Why digital enhancement of rock paintings works: rescaling and saturating colours.

David, Bruno ; Brayer, John ; McNiven, Ian J. 等

Introduction

Rock paintings are found on all continents except Antarctica. Yet rock-art research is faced with a very real problem. Like other items of material culture, it is subject to taphonomic processes and various forms of destruction over the long term. Such problems include granular disintegration of decorated rock surfaces; varied biological problems such as lichen, algal or fungal growths, insect nests and the rubbing of cattle against rock walls; water damage; graffiti; and the accumulation of dust mantles and other mineral crusts over rock surfaces. The result is an increasing visual obstruction of the art through time. These problems are so acute that significant proportions of pictographs in many art sites around the world remain too unclear to record adequately, their forms often being relegated to the general category `indeterminate'. In Australia, this problem is well illustrated by various recording programmes in the tropical north. On the granite boulders of Torres Strait, exactly 50% of pictographs were identified as indeterminate by McNiven et al. (2001); and in Australia's Cape York Peninsula, 39% are indeterminate on the granites near Yarrabah (David 1994); 30% on the granite boulders of Bonney Glen Station (David 1991); 22% in the Mitchell-Palmer limestone zone (David, unpublished data); and 10% on the Ngarrabullgan sandstones and conglomerates (David 1992). In the vast majority of cases, it is fading (caused by a variety of processes, see below) that has rendered the motif forms indeterminate.

Digital image enhancement technology has been well known since it was developed for space exploration in the 1960s and 1970s (e.g. Castleman 1996; Pratt 1978). Numerous archaeologists are now using digital enhancement techniques to help them better see the rock paintings that are or were once obvious on rock walls (e.g. Clogg & Diaz-Andreu 2000; Henderson 1995; Mark & Billo 2000; Read & Chippindale 2000). Relatively cheap yet powerful computer enhancement programs such as Adobe Photoshop or Corel Photo-Paint are now easily available over the counter, ideal for rock-art applications. Yet few archaeologists understand why enhancement works -- what happens when an image is enhanced -- or how far enhancement can be taken. This paper aims to rectify the situation by explaining why digital enhancement of rock paintings works. We use one of our own case-studies from the islands of Torres Strait, between Australia and New Guinea, to illustrate two principles of image enhancement -- colour rescaling and manipulation of colour saturation -- to increase the visibility of rock paintings to the human eye.

Measuring the colour of rock images

What is light?

Light travels through space as electromagnetic waves of varying lengths. The entire spectrum of wavelengths is called the electromagnetic spectrum, but only a small section of this spectrum is capable of producing visual sensation (Jacbson et al. 1988: 8). Other parts of the spectrum include radio waves and infra-red waves that have wavelengths longer than visible red light, and ultra-violet light that has wavelengths shorter than visible violet light. Despite some experimental efforts to use these non-visible wavelengths of light to record rock-art images, results have been mixed and most recording is accomplished using visible light.

What is colour?

The human retina contains millions of light-sensitive cells that can be divided into two major types, rods and cones. Through variations in chemical composition, three types of cones enable us to discriminate among different colours.

The light that reflects off any surface (e.g. a rock background or a painting) consists of a `rainbow' or spectrum of electromagnetic waves of different wavelength (e.g. red, yellow, green, cyan, blue, magenta). These surfaces differ in colour when the reflected light has different amounts of light intensity at different wavelengths (FIGURE 1). A red surface has more intensity in the red wavelengths. A blue surface has more intensity in the blue wavelengths.

[FIGURE 1 OMITTED]

How is colour measured?

A digital camera records the light reflected from a rock surface on a sensor consisting of a rectangular array of about 500x500 pixels (or picture elements). Each pixel represents a tiny square area of the image, that is, the light reflected from a region on the surface of the rock. For a colour camera, each pixel is really three different sensors, each sensitive to a respective colour, called red, green and blue (R, G and B). Three sensors are considered adequate because they are roughly equivalent to the way the human eye perceives colour using its three types of cones. So three numbers, the R, G and B intensities determine the colour at a point or pixel in the image.

What is colour space?

Each pixel colour, represented by the three R, G and B numbers or coordinates, can be considered a point in a three-dimensional `colour space' (FIGURE 2). Colour intensities range between 0 and 1 (or, in the computer, 0-255, because they are stored as unsigned integers in an 8-bit byte) and the colours at the 8 corners of the `colour cube' are the `primaries', Red, Green and Blue, the `secondaries', Cyan, Magenta and Yellow, and finally Black and White. Virtually any colour can be generated from a combination of the three primaries. The line from Black to White along the diagonal of the colour cube is called the `grey line' and consists of all shades of grey from black at the origin to white at the far diagonal corner. Two colours near-by each other in colour space will appear similar. Points in colour space near the grey line appear greyish or pastel or unsaturated. The less grey is contained within a colour, the more that colour is said to be `saturated'. So colours near the three primary corners of the colour cube appear deep and pure and are called saturated. This is a useful point to remember when digitally enhancing colour images, for minimizing the proportion of grey within selected colours may increase the purity and thus the distinctiveness of the targeted colours. A non-selective saturation of all colours in an image may bring out the distinctiveness or contrast of all coloured pixels and thus coloured regions within the image. And likewise, maximizing the amount of grey can desaturate background colours on a surface and make it appear more uniform.

[FIGURE 2 OMITTED]

What are some of the ways to represent a colour in colour space?

There are several equivalent ways to represent colours. These ways are essentially different coordinate systems for colour space analogous to cylindrical, conical or spherical coordinates for 3D positions in surveying. Three of the more important colour systems are discussed below.

First, in the RGB system, we use the measured coordinates (R, G, B) directly (Jain 1989). The resulting colour coordinates form a rectangular coordinate system and the colour space is a cube as shown in FIGURE 2. These coordinates correspond directly to the physical light measurements made by the camera. They are easy to use and to understand. However, they are not always the most intuitive to use for certain types of colour manipulation and enhancement.

Second, in the HSB (or HSL) system, the coordinates are a 3D conic coordinate system for 3D colour space (Jain 1989) (FIGURE 3). H is the `hue', or dominant spectral wavelength of the colour. Hue, H, is an angular coordinate around the circumference of the conic colour space. The hue identifies the dominant wavelength, typically between 400 and 700 nanometres along the electromagnetic spectrum. S is the saturation of the colour, that is, the `purity' of the colour or the extent to which other colours (such as grey) are not mixed into the dominant wavelength. Pure red is saturated, but pink is unsaturated because it has white (a combination of all colours) mixed in. Saturation, S, is a radial coordinate, measuring distance away from the centreline or axis of the conic colour space, called the grey line. B is the `brightness' or lightness (L) or total intensity of the colour. It refers to the total amount of incident light a colour reflects. Brightness, B, measures the distance along the centreline or axis of the cone (the grey line) away from the origin of colour space (The origin is the point representing the colour black). Artists and image enhancers use the (H, S, B) system because the intuitive meaning of the coordinates frequently matches the manipulations they wish to apply.

[FIGURE 3 OMITTED]

Third, the Commission International pour l'Eclairage (CIE) have defined a uniform colour space using the L*a*b* system, in which equal distances correspond to equal perceived colour differences (Pratt 1978). In this system the L* is related to the colour lightness and is conventionally reported on the vertical axis of the 3-D colour space, but is really very similar to the B or L coordinate in the HSB system, i.e. distance along the grey line. The axes a* and b* are each determined by hue and saturation and are reported on the plane orthogonal to L*. Thus brighter colours move towards higher L* values, while increasing saturation moves the point representing colour out from the origin in the a*b* plane. Red and green are characterized by positive and negative values of az, respectively. Yellow and blue are positive and negative values of b*, respectively. A commercially available chromametre is normally used in conservation studies, particularly to monitor degradation of paintings and to measure the surface colours. A white tile is the calibration standard.

What is the colour space representation of 2 differently coloured regions?

Consider FIGURE 7a, a computer-generated, simulated image of a darker grey star foreground figure on a lighter grey background area. As is typical with rock-art, a sample of pixel colours from the background region is frequently similarly coloured, forming a cluster or cloud of points in colour space. A group of pixels selected from the foreground star figure region of the image will also be coloured similarly to one another, forming a cloud or distribution of points in colour space. However, these points should be coloured differently from background pixels. So this cloud of points should be separated in colour space from the cloud of background colours (FIGURE 4).

[FIGURE 4 OMITTED]

What is contrast?

If the colour clouds of different image regions are widely separated in colour space, we say the image has good `contrast'. That is, we have good contrast if the inter-cloud distances in colour space are much greater than the intracloud distances. If the colour clouds of image regions are close together or even overlapping in colour space, we say the image has poor contrast. So contrast is a measure of how spread out in colour space the image colour points are. We usually express contrast in qualitative rather than quantitative terms. So we talk about `good', `poor' or `better' contrast, but usually do not attach a quantitative measure to the amount of contrast.

What is fading?

Rock-art images generally start out having good contrast, but over time, by a variety of taphonomic processes, the images undergo `fading', which tends to reduce the contrast between foreground figures and the background rock. The process of fading tends to move the colour distributions closer to one another and closer to the grey line so that differences may eventually disappear. Sometimes, the colour differences do not totally disappear in an image, but just become not noticeably different.

Photochemical change in rock painting pigments is thought possible by oxidation and subsequent leaching, but for this type of fading to occur the electromagnetic radiation must be absorbed by components in the paint (e.g. Tweedie et al. 1971). Subsidiary factors assisting in this process are temperature and humidity.

Wavelengths in the ultra-violet region of light could have sufficient energy to break molecular bonds that initiate changes in organic binders in paints, but it is unlikely that traditional rock painting pigments, such as clays, iron oxides and charcoal, will be affected by prolonged exposure to sunlight.

Generally, rock paintings can fade over many years because of the slow accumulation of dust and salts. Over time, the fading of paintings results from dust that initially obscures and then eventually masks the figures by forming an indurated opaque crust (e.g. Watchman 1993).

Loss of clarity or fading of paintings also takes place by the slow accumulation of a thin translucent film of amorphous silica (Watchman 1992), particularly on sandstone and granitic rocks. Again the effect is produced by the natural deposition of material directly over the art.

Another weathering condition that produces faded paintings is loss of paint. Granular disintegration or micro-exfoliation of the surface gradually leads to thinning and losses of paint mass, often exposing the underlying rock. Paint composed of pigment and filler and with only water as the medium may gradually fade through exfoliation because of weak cohesion between the fine particles. Cohesion is stronger where a binder exists. Similarly, paint loss will occur where adhesion to the rock surface is poor, such as for clay-based paints. The well-known fugitive nature of clays is the result of their platy structure and potential for large expansion and contraction movements during periods of wetting and drying respectively. Crystallization of salts on the surface and under a paint will also contribute to fading either through loss or masking of paint.

What are the minimum colour differences that can be perceived?

If colours are close to one another in colour space, the differences may not be `noticeable' by a human observer. Colour scientists define the term, `Just Noticeable Difference' to mean the distance apart two colour points must be in colour space for the colours to be just barely noticeably different (FIGURE 5) (Pratt 1978). This is the distance in colour space at which the difference just becomes perceptible by a human viewer. For simplicity, FIGURE 5 shows the region of `Not Noticeably Different' colours as a circle. This would be the case if a `uniform' colour system such as L*a*b* were used. In a `non-uniform' colour system such as RGB, this region would be an ellipse and equal colour distance changes will produce different perceived changes depending on the direction of the change (e.g. toward red or toward green). Thus in a nonuniform colour system such as HSB it is possible to enhance the perceived differences just by rotating the hues of all pixels so that the colour differences are in a more sensitive direction.

[FIGURE 5 OMITTED]

Digital enhancement

How it works

`Contrast enhancement' is a process in which we transform the colours of all the pixels of an image to try to increase the distance in colour space between the colour clouds of the foreground and background pixels. If we do it well, we can convert differences that are not noticeably different into differences that are noticeably different.

It is important to understand that in the type of enhancement discussed here we transform the image `pointwise', by mapping the colour of each and every pixel using the same transformation. The new colour of each pixel is dependent solely on the old colour of the pixel. In no case here do we selectively apply the transforms to portions of the image selected by using human interpretation or automated image recognition.

What are some of the ways to do contrast enhancement?

There are several ways to accomplish colour contrast enhancement. The basic difference between the different ways of enhancement is that they expand the colour differences in different directions in colour space. One way is simply to `re-scale' one of the basic colour coordinates, red, green or blue. This expands or spreads out colour points parallel to that colour axis. FIGURE 6 illustrates expanding colour points along the green direction. For example, in FIGURE 7a we present a simulated rock-art image. The colour space distribution of pixels for this image was shown in FIGURE 4. The green stretching enhancement is applied to this image and the result is shown in FIGURE 7b. The colour distribution of pixels for the FIGURE 7b image is seen in FIGURE 8. Notice how the colour points are spread out along the vertical (green) direction. We could obtain similar results by rescaling along the red or blue directions. By rescaling all colours simultaneously, we rescale along the grey line, which is equivalent to brightness contrast enhancement.

[FIGURES 6-8 OMITTED]

Another way to enhance an image uses hue, saturation and brightness (HSB). As saturation represents the distance of a colour from the centre axis of colour space (the grey line), increasing the saturation of colours moves the colours proportionately radially away from the grey line, increasing the distances between colours (FIGURE 9). We can illustrate this effect using an actual rock-art image from Torres Strait (FIGURE 10a). Sixty pixels were selected from small regions of the foreground and background of the image. The red and green colour components (in the range of 0-255) of these pixels are plotted in a scattergram (FIGURE 11). This plot is the same information as in FIGURE 4 except data from the real image of FIGURE 10a is used.

[FIGURES 9-11 OMITTED]

The background pixel colours in this image are clustered slightly above (greener than) the diagonal (the grey line, or in this red-green projection, the yellow line). The foreground pixel colours are clustered lower and to the right in the red direction. But these distribution clouds are very elongated and very close to one another. Increasing the saturation will make the slightly yellowish background pixels more yellowish, the slightly bluish background pixels more bluish and will make the slightly reddish foreground pixels more reddish, moving all the pixel colours away from the grey line (lower-left to upper-right diagonal of the plot). This increases the colour distances along the upper-left to lower-right direction and thus spreads the perceived differences between the clouds. The results on the colour distributions of increasing the saturation moderately (about 70%) are shown (FIGURE 12). Notice that the pixel colours indeed are more spread in a direction perpendicular to the grey line. The corresponding enhanced image is shown in FIGURE 10b.

[FIGURE 12 OMITTED]

These two techniques, increasing contrast by colour rescaling and altering the saturation of colours, are usually undertaken manually. There are also sophisticated statistical techniques that compute the centres and shapes of the foreground and background clouds and then rescale all the image colours to maximize some mathematical distance between the clouds.

How can researchers get started using contrast enhancement?

Below is a very brief list of steps a researcher can follow to begin using image enhancement.

1 Obtain and install an image editor such as Adobe Photoshop.

2 Scan or import images into the computer.

3 In the image editor, adjust the brightness and contrast of each image so it uses the whole dynamic range of brightness values.

4 Do a 100% saturation of all colours to assess hue distribution and increase colour distances (often a lower saturation level, typically 70-85%, works best as the outlier background colours intrude less onto the paint colours).

5 Successively and selectively apply saturation to only the perceived foreground colours (e.g. reds) then desaturation to only the background colours.

6 Rotate selected hues so that differences `appear' maximum.

7 Repeat steps 5 & 6 until no improvement.

8 Experiment with other enhancements available in the editor.

None of the above methods is inherently better than another, and indeed various combinations often prove to produce the best results. A method's performance depends upon the particular image being enhanced and its particular colour cloud distribution. The performance is usually evaluated by human perception so the different enhancement methods must be evaluated and modified interactively. For these and other reasons, simple methods such as colour rescaling and enhancing colour saturation tend to be used more frequently than other, more complex methods. The above list is a very brief potential starting point, and broad experimentation is always recommended. But we must also always be aware that enhancement can create erroneous artefacts. It is also imperative that any published enhancements be fully documented in the literature.

Enhancing the rock-art of Torres Strait

Let us now apply these principles to one rock-art site in Torres Strait.

Torres Strait is a 150-kin wide watery realm separating the Australian and New Guinean mainlands. It is home to `Torres Strait Islanders' who, like their ancestors, harvest their seas for fish, turtle and dugong. On some islands, population densities of over 100 people per sq. km existed at the time of initial European contact, which is an order of magnitude greater than the highest densities documented for Aboriginal Australia. Globally, Torres Strait is most famous as the place where the Melanesian and Australian (Aboriginal) cultural and ecological domains meet and as a transition zone between the horticultural and hunter-gatherer worlds (Harris 1977). Coring of reefs and islands reveals the Straits were established 8000-7000 years ago and that island formation is ongoing (e.g. Barham 2000). Thus, Islander society must have developed within this period and functioned as a bridge and barrier for diffusion of cultural traits (and flora/fauna) between northeast Australia and Melanesia.

Little is presently known of the nature of interaction between Australia and New Guinea, via Torres Strait, in prehistory. One avenue of research is stone tool exchange (McNiven 1998) while another is through investigation of spatial and temporal trends in artistic conventions. Unfortunately, much of Torres Strait's rock paintings are faded, and in many instances only the most recent artistic phases appear to have survived. In some cases, however, bare traces of earlier art, sometimes superimposed by more recent pictographs, are apparent, although the shapes of individual motifs are too obscure to be determined with the naked eye. In such cases, digital enhancement offers promise for the recovery of images. We illustrate here such recovery through the selective enhancement of colour saturation and colour rescaling at the site of Kabadul Kula on the island of Dauan (McNiven et al. 2001).

Kabadul Kula

Oral tradition indicates that the most recent phase of rock painting at Kabadul Kula likely took place during a head-hunting raid from southern Papua in the 19th century AD. The paintings then undertaken appear relatively clear, but faint traces of earlier artistic episodes are barely visible at the site today. The motifs purportedly associated with the most recent phase have obvious associations with southern Papuan artistic conventions, including clan designs. A vexing question in Torres Strait prehistory has long concerned the antiquity and nature of interactions with Papuans, and towards this end we were curious to identify the more faded and apparently older motifs at Kabadul Kula.

The site was systematically photographed with two Nikon Coolpix 950 digital cameras by the authors. Sixteen out of a total of 44 pictographs at Kabadul Kula were clear enough to the naked eye to make out their motif forms. Seven previously unknown, faded motifs were revealed by rescaling and saturating colours (16% of the art at the site). At least some of these faded motifs appear to represent the earliest artistic phase(s) at the site, the timing of which we aim to further investigate through forthcoming excavations (to obtain discarded pieces of ochre datable through stratigraphic association). The seven newly revealed images include a `face' or `mask' and an adjacent shield reminiscent of Papuan forms (shields are not known from Torres Strait ethnographically; this image may therefore signal Papuan cultural influences predating the ethnographic situation) (McNiven et al. 2000: figure 2); an anthropomorph and a fish head-dress akin to those recorded by Haddon (e.g. 1912) during the Cambridge Anthropological Expedition to the Torres Straits in 1898 (FIGURES 13 & 14); an insect-like figure (FIGURES 10a, 10b, 15); a face or mask following Papuan conventions; and what appears to be a claw-sail canoe typical of the region to the immediate east. None of the underlying art is reminiscent of Austronesian conventions, so common today as during much of the late Holocene in Island Melanesia to the east and northeastern Indonesia to the west. The implication of the rock-art conventions evident in Torres Strait's faded paintings is that Papuan influences in Torres Strait, to the exclusion of Austronesian influences, may have some antiquity, in support of recent models of Islander prehistory recently formulated by Tony Barham (2000).

[FIGURES 13-14 OMITTED]

The faded rock-art of Kabadul Kula was revealed using various methods of digital enhancement; we illustrate here how and why computer enhancement worked on one of these paintings, following the principles discussed above.

Further attempts to enhance FIGURE 10A were made. FIGURE 15 shows the image after saturating all hues by 100%, and rotating hues -12 along the colour wheel (i.e. all hues were shifted, the red hues now showing greater contrast to the background). The most noticeable effect here is caused by a general saturation of both foreground and background pixels. The rotation of all hues by -12 has the effect of shifting the foreground pixels, which appear yellowish after the first saturation, to more reddish-magenta hues. Thus the hue shift enhances the perceived contrast without affecting the mathematical colour distances between foreground and background. Other enhancements were tried by us and also worked well; the one presented here is simple to perform, just one amongst many possibilities. It is by understanding the effect of such enhancement steps on the colour distributions in colour space that the end-product could be appropriately interpreted, with our research programme in mind.

[FIGURE 15 OMITTED]

Conclusion

The florescence of rock-art research that began in the 1980s continues into the 21st century, not least due to the emergence of new dating and digital enhancement techniques enabling its characterization in time and space. The wide range of environmental and technical contexts of rock-art production, including the geology of the rock surface, sources of erosion, human pressures on the art's survival, rock-art techniques and pigment types will always mean that the best digital enhancement techniques for any given situation will require a combination of general principles and unique address.

Acknowledgements. We thank Gary Swinton (Geography & Environmental Science, Monash University) for drafting the figures; the University of Melbourne for a grant to help fund colour reproductions; the Australian Research Council for funding; the Australian Institute of Aboriginal and Torres Strait Islander Studies for a grant to record Kabadul Kula; the Dauan Island community for supporting our research; two anonymous referees; and BD thanks Monash University for a Logan Fellowship.

References

ANGEL, E. 1997. Interactive computer graphics: A top-down approach with OpenGL. Sydney: Addison-Wesley.

BARHAM, A.J. 2000. Late Holocene maritime societies in the Torres Strait Islands, northern Australia -- cultural arrival or cultural emergence? in S. O'Connor & P. Veth (ed.), East of Wallace's Line: modern Quaternary research in Southeast Asia: 223-314. Rotterdam: A.A. Balkema.

CASTLEMAN, K.R. 1996. Digital image processing. Upper Saddle River (NJ): Prentice Hall.

CLOGG, P. & M. D~AZ-ANDREU. 2000. Digital image processing and the recording of rock art, Journal of Archaeological Science 27: 837-43.

DAVID, B. 1991. Preliminary report on Aboriginal rock art sites on Bonney Glen Station, southeast Cape York Peninsula. Report to Wujal Wujal Community Council, Wujal Wujal.

1992. Archaeological investigations at Nurrabullgin: the Mt. Mulligan Project Stage 1. Report to the Australian Institute of Aboriginal and Torres Strait Islander Studies, Canberra.

1994. The Trinity Inlet Ethnographic Study: planning the management of traditional Yirrganydji, Yidinji and Gunggandji country. Report to the Trinity Inlet Management Program, Cairns.

HADDON, A.C. 1912. Reports of the Cambridge Anthropological Expedition to Torres Straits 5: Arts and Crafts. Cambridge: Cambridge University Press.

HARRIS, D.R. 1977. Subsistence strategies across Torres Strait, in J. Alien, J. Golson & R. Jones (ed.), Sunda and Sahul: prehistoric studies in Southeast Asia, Melanesia and Australia: 421-63. London: Academic Press.

HENDERSON, J.W. 1995. An improved procedure for the photographic enhancement of rock paintings, Rock Art Research 12: 75-85.

JACOBSON, R.E., S.R. RAY & G.G. ATTRIDGE. 1988. The manual of photography. London: Focal Press.

JAIN, A.K. 1989. Fundamentals of Digital Image Processing. New York (NY): Prentice-Hall.

MARK, R. & E. BILLO. 2000. Application of digital image enhancement in rock art recording. http://www.infomagic.net/~rockart/Enhancement.pdf

McNIVEN, I.J. 1998. Enmity and amity: reconsidering stoneheaded club (gabagaba) procurement and trade in Torres Strait, Oceania 69(2): 94-115.

McNIVEN, I.J., B. DAVID & J. BRAYER. 2000. Digital enhancement of Torres Strait rock-art, Antiquity 74: 759-60.

2001. Dauan Rock-Art Project (Torres Strait). Report to the Australian Institute of Aboriginal and Torres Strait Islander Studies, Canberra.

PRATT, W.K. 1978. Digital Image Processing. New York (NY): Wiley-Interscience.

READ, E.J. & C. CHIPPINDALE. 2000. Electronic drawing or manual drawing? Experiences from work with rock-paintings, in C. Buck, V. Cummings, C. Henley, S. Mills & S. Trick (ed.), UK Chapter of Computer Applications and Quantitative Methods in Archaeology: 59-79. Oxford: Archaeopress. BAR International series S844.

TWEEDIE, A.S., M.T. MILTON & P.Z. STURGEON. 1971. How reliable are artificial light sources for predicting degradation of textiles by daylight? Journal of American Association Textile Chemists and Colorists 3(2): 22-35.

WATCHMAN, A. 1992. Composition, formation and age of some Australian silica skins, Australian Aboriginal Studies 1: 61-6.

1993. Evidence of a 25,000-year-old pictograph in northern Australia, Geoarchaeology 8(6): 465-73.

BRUNO DAVID, JOHN BRAYER, IAB J. McNIVEN & ALAN WATCHMAN, David, Department of Geography and Environmental Science, Monash University, Clayton, Victoria 3800, Australia. Bruno.David@arts.monash.edu.au Brayer, Department of Computer Science, University of New Mexico, Albuquerque NM 87131-1386, USA. Brayer@cs.unm.edu McNiven, School of Fine Arts, Classical Studies and Archaeology, University of Melbourne, Parkville, Victoria 3010, Australia. i.mcniven@unimelb.edu.au Watchman, School of Anthropology, Archaeology and Sociology, James Cook University, Townsville, Queensland 4811, Australia. alan.watchman@jcu.edu.au