High-resolution digital photomosaic recording of rock-art.

Ford, Bruce

Baseline recording should be the top priority in the fight to retain and preserve rock-art. Not only are old and so far durable images facing new environmental and industrial threats, but an entire body of recent paintings, including those of the colonial period, which have not already been selected for long-term survival by fortunate and very rare taphonomic circumstances (Bednarik 1993) are rapidly disappearing, with little prospect of effective intervention to save them. This is certainly the case on the Batavia coast near Geraldton in Western Australia where the predominant white-clay pigments have undergone serious weathering since photographs were first taken in the early twentieth century (Ford 2005).

Many rock-art sites have been recorded by archaeologists over the years. However, these recordings, even if they can be found in various archives, do no necessarily assist the conservator seeking to determine the rate and mechanisms of deterioration because, on the whole, conservators and archaeologists view rock-art through different eyes, each selecting features most relevant to their purpose from an almost infinite source of potential data. What is included and left out depends on a complex range of factors, including Indigenous cultural values, site promotion and management exigencies, theoretical and ideological bias, personal interest, aesthetic considerations, conservation awareness, time and money, and the recording method itself (Gunn 1995a).

Archaeologists find that photographs, while ostensibly objective, not only fail to capture relevant detail but contain too much distracting 'noise'. Some employ a graphic recording method involving a combination of sketching, scale drawing, photography and tracing (direct or from photographs), designed to accurately abstract motifs (ironically) from the same physical context which is of most interest to the conservator. It is time-consuming, expensive and generally speaking quite useless as a conservation record, but encourages intensive scrutiny of the often subtle physical indicators of cultural significance such as style, sequence, repainting and marking techniques and often reveals images that are less visible to the casual observer or on photographs. Tracing, which in its direct form is sometimes criticised as a 'contact' and potentially damaging technique (e.g. Gunn 1995b), locates each mark accurately and avoids common perceptual traps such as recording what one expects to see. Unlike wide-angle photography, especially of surfaces with complex topography, there are no significant focal length distortion and parallax errors.

Site registers accurately reflect the diversity of rock-art recording purposes and techniques. Not infrequently they consist of a mish-mash of site diagrams, sketches, close-up and wide-angle photographs and sometimes extremely detailed graphic representations of particular motifs, which, however, do not amount to a visual record suitable for conservation monitoring. Photographs tend to out-survive or become separated from reports, and transparencies in particular are inconvenient to refer to in the field. There is a particular tendency for close-ups to be misfiled because the site they belong to cannot be identified--most researchers have a cache of troublesome photographs waiting to be reunited with their companions. Perhaps there is also a tendency for site managers and funding agents to consider recording finished if, for example they have an archaeological interpretation of the site, not knowing how much the purpose of recording affects its content.

Many recording objectives can now be achieved by digital image enhancement of digitised or digital high-resolution photographs, selecting and enhancing colour, contrast and edges to highlight particular features (e.g. McNiven et al. 2002). In this way at least some of the data selection process can be moved from the field to the laptop, irrespective of the purpose, as long as the original photographs contain sufficient detail and are of an appropriate scale. Enhancement also works best when pigment colours and other discriminating features are reasonably distinct from an uncomplicated background and when the recorder already knows that the features to be enhanced exist.

In addition to extracting useful archaeological detail, it has also been used to answer questions in conservation. For example, Ford and Officer (2005) identified an algal-like lichen growth pattern, which was not obvious in the field, using digitally enhancing photomosaics of the Nursery Swamp II rock-painting site near Canberra.

The expense of film photography made the balance between inclusiveness and resolution in photographic rock-art recording a difficult one to strike. Wide-angle shots (encompassing complete motifs or series of motifs) lack useful resolution for most conservation and many archaeological purposes, and suffer from parallax and focal-length distortions; close-ups, essential for recording pigment condition and marking techniques, are most strongly influenced by researcher bias and necessarily exclude the wider physical and representational context, even to the extent of rendering it difficult to identify the subject in the field--a task which may also be complicated by changes due to weathering, dust, water salts, and so on. More recently, high-resolution digital photography and cheap mass-storage technology have largely eliminated the expense involved in more or less indiscriminately photographing everything, irrespective of one's interest. However, the problem of usefully organising and clearly interpreting photographic data grows with the number of photographs, especially for subsequent reference in the field.

Digital photomosaics, stitched together from individual pictures, are a potent means of displaying potentially large numbers of close-ups in a highly contextualised form. In addition to other desirable properties, they retain the resolution of the constituent pictures within a much larger visual context and are therefore easy to consult. The user can zoom in and out from pigment to motif on the same image which is, in addition, uniquely suitable for digital image enhancement at all scales.

Taking digital photomosaics

The original photomosaic in Figure 1 is composed of approximately 70 digital colour photographs each covering an area of about 1000 x 750 mm (in portrait format) at 5.1 megapixel resolution. Viewed at 72 dpi, or screen resolution, the final picture is close to a full-size facsimile of the original. The file is a fairly hefty 244mb (or 18mb compressed as a high-quality JPEG), allowing zoom from pigment- to shelter-level within the same photograph. It would not be possible to take a single photograph of the entire painted surface at this site without a fisheye lens because the front of the shelter is hemmed in by rocks and trees at the dripline (Figure 4). Aperture locking and manual control over flash intensity ensured that digital camera (Olympus C5060) settings were consistent across the board.

[FIGURES 1&4 OMITTED]

The image plane of the camera was oriented, as nearly as possible, parallel to the painted surface (wall and ceiling) at a distance of 1.5 m using a 55 mm (35mm camera equivalent) focal length for minimum distortion. In this case it was convenient to take the pictures in a pre-planned series of vertical strips of three to five photographs (Figure 3), 18 of which were subsequently joined sideways into a flattened representation of the roughly 10 m long by 3 m high painted part of the shelter.

[FIGURES 2-3 OMITTED]

A string line was laid along the shelter floor to approximately guide tripod placement. However, for each frame, the camera position was carefully measured relative to the rock surface to preserve scale and provide sufficient overlap top and bottom and to the sides of pictures taken earlier in the sequence. Each photo was recorded and annotated on a gridded sketch to keep track of progress and facilitate later reconstruction of the sequence. Where there was doubt about side-to-side overlap, the photograph of the earlier recorded adjacent rectangle could be referred to on the camera display. At this shelter, which has a fairly regular topography, planning and photography took one person about five hours.

The lighting choices are the same as for any photograph and in this case a dual diagonal orientation designed highlight surface texture was tested, but abandoned in favour of a simple front-on flash because the results did not appear to be markedly different (there was little pigment surface relief to capture) and the dual flash rig was more difficult and time-consuming to set up and use.

While digital photomosaics are more accurately and easily merged if the photographs are accurately composed and shot using a tripod-mounted camera, it is possible to obtain quite good ad hoc results handheld. Figure 5, for example, is a flattened composite of four pictures of an extended motif in a three-metre-high alcove within a hollow, dome-shaped granite boulder which was difficult to access and where the motif curled back over the viewer's head like a wave. It was impossible to use a tripod and therefore difficult to estimate camera distance accurately because of the highly irregular topography. The red pigment was heavily coated in dust and not easy to distinguish from the background; however, digital image enhancement of the resulting photomosaic provided a usefully clear representation of the complete complex motif--much better than being there in fact. The photography took minutes and the photomosaic creation and image enhancement less than an hour. The enhanced motif was taken back into the site on a laptop the next day.

[FIGURE 5 OMITTED]

Stitching photographs into photomosaics and panoramas

Several photomosaic and panorama stitching programs are available for the Mac and PC that promise to join individual pictures, more or less seamlessly and automatically, into panoramas and photomosaics. The Willigulli photomosaic was created on a PC using a combination of the new 'Photomerge' function in Adobe Photoshop[c] CS and a program called Panorama Factory[c] by Smoky City Design. Neither program was able to do the whole job alone. Other programs trialled included PanaView Image Assembler (PC) and Realviz Stitcher (Mac & PC).

The programs offer a variety of mathematical image-matching and re-projection algorithms depending on the purpose of the final image. Panoramas up to 360[degrees] may be mathematically projected on an imaginary cylinder in which the observer stands at the centre, whereas a hemispherical projection would be chosen for a 3-D interactive walk-through. It is important to recognise that the stitching process relies on distortion to match overlap areas, although the effects can be minimised by careful composition (and a regular rock topography). The original undistorted digital photographs are stored with the photomosaic for reference. It is in this sense that the photomosaic might be regarded as a viewing aid rather than the definitive recording itself--which remain the individual photographs, which not incidentally retain all of their relevant EXIF data including f-stop, effective ISO, focal length, flash setting, colour space, resolution, date and time of day, camera make and model. The embodied meta-data in digital pictures saves a great deal of documentation in the field and, with the visual clues, probably would allow quite precise reconstruction of observer position at a later time if required.

Although the process is simple, even automatable in theory, the irregular shape of most shelter surfaces and the difficulty of exactly controlling camera position in the field means that a fairly high degree of skill and experience with particular programs is required to successfully massage many pictures into a usable and passably accurate photomosaic representation.

While, on the whole, it was easy to join the individual photographs vertically to form the strips in Figure 3, it was much more difficult to merge the 18 composite strips into the final form depicted in Figure 1. This is because the stretching and bending required to merge pictures in one direction affect the overlap fit in the other. It was necessary to manipulate some of the individual strips 'by hand', matching the size and shape of overlap regions using the various 'transform' functions in Photoshop[c] before the Photomerge or Panorama Factory[c] stitching algorithms would function correctly--that is, without producing unwanted artefacts like ghosting and double images or constructing completely fake areas out of confused material in an attempt to merge the two pictures. These areas, which may be confined to quite small parts of the overlap region, can be hard to spot but are usually recognisable by repetition of particular surface features, which may or may not be accompanied by ghosting. They are more common when the overlap area is bland.

Programs offer several options for assembling photomosaics and panoramas including interactive 3-D and perspective views. However, for conservation recording, the simplest cylindrical projection, 'rolled out' into a flat rectangular representation, was chosen for ease of creation, reference and (generally) minimum distortion. Fully three-dimensional interactive photomosaics are currently extremely difficult to set up in the field with sufficient accuracy, particularly where the shelter surface has a complex shape, and require propriety software to interpret. The author routinely takes 360[degrees] 'Quicktime[c]' panoramas consisting of five to ten individual overlapping photographs which allow a complete circular view around a virtual standpoint, not so much for rock-art recording per se, but as a convenient way of putting the shelter in its environmental context and as a community resource for children and others who may never otherwise visit the site.

Conclusions

Digital photomosaics embody all of the advantages of high-resolution digital photography and, with the assistance of digital manipulation, capture many of the benefits of hand-recording or tracing. The only practical limits on resolution are the quality of the camera and the computational power required to stitch and view very large composite files.

Because the image plane of the camera, like an acetate tracing sheet, is as close to parallel to the rock surface at every point as possible (on the scale of the photograph), the photomosaic is very similar in its representation. Resolution is vastly increased compared with a single photograph of the same area, camera-to-surface parallax errors are minimal, lighting can be held consistent across the entire recording, and at 55 mm equivalent there is little focal-length distortion. Consulting the photomosaic in the field or office is much easier than organising a pile of individual photographs because the researcher can zoom in and out of the picture at will. Someone who has never been to the site has all of the context to hand in a very amenable format.

[FIGURES 6-7 OMITTED]

Some situations, for example very low or enclosed shelters with complicated topography, make it difficult or impossible to record an entire rock-art assemblage in a series of photographs which can be assembled into a rectangular 'flattened' photomosaic. However, even in these cases, which are equally difficult to record by other means, it is usually possible to record complete motifs, if not the entire painted or carved surface.

Digital photography is so convenient, and digital photomosaics such an efficient means of recording large areas quickly and accurately, that it is hard to justify not attempting to capture a site in this format, even if there is no intention to acquire the skills or software to assemble it.

Acknowledgments

The author thanks AIATSIS and the Bay of Islands Aboriginal Corporation (BIAC) in Esperance, Western Australia, for funding the Willigulli (grant G2000/6428) XX 6458 and Mt Ridley work, respectively. The projects from which the examples above were taken would not have been possible without the assistance, knowledge and goodwill of Mr Bill Bennell and Graham Tucker of BIAC and the staff and traditional Aboriginal owners of the Yamatji Land and Sea Council in Geraldton. The Willigulli rock-painting complex is within Naaguja traditional territory and the author gratefully acknowledges the guidance and assistance of Mr Keth Counsillor and Mr Terry Radford in the conservation recording and assessment project at several sites within their area, including the fascinating Willigulli complex.

REFERENCES

Bednarik, RG 1993, 'A taphonomy of palaeoart', Antiquity 68:68-74.

Ford, BL 2005, 'Batavia coast rock-art project: final report to Yamatji Land and Sea Council and the Australian Institute of Aboriginal and Torres Straight Island Studies', unpublished report.

Ford, BL & Officer, K 2005, 'Micro-environmental study of lichen invasion at a Ngunnawal Aboriginal rock-art site in the eastern Australian sub-alpine region', ICOM 14th Triennial Conference, The Hague, September (preprints).

Gunn, RG 1995a, 'Recording Aboriginal rock images for management purposes', in GK Ward & LA Ward (eds), Management of rock-art imagery, Australian Rock-art Research Association, Melbourne (Occasional AURA Publication 9), pp. 93-6.

--1995b, 'Guidelines for recording Australian Aboriginal rock imagery', in GK Ward & LA Ward (eds), Management of rock-art imagery, Australian Rock-art Research Association, Melbourne (Occasional AURA Publication 9), pp. 126-7.

McNiven, IJ, David, B & Brady, L 2002, 'Torres Strait rock-art: an enhanced perspective', Australian Aboriginal Studies 2002/2:69-74.

Bruce Ford

Canberra

Bruce Ford completed an honours degree in chemistry in 1976, worked as a conservation scientist for various museums from 1982, and was Head of Conservation at the National Gallery of Australia. He holds a postgraduate diploma in rock-art conservation from the Getty Conservation Institute--University of Canberra and has worked on rock-art conservation, microclimate research and site management issues since 1989.

<bford@netspeed.com.au>