Principles & Setup

A structured approach to artefact photography

The PHOTARCH method is built upon a set of clear and repeatable principles designed to ensure visual consistency, scientific comparability, and long-term usability of archaeological imagery. Its goal is to make object photography replicable not only within a single institution or excavation, but across collections, countries, and decades. While individual images may appear deceptively simple, each photograph produced under PHOTARCH follows a strict logic regarding lighting, angle, background, measurement and framing. This page outlines the core principles of the method and the recommended setup for implementation, whether in a museum environment or in the field.

1. Lighting: Controlled and Repeatable

Lighting is at the heart of PHOTARCH. A single fixed light source (key light) is placed at a 45° angle from the object's side and above — if the object lies at the centre of a clock face, the key is at 10:30 — ensuring consistent modelling of shape and surface. A fill light or reflector is used on the opposing side to soften shadows without eliminating them. The aim is not to create artistic mood, but to produce measurable, interpretable, and repeatable lighting conditions across sessions and objects. Where possible, the light source should be colour calibrated and spectrally neutral.

1. KEY LIGHT — 10:30 · 45° above
2. CAMERA — plumb view, straight down
3. FILL LIGHT — opposite the key
4. CAST SHADOW — toward 4:30, as long as the object is high
5. OBJECT
6. WHITE GROUND — RGB 255

The upper-left key is not a preference. Three independent lines converge on it. The visual system assumes light from above and slightly left, and finds shape fastest when it is there (Sun & Perona 1998; Mamassian & Goutcher 2001). Western painting adopted the same from about 1420, measurably so: of 9,469 dated paintings, 65–68% are lit from the left (Carbon & Pastukhov 2018). And archaeological illustration codified it as a drawing rule long before photography — objects are shaded as if lit from the top left of the page (Martingell & Saville 1988). The angle was known in the Renaissance, and Leonardo da Vinci noted it: a shadow as long as the object means a light at 45° (Leonardo, Richter §516).

Two limits belong here rather than in a footnote. The leftward part is a convention, not a universal — the bias weakens in readers of right-to-left scripts (Andrews et al. 2013). The assumption is also weak against real cues: whatever direction you actually light from will be read as the direction of the light (Morgenstern et al. 2011). What matters is that we agree on one direction and keep to it.

Direction, not evenness

PHOTARCH does not want light that falls evenly from every side. It works against it. The light has a direction: it comes from the upper left and falls toward the lower right, in every photograph.

This is something the method arrived at, not something it inherited. While the ground was part of the finished picture, directional light had a price: its fall-off across the ground made every photograph look unevenly lit. Taking the ground to white removes that price and keeps what direction gives — relief, form, and a shadow that places the object on its ground. With the ground gone, an object lit evenly from all sides looks like a cut-out. Lit from one side, it looks like the thing it is.

Here PHOTARCH departs from the digitisation standards, and it does so on purpose. FADGI and Metamorfoze measure how evenly the light falls across the field, because they were written for flat originals, where an even field is the whole point. For an object it is the opposite of the point. A PHOTARCH photograph is not meant to meet those requirements, and nothing in the method tries to — see Standards. PHOTARCH's own assessment of what a photograph is fit for leaves evenness out and says so, and the direction of the light is the photographer's to record with the pictures. Flat originals — paintings, documents, prints — are another matter: they are lit evenly from two sides, and for them the standards can be turned on.

2. The kept shadow

The contact shadow is kept, deliberately, and it is the second best-supported part of the method.

A shadow where the object meets its ground places the object on that ground and gives it height. Vision uses contact shadows to settle position and depth, and tolerates a good deal of inconsistency elsewhere before it stops believing them (Casati & Cavanagh 2019). Shape can be recognised from shadow alone (Norman et al. 2000). Remove the shadow and the object floats: the picture becomes a cut-out, and the viewer loses the one cue that says how far the object stands off the surface.

This is a departure, and it should be said plainly. Some collections raise small artefacts on glass above the background for the express purpose of eliminating shadows. PHOTARCH does the opposite on purpose: the shadow is evidence of form, and an image meant to be compared with another image is better with it than without.

3. Camera angles: named, not numbered

In PHOTARCH, the plumb view — straight down onto the object as it lies on the board — is the primary image. This perspective offers clarity, neutrality, and easy comparability: a uniform overview that places all objects on equal visual footing. It is the picture the archive shows first and measures from.

A single image rarely tells the whole story. Supplementary angles capture form, surface texture, and depth, and they remain pictures of the same object rather than objects of their own.

Angles are named positions on a sphere, not coordinates on a grid. An object lying on the board has two faces, like a coin, two ends and two sides. Front is the upper face as it lies — the plumb view, and the main picture — and Back is the underside, the face it rests on. Top is the end at the top of the plumb view and Down the end at its bottom; Left and Right are the sides at the left and the right of the plumb view. The ends and the sides are each photographed at eye level with the object turned on the board. ATL, ATR, ADL and ADR — A as in angle — are the four diagonals, looking down at 45° at the corner where an end meets a side: angle top left, angle top right, angle down left and angle down right.

Nobody has to remember an abbreviation to know which side of an object they are looking at, and the plumb view says which way round it lies: the end at the top of the main picture is the top.

1. F · FRONT — the plumb view
2. B · BACK — the face it rests on
3. T · TOP — the top end
4. D · DOWN — the down end
5. L · LEFT — the left side
6. R · RIGHT — the right side
7. ATL — angle top left
8. ATR — angle top right
9. ADL — angle down left; about the view of this drawing
10. ADR — angle down right

The camera moves to each view; the lamp stays where it is. Each picture therefore carries both its position and the light that position gives — which is what makes the same angle from two different institutions the same angle.

4. Background: Clean, White, Consistent

PHOTARCH photographs onto a pure white ground (RGB 255,255,255) and keeps the shadow the object casts on it.

Grey has one real argument behind it, and it deserves stating rather than brushing aside. A white surround makes an object look slightly darker and less saturated — simultaneous contrast, described by Chevreul and demonstrated by Albers, and the reason ISO 3664 specifies the surround in a viewing booth — a neutral one, in its 2009 edition. But note what that standard governs: the room and the frame around the picture, never the picture's own background. No international standard mandates a grey background for an object image, because no international standard addresses object images at all.

White answers the argument. The visual system takes the highest luminance in a scene as white and scales everything else to it (Gilchrist et al. 1999). A pure white ground gives every image the same anchor. On grey, the anchor becomes whatever happens to be brightest in the object itself — and it moves from object to object, so two pictures are scaled differently before anyone compares them. For a method whose whole purpose is comparison across thousands of images, that decides it. And because every PHOTARCH image shares the same white, the induction grey would avoid is constant here, and cancels in the comparison.

1. ANCHOR · THE GROUND — one anchor, every image
2. ANCHOR · A HIGHLIGHT — it moves with the object

Grey is the reference for making the picture. White is the ground for showing it.

Some collections photograph on a light grey, others on grey or black with the shadow removed. PHOTARCH chooses white with the shadow kept, and it is in good company practically as well as theoretically: the Portable Antiquities Scheme built 1.5 million records on a white-background rule, natural-history digitisation defaults to white, and uniform backgrounds are a stated prerequisite for machine-readable specimen images (East et al. 2025).

White also has two plain working virtues: clipping is visible in the histogram, and no cut-out is needed, so the object keeps its spatial grounding.

5. Scale: measured, not laid beside

There are numerous guidelines for how best to position a scale bar next to an artefact. A physical scale bar, however carefully placed, confirms the size of the scale bar. It sits on a different plane from the object, at a different distance from the lens, and it tells you what a millimetre is there. The board has the same limit for a thick object — its markers lie on the surface the object rests on, not on the object's upper face — and PHOTARCH does not escape it. It says so: every measurement records the plane its scale holds for and how it was set, and what that means for an object with height — how large the error is, and how to state it — is worked through under Scale in artefact photography.

PHOTARCH takes the scale from the object itself: its longest side, measured with a digital caliper and entered with the picture. That is the measurement that holds for the object. The PHOTARCH board is the calibrated alternative — a surface with markers in its corners, which the application finds and reads to derive how many millimetres one pixel covers on the plane the object rests on, correcting the perspective in the same operation. A ruler in the frame is only a fallback, and the file records which of the three it was.

That number is then written into the file as Image Scale Metadata, together with the pixel grid it applies to and an uncertainty figure with its assumption spelled out. Any program can read it and draw a true scale bar — after the file has been resized, or turned; a copy cropped to another shape is recognised, and its scale is not used. The archive draws its scale bar from that number rather than from anything in the frame.

This is the shift that turns a photograph into a data point: from a reference object laid beside the artefact to a measurement carried inside the file.

The same problem, solved once already

Medicine reached this conclusion decades ago, and for harder reasons. A radiograph cannot be measured by laying a ruler beside the patient: the image is a projection, and anything nearer the source is magnified more than anything behind it. So the scale is carried as a field in the file rather than as an object in the frame — and DICOM, the standard that does it, distinguishes two of them.

Imager Pixel Spacing is the distance between pixel centres on the detector. Pixel Spacing is the distance in the patient. They are different numbers about different planes, and the standard is explicit that the first must never be quietly adjusted to stand in for the second — that is what the second exists for. A third field records the magnification factor, and a fourth records how the calibration was arrived at.

That is precisely what Image Scale Metadata asks of a photograph: the number, the grid it belongs to, and how it was obtained. Not because ISM copies DICOM — ISM is a handful of fields where DICOM is an industry — but because anyone who takes measuring from images seriously arrives at the same three questions. Getting the grid wrong is the error a radiologist is warned about and the one an archive makes when it writes the master's dimensions onto a smaller copy of it. The answer in both cases is the same: say which pixels the number is about.

The difference is that medicine has had this since the 1990s and the imaging standards for cultural heritage still have nothing of the kind — which is the gap described under Standards.

Try it on a file of your own

The argument is easier to believe when you can test it. The ISM viewer opens an image and draws its scale from the file itself — nothing is sent anywhere, it runs in the browser. It will also write a scale into an older picture that has a ruler in it: measure the ruler once, and the file carries its own scale from then on.

Beside it are a validator for checking what a file claims, and a set of test files with the answer a correct reader must give for each — including one that is wrong on purpose, because a crop to the same shape as a resize cannot be told apart from the dimensions alone, and an implementer should meet that limit deliberately rather than by accident.

Scale and colour targets are therefore not composed into the primary image. The colour target is read at capture and its result lives in the file's colour profile; the scale lives in the file's metadata. Where you photograph without a board, include a reference frame in the set showing scale and colour chart under identical lighting — it is the sensible fallback, not the main path.

6. Framing: square, with the same margin

Every PHOTARCH photograph is published square, and in every one the object takes the same share of it. The picture is taken in the camera's own frame — 3:2 on most cameras, 4:3 on medium format — and delivered as 1:1: it is cropped to a square in which the object's longest side takes 60 % of the side, leaving 20 % clear at each end — or 75 %, leaving 12.5 %, for a slender object, one at least 1.5 times as long as it is wide. The object is always in the middle of the square. Across the object the margin is whatever its width leaves, so the air round an object is the same along its length and nearly the same round the rest of it.

Inside the square is the safe zone: a circle just inside its edge, 96 % of the inscribed circle, outside which the picture is white and nothing else. The object and its shadow must both lie inside it. The shadow is not counted in that share, but where it would leave the safe zone the square is made larger, with the object still in its middle, until it fits — so no shadow ever touches the edge of the picture. The board stays square, whatever the frame.

1. THE CAMERA'S 3:2 — the frame as taken
2. THE SQUARE KEPT — always centred on her
3. HER LENGTH — 60 % of its side
4. THE SAFE ZONE — white outside it; her shadow inside

Square, because objects have no orientation. A blade lies long, a bead is round, and the same figure seen end on for the Top view is suddenly as wide as it is high. Kept in the camera's 3:2, every picture would ask for a decision — portrait or landscape — and a sitting of ten views would end up in two shapes. Cut to a square, every object and every view ends up in the same frame, and the camera never has to be turned on its stand: its long side stays where it is, and the square is cut where the object lies.

What it asks at the camera is room. The square is cut from the frame's short side, so a long object must not be framed to fill the frame along its long side: the square cut from that frame would be too small to hold it. Frame so that the object's longest side takes no more than about half of the frame's short side, and less for an object that stands high, whose shadow needs room; then the square can always be cut round it. A camera that can show a square in the viewfinder makes this easy to judge; the crop is made afterwards either way, centred on the object.

Squares also line up. The archive's grid, a contact sheet, the ten views of one object, the same view of a hundred: squares tile without bands or gaps, and without one picture standing on end among the others.

The same share, so that objects are compared and not their framing. When every object's longest side takes the same share of its picture's side — 60 %, or 75 % for a slender object, and less only where a long shadow needs the room — a row of different objects is read for their form, not for how close the photographer happened to stand. The share holds at any size the picture is shown at: in a picture 3,000 pixels square the object's longest side is 1,800 pixels, in a thumbnail 200 pixels square it is 120.

It follows that a 30 mm bead and a 300 mm blade fill their pictures alike, and nothing in the framing says which is larger. That is the price of the same share, and the reason the size travels with the file instead: the scale in its metadata (section 5) is what gives the size back. The archive draws its scale bar from it, and any program that reads it can show two pictures at their true relative size.

1. AS TAKEN · THE CAMERA'S 3:2 — the square to keep, centred on the object
2. AS DELIVERED · 1:1 — the longest side 60 % of the side; the safe zone (dotted) holds object and shadow

Why a margin at all. A frame the object fills would put more pixels on it, and three things speak against it.

• The shadow. The kept shadow falls toward 4:30 and is as long as the object is high (section 2). It belongs in the picture, and like the object it lies inside the safe zone. At 60 % the margin holds the shadow of most finds lying on their ground: the Venus, front up or on her side, fits at 60 % with her shadow inside the safe zone. An object stood on end casts a shadow as long as it is high and gets a larger square: the Venus on her end for the Top view, 100 mm high, takes 22 % of the side.
• The white, on every side. The ground is the picture's anchor for white (section 4), and outside the safe zone the picture is nothing else, so it is there on all four sides and in every corner, and the picture can be shown cut to a circle without losing anything. It also shows the whole object is in the picture: an object that touches the edge of the frame may have been cut by it.
• The middle of the lens. Distortion, fall-off and softness grow toward the edges of the frame. An object in the middle of the square, with a margin on every side, stays clear of the corners, where the lens draws worst.

A rule, not a judgement. The square is worked out from the object's own outline, not chosen by eye, so two people photographing the same object arrive at the same picture, and a collection photographed over years by different hands still lines up.

What the crop does not change, and what it costs. The crop removes pixels; it does not resize them. The scale stays as it was measured, measurements are made on the uncropped picture, and the board's markers, read before the crop, can fall outside it: the scale they gave is in the file. The cost is pixels. Cutting a square from the frame discards at least a third of a 3:2 picture and a quarter of a 4:3 one, and at most 60 % of the pixels across the square fall on the object's longest side (75 % for a slender object). A 100-megapixel camera, 11,648 × 8,736 pixels, gives a square at most 8,736 pixels across, and a 100 mm object at most 5,242 of them: 52 pixels per millimetre.

Conclusion

Followed together, these principles give uniform photographs of artefacts that can be measured, again and again, with whatever equipment is at hand. The method is designed to serve as a reliable foundation for documentation, analysis, and communication. As the volume of cultural heritage imagery continues to grow, consistency becomes not only a technical concern, but an ethical one. If the past is worth preserving, it is worth photographing properly.

References · light direction and shadow

• empiricalSun, J. & Perona, P. (1998). Where is the sun? Nature Neuroscience 1, 183–184.Convex shapes are found fastest when light is assumed 20–30° left of vertical.
• empiricalMamassian, P. & Goutcher, R. (2001). Prior knowledge on the illumination position. Cognition 81, B1–B9.An above-left prior, about 26° left of vertical, in shape-from-shading.
• empiricalCarbon, C.-C. & Pastukhov, A. (2018). Reliable top-left light convention starts with the Early Renaissance. Frontiers in Psychology 9:454.9,469 paintings; 65–68% lit from the left from about 1420 onward.
• caveatAndrews, B., Aisenberg, D., d'Avossa, G. & Sapir, A. (2013). Cross-cultural effects on the assumed light source direction. Journal of Vision.The leftward bias weakens in readers of right-to-left scripts.
• caveatMorgenstern, Y., Murray, R. F. & Harris, L. R. (2011). The human visual system's assumption that light comes from above is weak. PNAS 108(30).Real lighting cues override the assumption; light from where you mean to.
• conventionMartingell, H. & Saville, A. (1988). The Illustration of Lithic Artefacts. AAI&S / Lithic Studies Society."It is conventional to assume, when shading, that objects are illuminated from the top left-hand corner of the page."
• primaryLeonardo da Vinci. Notebooks (Richter), §514–520.§516: a shadow as long as the object — a light at 45°. Leonardo fixes height and softness, not direction.
• monographCasati, R. & Cavanagh, P. (2019). The Visual World of Shadows. MIT Press.Contact shadows anchor objects to surfaces and carry position and height.
• empiricalNorman, J. F., Dawson, T. E. & Raines, S. R. (2000). Perception and recognition of natural object shape from deforming and static shadows. Perception 29, 135.
• counterCHIN / Canadian Museum of Civilization. Digitisation standards, artefacts.Lamps at 45°; objects raised on glass to remove shadows; black or neutral grey ground. PHOTARCH departs from this deliberately.

References · measuring from an image

• standardDICOM PS3.3. Imager Pixel Spacing (0018,1164).The distance between pixel centres on the detector. "Shall never be adjusted to account for calibration against an object of known size" — Pixel Spacing exists for that.
• standardDICOM PS3.3. Pixel Spacing (0028,0030) · Estimated Radiographic Magnification Factor (0018,1114) · Pixel Spacing Calibration Type (0028,0A02).The distance in the patient, the factor between the two planes, and a field recording how the calibration was obtained.
• standardNEMA. Correction Proposal 586: Pixel spacing and calibration in projection radiography.Why the distinction had to be made explicit, and what goes wrong when it is not.

References · the ground

• empiricalGilchrist, A. et al. (1999). An anchoring theory of lightness perception. Psychological Review 106, 795.The highest luminance in a scene is seen as white and anchors the scale.
• classicChevreul, M. E. (1839). De la loi du contraste simultané des couleurs.The honest case for grey: induction from the surround.
• classicAlbers, J. (1963). Interaction of Color. Yale University Press.
• standardISO 3664:2025. Graphic technology and photography — Viewing conditions, 4th edition, replacing ISO 3664:2009.The surround — the area next to an image's border, in the plane of viewing — can affect how its tone and colour are seen, and is specified (Introduction, §3.17). The 2009 edition required a print's surround to appear neutral, with a CIELAB chroma no greater than 2 (§4.1.2). Governs the room, not the image.
• institutionalRijksmuseum (2017). Manual for the Photography of 3D Objects.Neutral background at L* 80; whitest point L* 97. "We are not creating art, we are photographing art."
• practicePortable Antiquities Scheme. How to take a photograph. British Museum (accessed 4 October 2026)."A white background is used for most objects." The key light from slightly above at 45°, usually from the top left.
• empiricalEast, A. et al. (2025). Optimizing image capture for computer-vision-powered identification. Methods in Ecology and Evolution.Uniform, consistent backgrounds as a prerequisite for machine-readable specimen images.