FADGI
FADGI — the Federal Agencies Digital Guidelines Initiative — is a collaboration of United States federal agencies, hosted by the Library of Congress, that publishes technical guidance for digitising cultural heritage. Its Technical Guidelines for Digitizing Cultural Heritage Materials are used as an image-quality reference well beyond the United States. The current version is the third edition, May 2023, and it is free to read.
PHOTARCH does not assess objects by FADGI. It assesses them by its own three levels — record shot, study and reproduction, set out under PHOTARCH's levels — and offers FADGI, with ISO 19264-1 and Metamorfoze, for paintings, documents, prints and other flat originals, which is what it was written for: a setting in the PHOTARCH app turns them on, clearly labelled. This page explains the standard itself.
What it is for
FADGI describes what a good digital copy of a cultural heritage original is, and how to prove it. It does this per type of material and per quality level, with numbers: how many pixels per inch of the original, which bit depth and colour space, and how far each measured property may deviate from a reference.
The guidelines call themselves informative, not prescriptive. They are not a manual for getting a result, and they say plainly that no set of metrics replaces the judgement of an experienced photographer. But they are equally plain about the reverse: without a managed quality-assurance programme, no claim of FADGI conformance can be made.
The star levels
FADGI grades on four levels. Three of them correspond to the levels of ISO 19264-1 (§2.1):
| Level | ISO 19264-1 | What FADGI says it is for |
|---|---|---|
| ★★★★ Four-star | Level A | The best imaging practical today; state of the art, best suited to smaller projects |
| ★★★ Three-star | Level B | A very good professional image, appropriate for most uses |
| ★★ Two-star | Level C | The minimum acceptable for most professional digitisation |
| ★ One-star | — | A reference to locate the original; not ISO conformant |
FADGI's own advice is to digitise to three-star where possible, so that nothing has to be done twice, and to be careful with four-star unless the programme is very experienced. A star rating belongs to an image only when it has been reached through image analysis — a level is measured, not declared.
What it measures
The parameters are measured on a test target photographed with the same set-up as the originals, and analysed in software. Since 2023 they are expressed in CIELAB rather than RGB count values; older measurements no longer validate conformance.
| Parameter | What it tells you |
|---|---|
| Sampling frequency | Pixels per inch of the original — how much detail can be recorded |
| Tone response (OECF) | Whether greys come out at their true lightness, ΔL* |
| White balance | Whether neutral greys stay neutral, ΔE(a*b*) |
| Lightness uniformity | Whether the light falls evenly across the field |
| Colour accuracy | Mean and 90th-percentile ΔE2000 over the colour patches |
| Colour channel misregistration | Whether red, green and blue land on the same pixel |
| SFR | Sharpness: how well fine detail is resolved, and whether it is aliased |
| Reproduction scale accuracy | Whether the stated resolution matches the real one |
| Sharpening | Whether edges have been artificially boosted |
| Noise | Random variation in flat areas, as a standard deviation of L* |
| Gain modulation | Clipping in the tone curve; informative only in 2023 |
Besides the measured values, every level names the file formats allowed for the master (TIFF or JPEG 2000), the bit depth and the colour spaces.
Fifteen kinds of material — none of them an object
Chapter 3 sets the numbers per material category, from bound volumes and manuscripts to maps, photographic prints, negatives, X-ray film and microfilm (§3.1–3.15). There is no category for three-dimensional objects. The nearest is §3.13, Paintings and Other Two-Dimensional Art, which is itself marked provisional and notes that artwork is rarely truly flat.
For that category the levels look like this, in part:
| ★ | ★★ | ★★★ | ★★★★ | |
|---|---|---|---|---|
| Resolution at the original | 300 ppi | 300 ppi | 450 ppi | 600 ppi |
| Bit depth | 8 | 8 | 16 | 16 |
| Colour space | sRGB or wider | sRGB or wider | Adobe RGB, ProPhoto, eciRGB v2 | Adobe RGB, ProPhoto, eciRGB v2 |
| Mean colour error, ΔE2000 | < 6.5 | < 5 | < 3.5 | < 2 |
| Noise, std. dev. of L* | < 4 | < 3 | < 2 | < 1 |
The resolution limits carry a small tolerance for scale accuracy, so 300 ppi in the table reads as more than 292.5 ppi.
Conformance is a programme, not a file
FADGI describes four parts that together make a conforming programme (§2.2): the technical guidelines themselves, documented best practice, a repeatable digital image conformance evaluation with targets and analysis software, and trained professional staff. Conformance is established for a set-up, and every change to equipment, light or workflow means testing again.
This is the most common misunderstanding about FADGI. A single file cannot carry a FADGI certificate. What it can carry is evidence: the measurements, the target, the set-up it was made with.
How to relate to it
- Decide the level by use, before you photograph. Three-star for most collection work, four-star for originals where colour and detail are the point. A level chosen after the fact is only a description.
- Measure what you can, and say what you measured. A target in the sitting is the minimum; full conformance needs analysis software and a resolution target.
- For objects, don't claim it. No FADGI category describes an object, and none says how it should be lit, laid out or scaled; the nearest, flat art, assumes an original lit evenly across the field. PHOTARCH assesses objects by its own levels instead — a study asks for 12 px/mm at the object, about 300 ppi, and a reproduction for 24, about 600 — and leaves evenness of light out. That is the gap the PHOTARCH method fills.
What PHOTARCH measures, and what it does not
The PHOTARCH app reports FADGI only when the setting for flat originals is on. It then grades the photograph against FADGI §3.13 on what it can measure from the picture itself:
- the file format, colour space and bit depth of the master, the bit depth judged on the file as FADGI asks — a 14-bit raw written into a 16-bit TIFF counts as 16 — with the bits the camera recorded, read from the raw file's own header, shown beside it;
- the resolution at the object, in pixels per inch of the object's own size, from the scale set in Measure;
- colour accuracy, white balance and tone response, from the ColorChecker photographed in the same sitting — mean and 90th-percentile ΔE2000, white balance ΔE and ΔL* on the neutrals.
It does not measure sharpness on a target (SFR), sharpening, noise, lightness uniformity, colour channel misregistration, reproduction scale accuracy or gain modulation, and it says so next to every grade. Flat originals are lit evenly from two sides, as FADGI assumes, but the app does not measure how evenly. A green light therefore means nothing that was measured stands in the way of this level — a well-founded indication, not a conformance claim.
Sources
- guidelineFADGI (2023). Technical Guidelines for Digitizing Cultural Heritage Materials, 3rd ed., May 2023. Library of Congress.§2.1 star system and ISO correspondence · §2.2 components of conformance · §2.4 evaluation parameters · §3.13 paintings and other two-dimensional art.
- standardISO 19264-1:2021. Photography — Archiving systems — Imaging systems quality analysis — Part 1: Reflective originals.See ISO 19264-1.