Digitising film archives

Scanning magazine pages and tearsheets without moiré

Two workers handle a ladle beside industrial machinery in a black-and-white workshop photograph.
Lead pipe making, Paisley, 1975.Photograph by Robert Harper. Robert Harper Photograph Library, Archive.London. From the archive, shown for illustration.

A printed photograph is not a photograph. It is four grids of dots pretending to be one, and a scanner is a fifth grid laid on top. The patterns that follow are arithmetic, and so is the cure.

Scan the film for the picture and the page for the proof; a tearsheet is a certificate of publication, not a second original.

Scan a magazine page at the wrong resolution and a pattern of bands, crosses or watered silk appears across every photograph on it. Nothing is wrong with the scanner. The pattern is moiré, the interference between the dots the printer used and the pixels the scanner uses, and it can be avoided by choosing the resolution with a little arithmetic instead of from habit. Most advice on the subject amounts to "tick the descreen box". Our angle is the archivist's: the dots are part of the object, a tearsheet is evidence of publication, and the right question is what the scan is for. The method follows from the answer.

Why do scanned magazine pages show moiré?

A printing press cannot lay down grey. It can only print ink or leave paper bare. To reproduce a photograph, the image is broken into a grid of dots that vary in size: large dots where the picture is dark, small where it is light. From reading distance the eye blends them into tone. This is the halftone, and it made the illustrated press possible. Frederic Ives of Philadelphia patented the first commercially successful method in 1881, and Georg Meisenbach patented his autotype process in Germany in 1882.

Colour printing uses four such grids, in cyan, magenta, yellow and black, each set at a different standardised angle so that the dots form small rosettes instead of clashing. Printers chose those angles precisely to keep their own moiré too fine to see.

A scanner then lays a fifth grid over the four. It records the page only at discrete sample points. Where its sample spacing nearly matches the spacing of the dots, samples land on dots in some places and between them in others, and the drift between the two shows up as a slow, large-scale pattern. In signal terms this is aliasing: a fine regular pattern sampled too coarsely reappears as a false coarse one. The moiré is in the file and not on the paper, which is why the same page can scan cleanly at one setting and badly at another.

What is a halftone screen, and how fine is it?

Screens are measured in lines per inch, the number of rows of dots in an inch. The coarser the paper, the coarser the screen has to be, because ink spreads.

Printed matter Typical screen ruling Samples per dot at 300 ppi At 600 ppi At 1,200 ppi
Newspapers around 85 lpi 3.5 7 14
Magazines around 133 to 150 lpi 2 to 2.3 4 to 4.5 8 to 9
Art books and fine printing 175 lpi and above 1.7 or fewer 3.4 6.9

The rulings are typical figures, not rules, and they varied by press, paper and decade. You can measure a particular page with a magnifier and a ruler, or estimate it from a high-resolution test scan.

The last three columns are the useful ones. Two samples per dot is the theoretical minimum for registering a regular pattern at all, and in practice it is the setting at which moiré is worst. This is the trap in the common habit of scanning everything at 300 ppi: for a magazine at 150 lpi it is very nearly the worst possible choice.

What resolution avoids moiré?

There are two honest strategies and one bad one.

Resolve the dots. Scan finely enough that each dot is recorded as a shape, with at least four samples across it and preferably more. For magazines that means 600 ppi as a floor and 1,200 ppi where the original is small or the screen is fine. The file then shows the page as it is: paper, ink and dots. There is no moiré because nothing is being under-sampled. Our scan resolution calculator will give pixel dimensions and file size for any page size and resolution, which matters, since an A4 page at 1,200 ppi is roughly 9,900 x 14,000 pixels.

Blur the dots optically. Capture so coarsely, or with enough optical softness, that individual dots never register. Some scanner descreen modes work partly this way. It can look acceptable at screen size, but it discards information at the moment of capture.

The bad one is the middle: a resolution near twice the screen ruling, with no filtering. That is where the patterns live.

Turning the page a few degrees on the glass is often recommended. It changes the pattern and can help with a black-and-white halftone at a single angle. With four-colour printing there are four screens at four angles, and rotating the page away from one tends to bring it nearer another.

Should you descreen at all?

That depends on what the file is for, and an archive usually wants two files.

Descreening, or inverse halftoning, tries to reconstruct continuous tone from dots. The simplest method is a low-pass filter, such as a Gaussian blur with a radius about the size of the dot spacing, followed by downsampling and a little sharpening. It removes the screen, and it softens real edges such as type and fine line work along with it. More elaborate methods treat edges separately, and recent ones use neural networks trained on pairs of halftone and original images. Those can produce handsome results, but they are inferring detail that the dots do not contain, and their output depends on what they were trained on.

For that reason the scanner's built-in descreen option is the wrong tool for a master file. It runs before the image is saved and cannot be reversed. The sounder routine is:

  1. Master. High resolution, dots resolved, no descreening, no sharpening, saved in a lossless format. This is a record of the printed object.
  2. Access copy. Made from the master by filtering and downsampling to something like 300 ppi, where the dots have merged and the page reads comfortably on a screen.

You can see the principle at work by applying a small amount of blur to a high-resolution scan with our blur tool and then reducing it with the image resizer: the screen vanishes at the point where the blur radius matches the dot pitch.

When is it better to photograph the page?

Often. Library and archive digitisation of bound and printed matter is now frequently done with cameras, for practical reasons.

  • Bound volumes. A magazine with a spine cannot be pressed flat on glass without damage. A copy stand with a book cradle photographs it as it lies.
  • Size. A broadsheet newspaper page is larger than most flatbeds.
  • Speed. A camera exposure takes a fraction of a second; a 1,200 ppi flatbed pass takes minutes.
  • Show-through. Thin magazine paper lets the reverse of the page show through under strong light. A sheet of black paper behind the page reduces it, with either method.

The sampling rules are unchanged. A camera's pixels are a grid like any other, and what counts is the resolution at the page: divide the sensor's pixel width by the width of the area photographed, in inches. A 100-megapixel sensor, about 11,600 pixels on its long side, records a whole A4 page at just under 1,000 ppi, which resolves a magazine screen comfortably. A 24-megapixel camera, at 6,000 x 4,000 pixels, manages about 480 ppi on the same page. That is ample for a newspaper and only about three samples per dot for a magazine at 150 lpi, which is moiré territory unless the page is photographed in sections.

Cameras add their own requirement: even lighting from two sides at about 45 degrees, a lens with a flat field, and the sensor parallel to the page.

What do the standards say about printed pages?

The cultural heritage imaging standards are written for exactly this kind of reflective original. FADGI's third-edition guidelines, published in May 2023, do not rate a file by its ppi alone. They measure sampling frequency, the system's real resolving power, tone, colour accuracy, noise, uniformity and geometry from a test target, and grade the result from one to four stars. The Dutch Metamorfoze guidelines, developed for paper heritage, assign magazines and newspapers to their Light and Extra Light quality levels, where correct tonal capture is the foundation and colour accuracy matters at the higher of the two. A plain-English account of both is in FADGI, Metamorfoze and ISO 19264 explained.

For a brief to a scanning bureau, the practical consequence is to specify the purpose: a faithful record of the printed page with the screen resolved, plus a descreened access derivative. Stating only "300 ppi" invites exactly the result this article is about.

What is a tearsheet actually worth to an archive?

Less than the original as an image, and more than it as a document.

As an image, a halftone is a severe reduction. Each screen cell carries, in effect, one tone, so a photograph printed at 150 lpi across a full A4 page contains information of the order of 1,240 x 1,750 cells. The 6x7 transparency it was made from, scanned at 4,000 ppi, runs to about 8,800 x 11,000 pixels. No descreening method recovers that difference. For reproduction, for study of the photograph itself, and for anyone assembling image data, the film or the camera file is the source to use, as set out in scanner, drum or camera.

As a document, the page does what the film cannot. It fixes a publication: the title, the issue, the date, the credit line, the caption, the crop the art director chose. Those are the facts a picture researcher or buyer asks for first, as we discuss in photograph provenance: what buyers ask. A tearsheet should therefore be catalogued with its publication details and linked to the original frame where that can be identified.

One caution, which is general information and not legal advice: a printed page reproduces a publisher's layout and other contributors' text and images as well as the photograph, so a scan of a tearsheet should be treated as a reference copy and not as an image available for licensing in its own right.

The magazine pages in the Robert Harper archive

The inventory of the Robert Harper Photograph Library includes a folder headed Magazine Pages, recorded as 79 shots across 364 frames, with pages from Elle, Inside Soap, Esquire, Cosmopolitan, Vogue, Wallpaper, The Face and Harpers & Queen, together with newspaper pages. A separate Newspapers folder holds front pages and other pages, 20 shots across 76 frames. They have not yet been matched to the transparencies they were printed from, and that matching is among the more useful cataloguing tasks ahead, because each successful match attaches a publication and a date to a frame of film. The folder-level figures are in the archive inventory.

More than half of the archive is still on film and has yet to be scanned. If your company digitises mixed collections of film and printed matter and has a view on how the two should be linked, the page for digitisation partners explains what we are looking for and how to reach us.

Frequently asked questions

Why do scanned magazine pages have a pattern on them?

Because printed photographs are made of regularly spaced dots, and a scanner samples the page on its own regular grid. Where the two grids nearly but not exactly coincide, a larger interference pattern appears. This is moiré, a form of aliasing. It is created by the scan and is not on the page itself.

What resolution should I scan a magazine page at?

High enough to resolve the dots clearly: at least four samples per dot. For a typical magazine screen of around 150 lines per inch, that means 600 ppi or more, and 1,200 ppi for small originals. Scanning at about twice the screen ruling, such as 300 ppi, is the setting most likely to produce moiré.

What does the descreen setting on a scanner do?

It blurs the scan just enough to merge the halftone dots into continuous tone, then usually sharpens the result. It works, but it is applied before the file is saved and cannot be undone. For archive work, capture the page without it at high resolution, and descreen a copy afterwards in software.

Is it better to scan or photograph a magazine page?

Either can work. A flatbed gives even light and high resolution for loose pages up to its bed size. A camera on a copy stand is faster, handles bound volumes and large newspaper pages without flattening them, and is the normal method in library digitisation. The same sampling arithmetic applies to both.

Can a tearsheet be used in place of the original photograph?

Only as a reference. A printed page has been screened, inked and trimmed, and it includes the publisher's layout and other people's words and pictures. It shows that, where and when a photograph was published, which is valuable as provenance. For reproduction, the original negative, transparency or file is the proper source.

Sources

Published by Archive.London. Corrections are welcome under our editorial policy.

Help digitise the archive

52,684 analogue originals, mostly medium-format transparencies, are waiting to be scanned and catalogued. We are speaking to scanning, cataloguing and tagging partners.

See the digitisation briefWrite to the archive