposure of photographs. The ideal digitalcamera used for acquiring visual data should therefore be chosen in relation to the level of desiredpreciseness for the study. Digital images of existing spaces are obtained according to an establishedprocedure [Demers 1997] to ensure validation of the analysis. It is important to relate to the procedureas it determines the level of lighting obtained in the images. For instance, as the eye adapts todifferent lighting levels, the pose-meter of the camera needs to consider the same adjustment inrelation to different lighting conditions. Whereas many scenes can be obtained from automaticexposures, images obtained in backlighting offer the most critical conditions as the photographer needto select the right exposure and even work with manual controls. Reflected light from mirrors and otherhigh reflectance surfaces can also affect the correct exposure. The preview of an image on the screenof the digital camera should therefore always be performed on site to verify the correspondencebetween the actual lighting condition and the recorded image.

Images are subsequently processed within the Adobe Photoshop software for digital images combinedwith plug-ins from Reindeer Graphics according to the methodology [Demers 2002] that allows severallevels of evaluation. Whereas chromatic contrast is subjectively assessed on people, intensity contrastremains more appropriate for computer analysis as it refers to the image brightness and is thereforeused in the research. The color information content of the digital images is therefore discarded,leaving grey scale pixel values that provide information on light and contrast. Previous validations[Demers 1997] carried out in an artificial sky using a physical model correlated photometric andphotographic measurements in an average interior lighting condition, but current research shows thatresults obtained in an existing artificially lit space using photocells as a measure of illumination levelsin comparison with grey level scales are also representative. Assuming a certain correspondence, thisresearch expands the use of lighting patterns in the analysis of a lighting condition. Several types ofpatterns can be extracted and used according to the character of the analysis such as the level ofdetail and the dominant pattern. The following sections explore some of the most interesting patternsand their potential to assist architects and designers in their analysis of light in space.

2.1 Level of detailThe pattern of the level of detail relates to the attention needed to “read” or respond to visual stimuli ofa space. Light has the ability to emphasize textures, details and structural features of a space,producing sometimes intricate patterns in architecture, and sometimes more minimalist and uniformcompositions. These attributes are important in the elaboration of an aesthetic description of a space.Color and luminance patterns of different surfaces are also responsible for interaction with light andthe perception of visual details in architecture. In the digital method of analysis, the level of detail isquantified as the overall percentage of contour lines contained in a scene in relation to the entire viewframe of the image. The contour filter should be applied to the grey scale digital image. This descriptorallows a discussion and comparison of images issued from spaces of different definitions of detail(figure 1). Figure 1a represents the highest level of detail with a line density of 36% amongst thepresented examples. The image corresponds to the view of strongly projected shadows of sunlight onthe already textured wall of the space. The image also highlights the intricate patterns of shades andtextures creating unexpected grids. The quasi-absence of lines in the vertical rectangle of the sameimage also emphasizes its visual presence, competing with the surrounding highly textured surfaces.Figure 1d presents the lowest level of detail with a 7% value. The contours lines of the chairs defineuniform surfaces that are surrounded by a somewhat delicately defined background details. The lightgrey lines correspond to more subtle definition of details whereas dark and thick lines such as in figure1a and 1b are more related to highly contrasting patterns in space. Surfaces showing a certain level oflighting uniformity therefore contain a lower number of pattern lines.


It is impossible to predict exactly how people would react to a particular scene, but the digital imageanalysis allows the possibility to identify the potential locus of attention, corresponding to the zone ofhigh density that could relate to areas of special interest in the visual field [Austen, 2000]. It is alsoused to refer to the extent, also described as its distribution in the space [Austen, 2000]. Theseanalysis are part of the qualitative definition of light and textures on surfaces and provide order in thediscussion of a high number of spaces or views. It should be acknowledged that the level of detailpattern does not take into account focal accents [Lou, 1996] in the environments such as the presenceof people or other details. Other complementary aspects of space such as acoustics would alsoconsiderably affect the global perception. However, it remains interesting to consider this pattern as acondition that can generate a particular attraction to the viewer. It can therefore be assumed that:

a) In two similar types of images or subjects, the level of attention could be raised as the numberof lines increases, creating interest for the eye.

b) Images containing a combination of surfaces of a high number of lines (visually textured) anda low number of lines (visually more uniform) are producing a high contrast of light-textures.These have the potential to provide areas of rest for the eye, defining a certain composition ofspace and light-textures.

Lighting patterns in artificially lit environments are usually constant whereas in daylit spaces, they aredependant of the amount of light and direction occurring on a surface. The latter type of patternsshould ideally be photographed under uniform and clear skies, with several sun angles when possiblefor a more comprehensive and accurate analysis of a lighting effect.

a) b) c) d)

Figure 1: Level of detail: digital images are in a decreased order or detail content.

2.2 Location of the dominant patternHopkinson [1969] acknowledges the phototropic nature of the eye, which accounts for our beingattracted or even sometimes diverted by sources of light of a high brightness in the visual field. Thelocation of bright patterns on the digital image allows the establishing of the physical relation betweenspace and light, essential to the perception of the visual environment. The quantification of thisvariable is established by applying the Adobe Photoshop “stamp” filter (figure 2) on a grey scale digital


image by using mid-scales of clear/dark and uniformity settings of the software of image analysis.Other settings could be applied, but it was found that where the extremes of these scales werechosen, the resulting images were either too bold or too precise to produce any conclusive analysis. Itis important to apply the same settings to each image to ensure correspondence of the results ofanalysis. Figure 2 presents an application of the filter on the same images that were previouslyanalyzed (figure 1). The filter ensures that only the brightest areas of the images are highlighted andboldly represented in black and white. The second row of smaller images identifies these zones ofhigh brightness for each example. This analysis evacuates all information related to gradation of lightsand detail, but provides a general overview of the location of the brighter sources of light in the space.Figure 2a presents a rather uniform brightness pattern that is slightly more densely at the top of thespace than in the lower parts. It also emphasizes the dark vertical rectangle of the window area of thewall as a light absorbent surface. Figure 2b shows an asymmetric composition where the upper left ofthe space corresponds to the brighter area whereas a series of converging architectural elementshighlighted by light underline the presence of horizontal reflectors associated with each balconycirculation. Figures 2c and 2d show a central occupation of a more precisely defined brighter zonesurrounded by darker areas. The vertical dimension of light presented in figure 2c associates with thehigh lateral window of the atrium space that admits sunlight whereas the horizontal repartition of lightin emphasized in figure 2d.

a) b) c) d)

Figure 2: Location of the dominant lighting pattern.

Superposition of patterns related to the level of detail (section 2.1) and the brightness location (section2.2) sometimes occur in the same zone. Figure 2d is an example of such superposition as thehorizontal central occupation of the brighter part of the image also corresponds to the zone of higherdensity of details. These combinations are interesting as they indicate whereas the brighter zones andlighting-pattern are acting separately of acting together, reinforcing a specific location of the space.

ScatteredEntire frame

DominanceUpper left

HorizontallyCentered

VerticallyCentered


2.3 Gradation of lightGradations from light to dark can be interpreted from the pattern as an indicator of contrast on certainareas of a space. The main direction of the pattern and the relative measurements between thebrightness zones on that axis become assessors of the gradation. Images are processed andsimplified into five brightness zones (0%, 25%, 50%, 75% and 100%) using the “posterize” commandof the software of image analysis. This five brightness zones system is based on the LuminanceBrightness Rating (LBR), which contains nine grey levels [Lou, 1996]. It is possible to refine the imageanalysis in separating the image in the nine grey levels (0%, 12,5%, 25%, 37,5%, 50%, etc…). Thisprocess adds only four new brightness zones between the existing ones, but the transitions betweenzones are generally too smooth and do not reveal as much the global pattern. The nine brightnessgrey level separation should therefore mainly relate to scenes of a very low contrast.

The original image of figure 3a is processed in these five brightness levels (figure 3b). It is possible tocalculate the distance between the limits of different brightness zones and use a comparative scale toevaluate the spreading of the patterns. The pattern of figure 3b is obtained when using the contourfilter. A coefficient of gradation [Demers, 1997] is introduced to provide a quantitative interpretation ofthe pattern. The interval between the brightness zones of a pattern is usually calculatedperpendicularly to the main axis of the spreading.

a) b) c)

Figure 3: Linear aperture: gradation of light obtained through brightness separation [from Demers1997].

Images of figure 3 were extracted from a series of experiments performed in the controlledenvironment of a physical model photographed under an artificial sky [Demers 1993; 1997]. For suchcomparative studies, all significant objects of the images should be at the same size to ensure thatcalculation of distances between zones is relevant for a certain space or type of light. For instance, theimage of the white wall (figure 3) was always of the same as apertures were changing sizes andpositions. These parametrical studies enabled a more theoretical understanding of lighting patterns inrelation to aperture configuration and position, as well a surface reflectance. This knowledge is nowapplied in actual spaces such as in figure 4. In this particular case, the five brightness levels arepresent in the scene, indicating a high contrast particularly between the top of the chairs and thebackground windows. This is clearly represented in the absence of transition between the 0% (blackchairs) and the 100% zones (bright windows). The patterns discussed earlier in section 2.1 (level ofdetail) in the window area of the central image are in the 100% and 75% brightness, indicating arelatively low contrast definition of the pattern. There is also a rather important definition of the ceilingsurface in relation to the background windows, also indicative of high contrast. Images have thepotential to identify zones of the visual field responsible for glare. High contrast at the eye level of thescene of figure 4 has the potential to produce glare for an observer situated in the room at the locationof the photographer.

x

0,15x

0,21x

0,19x

0,34x

100%

75%

50%

25%

0%


Figure 4: Brightness separation of image d (in figures 1 and 2).

2.4 General brightness of the imagePerhaps the most striking character of a space comes from its dark versus bright ambience. It ispossible to measure the lighting levels of a space and compare it with other spaces, but this can be along process and need specific meters that one does not necessarily carry all the time, especiallywhen unexpectedly visiting a building. The present type of digital analysis is practical to compare anddiscuss different situations.

Histograms are used to analyse the distribution of pixels of certain brightness on an image (figure 5,top right). When using the separation of brightness (section 2.3), it is possible to obtain a robusthistogram of its brightness content, enabling a more global assessment of the grey level dominanceand compare it to other images. Figure 5 shows the relative quantity of pixels for each of the fivelevels of the grey scale for the four images previously analysed. The histograms are translated intohorizontal bars corresponding to the five brightness separations discussed in section 2.3. Thehorizontal axis represents the proportion of pixels of a particular brightness on the image. The 100%brightness (white) is relatively constant for all four images, representing an average of 20% of the totalamount of pixels (from 15% to 25%). The 0% brightness (black) is much more variable from oneimage to another, from about 20% for the atrium to nearly 60% for the office and therefore affect moreconsiderably the character of the analysed spaces in terms of lighting ambiences. The three mid-brightness zones (25%, 50% and 75%) can be regarded as a group in this particular example. Theproportion of mid-brightness greys increases from the office (top image, about 25%) to the atrium(bottom image, nearly 60%). In this group, the lower and mid-brightness pixels are more relevant todiscuss the overall brightness of the image.

0%

100%

25%

75%

50%

25%


Figure 5: Comparison of the distribution of lighting zones in images.

3. CONCLUSIONThe digital image allows the comparison of several lighting conditions in a systematic manner. Itprovides also a relation between patterns of light and the composition of space in a language that ismeaningful to architecture. Four types of digital analysis have been presented in this research:

1 Level of detail2. Location of the dominant pattern3. Gradation of light4. General brightness of the image

These types can be used independently, or compared one to another. Previous research has showncorrelation between photometric and photographic measurements, especially in the mid-brightnessgrey scales [Demers, 1997]. Although not extensive, the study reinforces the potential for digitalimages to support more conventional measurement techniques.

The variables of digital imagery refer to physical aspects of light in space and therefore afford anartistic insight and potential descriptors of the lighting patterns in space. The results can bequalitatively and quantitatively interpreted, ensuring a complementary analysis of light on surfaces.The digital camera and the method of computer image analysis constitute the main tools for acquiringand analyzing lighting conditions. These tools have the advantage of being already used byprofessionals and students of architecture and therefore are available at any time. When performingthese analysis, researchers and architects should always bear in mind that the image and the viewframe of the camera is only a part of an actual space frozen in time. Although visual perception is verycomplex, the patterns developed in this research will eventually be part of a series of other patternsissued from other complementary aspects.

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

atrium

exchange

tranchée

office

proportion of zone per image brightness level (%)

0%

25%

50%

75%

100%


REFERENCES

AUSTEN, Erin & James T. Enns, (2000), Change Detection: Paying Attention To Detail, PSYCHE,6(11), October 2000.

BRACARENSE, Viana Junior, Da Silva Jota, Sad De Assis, (2002), “Public building lightingperformance analysis”, PLEA 2002 (Passive Low Energy in Architecture): Design with theenvironment, Proceedings of the 19th International Conference, Toulouse, France, July, p. 821-825.

DEMERS, Claude M.H., (1993), “Architecture de lumière: le contraste générateur d’ambiances” inQuébec ‘93, Energy efficiency and solar energy: Proceeding of the 19th Annual Conference of theSolar Energy Society of Canada, The Solar Energy Society of Canada Inc., July 9-13 1993, Québec,Canada, pp. I-22 - I-27.

DEMERS, C. M.H., (1997), The Sanctuary of Art: images in the assessment and design of light inarchitecture, PhD thesis, University of Cambridge, Emmanuel College, Angleterre.

DUBOIS, Marie-Claude, (2006), Integration of daylight quality in the design studio:from research to practice, Proceedings of the PLEA (Passive Low Energy in Architecture) 2006Conference, Geneva, Switzerland, 6-8 September 2006.

EVANS, Ralph M., (1959), “Eye, film, and camera in colour photography”, New York, John Wiley &Sons.

HOPKINSON, R.G., and J.D. Kay, (1969), The lighting of buildings, Faber and Faber, London.

LOU, Michel, (1996), Light: The shape of space / Designing with space ans light, Van NostrandReinhold, New York.

SIRET, D., “Proposition pour une approche declarative des ambiances dans le projet architectural:application à l’ensoleillement”, PhD thesis, (1997), Université de Nantes, France.

SCHILER, Marc, (2000), “Towards a Definition of Glare: Can qualitative issues be quantified?”, 2nd

EAAE-ARCC Conference on Architectural Research, July 4-8, Paris, France.

Assessing Light in Architecture : a Numerical Procedure for a Qualitative and Quantitative Analysis

Documents

Transcript of Assessing Light in Architecture : a Numerical Procedure for a Qualitative and Quantitative Analysis