How to digitise slides. Recommendations and working lists for the reproduction of a very special artefact

The scanner and its equipment – some reflections before the acquisition

It is important to decide before buying (or renting) and using a scanner what should be done with the created files: are they to be kept for a long time (which makes migration necessary) or just for a short while (2-3 software generations), to be used for high quality reproduction (e.g. projection, printing) or just for the archive’s low-resolution online-database to show the collection to third parties. Consequently, the archive has to make sure that the scanner can deliver what is needed. Unfortunately, there are very few specialised media shops and only a handful of qualified and competent people capable to inform about the “inner life” of an image scanner as we had to experience ourselves. Therefore, time-consuming research on the internet is (almost) obligatory.

 

The JISC Digital Media Group (formerly Joint Information Systems Committee), a not-for-profit company for higher education and research based in the UK, states: “Scanners are one of the most common methods of capturing digital images from original 2D analogue works and very few digitisation projects can manage without at least one within their selection of capture devices.” (http://www.jiscdigitalmedia.ac.uk/toolkit/digitisation-equipment/conclusion3)

 

As slides are rigid and flat, a flatbed scanner can be used. A number of scanners in DIN A4 and some in DIN A3 with different resolutions are on the market, most are good for light reflecting (opaque) objects (e.g. paper documents), some also for transparent artefacts such as slides. A flatbed scanner normally needs a special equipment to reproduce transparent objects which is integrated in the cover. (The white “reflection shield” for non-transparent objects has then to be removed). Diapositive scanners for small “homemade” diapositives are not suitable for lantern slides in a variety of sizes.

 

When planning to buy or rent a scanner, the most important elements are the kind of scanner, its lamp, and the sensor. The lens is also elementary.

 

  1. Scanner models

The sensor is placed according to the way it is supposed to work. Several models are on the market: a. scanners for opaque objects where lamp and CCD-line sensor are under the glass plane (suitable for lecture texts, manufacturer catalogues etc.); b. scanners with an additional light source in the cover working with a CCD-line sensor suitable for large artefacts (e.g. slides in the length of a DIN A4 sheet which is 29,7 x 42cm); c. scanners with a additional light source in the cover working with several small lenses and a CCD-line sensor suitable for small objects (e.g. slides in 8,5 x 10 cm).

 

For (un-)framed diapositive film, a diapositive scanner (aka “slide scanner”) is often recommended. It can take small formats such as 24 x 36 mm image (surface 864 mm²), and also smaller ones (24 x 33 mm, 793 mm², see Kraus 1998, p. 92), but is not suitable if one wants to also scan the frame with its 35 mm height and c. 38 mm width (1.330 mm²) between the two lines shared with the neighbouring frames. There are also scanners for professional diapositives of medium (e.g. 42 x 56 mm, 56 x 68 mm) and large size (e.g. 95 x 116 mm), but they won’t scan the whole frame either (see Kraus 1998, p. 92-93) and thus leave out original indications on the edges (so-called “edge codes”). These are not suitable for slides, and they seem to no longer be sold today. Given the range of slide formats, the only solution – a flatbed scanner for transparent objects – has to be chosen very carefully.

 

  1. Light sources

Depending on what has to be scanned, a bright light source sits under the pane of a scanner to illuminate an opaque sheet from below, or shines from above through the transparent object. The light source of the scanner influences the result as it produces light rays of different composition of different wave-lengths. Most flatbed scanners work with a Xenon bulb, LEDs or Cold Cathode Fluorescent light (CCFl), film scanners also use tungsten-halogen, mercury and metal halide lamps.

LEDs produce little heat and have a longer life span than CCFls. Every light source has a certain characteristic rage of light frequency it emits, as all have different wave compounds. Xenon lamps emit heat, but contain red, green and blue light in more or less equal portions, disposing thus of a continuous spectrum which means they are “neutral” and will not accentuate specific areas or omit parts of the visible light. (See also: “The potential of different external light sources for photographing voluminous objects” in the photographing section.)

 

Flatbed scanners work with diffused light; when used for a translucent object, a part of the light “pierces” it from different directions, the rest, which does not cross the slide, is simply lost. A collimated light source concentrates the beam on the object and achieves a higher brightness on the sensor, but when scanning with a collimated lamp (e.g. in a film scanner) it generally enhances the contrast of the film material in the reproduction.

 

The digitisation should not harm the slide. Especially photographically taken positives and negatives are sensitive to light. A longer exposure to intense light, or to a lamp that contains ultra-violet light, has to be avoided, as this can influence the silver-grains in the emulsion.

 

  1. The form of the sensor

A flatbed scanner which does line-scanning and is therefore considered “one-dimensional” (still-cameras have a “two-dimensional” array called a “matrix CCD”), works as follows: the slide lying flat on a glass pane is continuously illuminated by light rays. The reflected or transmitted rays are sent via a movable system of mirrors and lens(es) under the pane (see Ang 2006, p. 159) onto the image sensor. The sensor of a “linear CCD” is pushed by a motor and advances in mini-steps from one end of the pane to the other. As it is arranged in one or three rows, the sensor takes “samples” regularly and “line by line” (or rather “row by row”) from one end of the glass pane to the other, covering the entire surface of the slide. The sensor is arranged as a regular grit (or row) with the pixels being at a regular distance from each other. In the reproduction process, a scanner with a resolution of 600 by 1.200 spi (for its total surface) “dissects” the object’s continuous surface horizontally into 600 different “samples” per inch and vertically into (theoretically) 1.200 lines as it stops 1.200 times (“sampling rate”) on its way along the length of the scanner bed (Kraus 1998, p. 13, 240). (As even a big slide would not cover the whole bed, the scanning unit reproduces also empty space as it glides over the whole glass pane.) A sensor has to be thought of as a “grit” (even if the moving sensor has only one or three rows). Each pixel of this “grit” has one corresponding point on the slide. The pixel registers its physical characteristics. The more pixels are on a sensor, the more samples can be taken and the more corresponding points can be depicted later on a monitor or in print. The artefact which forms a continuous unit is turned by the digital reproduction into a raster of extracts (“samples”), each exactly at equal distance from its neighbours.

 

To register colour, the sensor is equipped with red, green and blue (RGB) filters which cover the light-sensitive zone. Older “three-pass scanners” pass three time along the document, each time with another of the RGB-filters in front of the pixel row, capturing one primary colour after the other and producing three different images in grey (three “colour separations”). It seems that passing three times over the pane can produce tracking errors. (http://printwiki.org/Scanning)  ”Single-pass scanners” have a sensor with three rows, each line with pixels covered by filters in just one of the RGB-colours; they take all information at once, which makes the registration process quicker (Kraus 1998, p. 23).

 

  1. The scanner lens

Like still-cameras flatbed scanners have an objective which sits in front of the sensor. It can be made of plastic, as there is no risk for scratching. It is situated under the glass pane and forms a unit together with the light source and the mirror system (CCD, CMOS). The light beam is sent on or through the slide, then reflected by an arrangement of mirrors and sent into the objective and on the sensor behind it. A scanner has a wide-angle lens built of spherical and aspherical lenses to correct aberrations in form of chromatic, spherical and diffraction errors. Given the short distance between the objective and the slide, it has to cover a wide angle of view. (http://www.kyocera-optec.jp/english/products/unit/scanner.html)  While the objective of a diapositive scanner has to cover a small area, the lens in a CCD- or CMOS-flatbed scanners has to cover a much larger field. This reduces the sharpness of the reproduction compared to the first. A CIS-scanner has a unit which combines a row of white LED lamps with a row of lenses and a lengthy sensor; as the length of the unit equals the width of the glass pane, the object should be equally sharp.

 

Parameters to checked when buying or renting a flatbed scanner

 

Before buying or renting a scanner, one should establish a list of what is needed and desirable. A check with local sales agents and on the internet is recommended to learn more about the following parameters. As the scanner is built as a unit, hardware changes (such as using different objectives with a still-camera) are hardly possible.

 

  1. scanning surface

If the glass pane (sometimes called “platen”) is really thick, it may absorb or refract a part of the light sent by the lamp through the slide. This would slightly falsify the result, but could be corrected in post production by adding some brightness. There is a certain chance that the pane will be (slightly) scratched, as slides have “corners and edges”, e.g. when they are broken, their frame is made of metal or it has metal parts (movable slides). Scratches can be seen as black lines in the scan, because light falling through the slides is diffused by the scratch and cannot reach the sensor. The pane should therefore be replaceable.

 

 

Hints by the scanner operator:

The glass panel should be cleaned with a glass cleaner for windows and a clean smooth cloth to avoid scratching the surface. Opticians use tea towels for eyeglasses, which makes them ideal for cleaning the scanning surface. Paper tissues are a high risk, as they produce abrasion and scratches.

 

  1. cover

Lantern slides can be of a certain “bulkiness”, as their wooden frames are up to 5 mm thick. (Some can be thicker.) To avoid unwanted disturbing light from other sources interfering with the scanning, the cover has to be closed. The cover should be flexible so that a thick slide can be placed and the cover still be closed. When the slide’s image is scanned, the light is emitted from above; to avoid losing too much light by e.g. scattering, the transmitting light unit in the cover has to lay flat and 100% parallel to the slide and the pane.

 

  1. light source

It is very important that the light source stays absolutely stable in colour temperature and brightness (no flickering) while passing along the pane. The light must be strong to avoid that the reproduction of the darker parts of the image suffer from noise, i. e. an irregular pixel pattern. Noise, also called “dark current”, sometimes also “grain”, is produced by an erroneous electronic information. According to photographer Tom Striewisch (2009, p. 44-45), noise is generated by light sensitive units on a CCD-sensor, they send weak charges without having received light. These charges are falsely read as light information. Generally visible in the dark parts, where they form clear dots, they are not noticed in the very light parts of the image. The light dots make details vanish and disturb the tonal harmony.

 

  1. depth of field

Flatbed scanners are made for documents that are “read” while in direct contact with the pane. As slides with protection glass (thickness of glass c. 1-2mm) or wooden frames (distance exterior frame – glass with image c. 3-4mm) have the image level at a certain distance to the pane, it is important that the focus can be adapted. The scanner has to have a sufficient depth of field, to make sure that image and carrier (both at a certain distance from each other) are equally sharp in the reproduction. Evidently, the image side of the slide has to be as close as possible to the pane (general focal point). Tests with thicker slides are imperative to learn about the scanner’s flexibility regarding its depth of field.

 

Post production software may be able to correct insufficient sharpness to a certain degree, but the sharpening tools are known to produce artefacts, e.g. light fringing along the edges of objects when the contrast is pushed. (For more on artefacts and the relation between image processing tools and increased noise see on Don Williams (2000) and Franziska Frey (2000).) Besides, a correction in sharpness give the slide a certain “unnatural” look, due to the enhancement of colour contrast, which stresses the lines in the image; the software does not correct the blurring, it only “simulates” sharpness by intensifying the difference between dark and light parts of the image. Photographic expert Freeman (2004, p. 257) recommends, in case this tool is needed, to use it at the end of all digital manipulations, as the increasing of the image’s contrast is not reversible. Similarly, the report on Benchmarking Art Images Interchange Cycles. Final Report 2011 (Frey 2011, p. 103-104) admits that sharpening an image may be necessary, as details can get out of focus due to the numerous steps in which files are handled. This should be done by the printer or the web designer with the purpose to go back to the original’s sharpness and richness in details.

 

  1. drivers and other accessories

A flatbed scanner is not usable without a computer by which the scanner settings are programmed, the files are saved, renamed and transcoded, the scan is made viewable on the monitor, the image quality is corrected in post production etc. A scanner is a peripheral hardware element (as is a camera or a printer) which communicates with the computer thanks to a software (“driver”) that is specially conceived. When buying a scanner second-hand, drivers have to be acquired as well, because not all older software packages can be found on the internet. Not all computers accept to read earlier versions, as today’s software producers do not guarantee backward compatibility for more than 2-3 product generations. Therefore, it is necessary to make sure that second-hand devices are capable of working with file formats that are currently in use, otherwise modern post production software may not be successfully implemented and applied. Also, hardware can create problems if e.g. connection cables cannot cope with the progress in data transfer technology, or changes in interfaces prevent communication between connected systems. The industry changes hardware and software regularly with sometimes fatal consequences as the Swiss conservation group Memoriav stresses: the “lack of software support can make fully functional hardware obsolete from one update to another” (Jarczyk et.al. 2017, p. 35). Thus, one obsolete element can ruin a perfectly functioning workflow. It is therefore imperative to test whether all devices work together, and to make sure they will do so in the future.  Besides, the time scale of renewals can be an indicator for longevity, the popularity of a product will help later to find “spare parts”. The use of open source solutions can be another way to keep the device future-proof.

 

  1. scanning speed

Scanning is a repetitive work and can be regarded as quite boring, especially when scanning a slide takes a certain time when belts or “rack and pinion systems” driven by an electrical motor move light source and mirror (CCD, CMOS) or the CIS’s “triple-unit” across the glass pane (https://itstillworks.com/flat-bed-scanner-work-4927087.html).  The mechanism is similar to a good old record player. A high scanning speed makes the act more varied, as the operator is constantly in action and does not spend time waiting. The higher the resolution, the slower the scanning speed. If anti-dust and anti-scratch software is involved, the pace decreases even more as the scratch-detection assigns areas to algorithms which correct the flaws.

 

  1. sensor resolution

The higher the number of pixels on a row, the more samples per inch can be taken and the more details will be registered. A high resolution allows the viewer to zoom into the image while details remain sharp. If the resolution is not high enough, the reproduction will quickly become “pixelated” and no longer look nice. A low resolution restricts the potential use (and may even force the archive to redo the reproduction with another device).

 

The resolution is determined by the sensor’s capacity to reproduce captured information in a high number of clearly distinguishable units (lines, dots). Manufacturers indicate the “nominal resolution” which is a theoretical, mathematically calculated indication and differs from the real “effective resolution”. Tests with test charts (presenting lines which are generally better visible than dots) have shown that the promised high “line rate” (lpi, lines per inch) is not kept. One of the best sensors had a “nominal resolution” of 5.000 ppi and a “sensational” effective resolution of 4.100 ppi (Wagner [2017]).

 

  1. colour depth

Promotional documentation on flatbed scanners state that they have a colour depth of between 24-bit and 48-bit. Indications such as 24-bit, 30-bit or 42-bit add up all three colour channels, thus the actual depth per colour is 24 : 3 = 8-bit, and accordingly 10-bit and 14-bit. (Some manufacturers express it in a roundabout way, e.g.  “24-bit colour, 16-bit grey scale”.) According to image scanner specialist Patrick Wagner, owner of a German scanner distribution company, this indication tells the user that the scanner’s analogue-to-digital transformer (A/D converter) is capable of transmitting the optical signals received from the sensor in each of its colour channels in 256 steps (3×8 = 24-bit) or in 65.536 nuances (3×16 = 48-bit) (Wagner [2017]). Kraus (199, p. 189) has calculated that am image with a colour depth of 16,7 million colour shades (256x256x256) would need a screen with a diameter of 64 inches (162,5cm) to be adequately reproduced. Nevertheless, a high bit depth guarantees a certain sustainability.

 

  1. additional automatic “services”

Tests with film scanners have shown that automatic “restoration” software can be wrong: it identifies image details as “errors” and “repairs” them, thus an element of the original is erased as a “mistake” or transformed for correction. Often it is better to correct flaws in post production manually (and not let the machine do it) to keep total control on what happens.

 

10.1 dust and scratch correction

Image scanners also use infra-red light for flaw detection. This only works for coloured, not for black and white slides. A tester of diapositive scanners (Kramer 2015, p. 124) welcomes this tool. Nevertheless, if the scanner proposes it as standard setting, it should be used with care as any automatic interference is based on algorithms. The location where the infra-red light detects dirt, dust or a scratch is marked by the software, which replaces the original pixels by copying and mixing information from neighbouring ones (interpolation). This can produce artefacts.

 

10.2. image enhancing

André Kramer, tester of diapositive scanners has mixed feelings as to colour correction and dynamic exposure extender, but was positive concerning automatic sharpening (2015, p. 124). Nevertheless, any automatic interference is pre-programmed by the manufacturers and therefore uncontrollable by the scanner operator. And it has consequences: colour correction can for instance reduce the potential colour range in post production. Image enhancing is better done by the professionals in charge of reproducing the slide for printing, presentation on the internet or on iPads used by visitors in museums. They are at the end of the workflow, know the needs of the consumers and will rework the scan with regard to the light conditions under which the reproduction will be presented. In the Benchmarking report the idea came up that, as light conditions influence the person looking at an object, the reproduction should be manipulated to “keep the appearance of the image at a consistent quality, independent of the viewing environment”: e.g. lighter reproductions for laptops with weak back-light, darker ones when the digital image is projected with a strong light bulb (Frey 2011, p. 124). The research has revealed that it is better for the use on the internet to enhance the scan’s contrast, as tests have shown that users evaluate lower contrast reproductions as “bad” images (Frey 2011, p. 106).