An overview of scanners
This document details the features of the typical flatbed scanner and what to look out for when choosing one for a digitisation project.
A scanner is a device that is used to capture objects by scanning them which produces a digital image file which you can then use in a range of ways. A scanner is used when you want to capture a real-world object.
The large fall in the cost of scanners over the last decade means that they are no longer the specialist item they once were. However the range of prices and specifications of scanners on the market can easily confuse the potential buyer. The different types of scanner (flatbed, film, drum and others) only adds to the confusion.
Before examining the different types of scanner, lets take a look at the underlying technology and consider some of the issues that need to be considered when choosing and using them.
Before choosing a scanner for a specific task you should consider:
- How large are the images to be scanned?
- How reflective are surface textures?
- Are some or all transparent?
- What are the conditions of your originals (are they fragile?)
- What quality do you require?
- What is your budget?
In recent years digital cameras have been used as an alternative to the use of a traditional scanner and thus most users will compare the two types of devices before embarking on a digitisation project. This guide will focus on scanners rather than digital cameras for scanning.
- A scanner is easier to operate and requires less technical knowledge to produce high quality, repeatable results than is required to use a pro-sumer digital camera. Common photographic problems such as focus, exposure or perspective errors are almost impossible with a scanner. Some libraries are installing simple publically operated ‘push button’ scanners for their users.
- The controlled lighting environment in a scanner eliminates exposure and colour temperature issues.
- Scanners can be set up to scan objects in batches, most flatbeds can scan several negatives, one at a time. Some models are designed to scan multipage documents automatically then pass the data through an Optical Character Recognition application to produce a readable and searchable digital version.
- Most scanners only require a small amount of desk space, cameras however require a formal or ‘make shift’ studio space with controlled lighting.
- Dedicated book scanners can digitise bound volumes without placing the object under stress.
- Less flexible than a camera, can normally only scan 2D objects. A camera can capture most types of original though additional lenses or other equipment might be required.
- Limited to a maximum scan area eg A4
- Specialist scanners can be very expensive £20,000+
- Scanners are typically deskbound. They are easily damaged in transit and so the original must be brought to the scanner, a camera in contrast can be taken to the original if necessary.
Choosing a scanner
Resolution is probably the specification that causes most confusion for newcomers to scanning. It is simply a measure of the ability to capture detail within the original work.
Scanner manufacturers normally specify two figures for this parameter: 'optical' and 'interpolated' - however only the 'optical' figure gives a true indication of a scanner's capabilities.
For scanners that use Charge Coupled Devices (CCD) as the sensor, the optical resolution is limited by the number of elements (pixels) in the detector array and how the array is moved relative to the image. Resolution is specified by the number of separate and distinct samples that the sensor can make in every inch and therefore measured in samples per inch (spi), although often the more generic 'dpi' (dots per inch) is given by manufacturers.
The figure given for 'interpolated' resolution is a result of the scanner software 'guessing' the values between pixels and presenting these intermediate values as 'real' values. This process gives no extra image information but merely smoothes the visual information and increases the size of the file. There are very few times when this process is in any way useful and when it is; it can be better undertaken within image processing software. So, remember that when assessing a scanner's resolution, it is only the 'optical' figure that should be taken into account.
Many scanners have their optical resolution specified as two figures eg. 1200 × 2400 spi. When this is the case, it is only the first and lower figure that gives the actual size (and resolution) of the CCD. The second figure is the smallest distance that the CCD element can be moved before it takes another reading (called the addressability). If this distance is smaller than the actual element itself then, again, interpolation is used to create the missing data. It is therefore only the lower figure that gives an honest guide of the resolution capabilities of the scanner.
As there is very little or no standardisation of presentation of an input device's specification, it is not surprising that manufacturers try every method that they can to make their scanners appear and sound as good as possible. When assessing the specifications of the scanner, it is important that one always makes sure that the figures are being compared in a standardised way to give a like-with-like comparison.
The 'colour depth' or 'bit depth' of a scanner is an indication of the range of colours that can be captured by the scanner (see JISC Digital Media's advice document: The Digital Still Image). It does not define the limits of the colour gamut (colour space) that is readable by the device (which is dependent upon the physical characteristics of the device) but simply specifies the number of separate definable colours that can be accurately described within it.
A higher figure will equate to a more accurate description of the colours available to the scanner but does not necessarily mean that they are available to the user at the end of process. Scanners will often capture at a larger 'bit depth' of 36-42Bit and then save or export from the scanner in standard 24Bit RGB colour. This extended colour depth is used internally by the scanner to produce the best possible quality original image data but is not normally available to the user. Although recently, there has been a move towards some scanners (with the appropriate software) allowing the full size 'hi-bit' version of the file to be saved and edited as a 48Bit TIFF or PNG.
Again the manufacturers have little agreement about the best way of specifying a scanner's colour depth. Some manufacturers give the number of bits for each colour channel (8-16 Bit/channel), whilst others prefer to give the number of bits for all colour channels (24-48Bit), either is accurate as long as you make sure you compare like-with-like.
The colour depth, in itself, does not provide much evidence of the quality of the scanner, however it does give some guide to how capable the scanner might be if it can use all the colour data it produces. The signal-to-noise ratio of the CCD sensors will always have more effect on the quality of the scanner's output. A scanner with a high signal-to-noise ratio and 30Bit colour depth will easily out perform a scanner with 42Bit colour and a low signal-to-noise ratio.
Recently, many low-end scanners are offering 36-42Bit A/D (analogue-digital) conversion although it is unlikely that the original source data created by the CCD within the scanner is really of a high enough quality to benefit from the additional bit-depth. Additional bit-depth allows a larger dynamic range to be captured by the scanner, but does not ensure that it exists within the original data collected by the CCD.
Dynamic range and dMax
The 'dynamic range' of a scanner is the range of densities in which it can differentiate detail within the original from the brightest highlight to the deepest shadow. The maximum density of the deepest shadow is called the 'dMax' and the minimum density of the clearest highlight the 'dMin'. The dynamic range is the difference between both values.
Dynamic range is measured on a logarithmic scale of optical density (OD) from 0.0 for a 'perfect' white to 4.0 for a very dense black. The quoted 'dynamic range' gives the range of OD values (between 0 - 4) that the scanner can distinguish. For a 'perfect' scanner this would of course be 4.0 (or just over) but in reality no scanner can come close, with cheaper flatbed scanners having a dynamic range of 2.5 - 3.0 and even the very best drum scanners only providing up to about 3.8. Of course images in the real world do not have perfect blacks or whites either and normally have a dynamic range significantly lower than 4.0.
A scanner with a 'dMin' of 0.2 and a 'dMax' of 3.8 would have a dynamic range of 3.6.
The 'dMax' can be considered a quantitative assessment of the scanners ability to read shadow detail.
|Scanner density ranges|
|Scanner||Typical dynamic range|
|Office grey-scale scanners||<2.5|
|24 Bit colour flatbed scanners (older models & low-end)||2.3 - 2.6|
|30 Bit colour flatbed scanners (midrange)||2.8 - 3.2|
|34-36 Bit colour flatbed scanners (high-end)||3.3 - 3.6|
|Desktop drum scanners||3.3 - 3.7|
|High-end drum scanners||3.4 - 3.8|
|Note: manufacturers often exaggerate the dynamic range values of their products. Values over 4 describe the maximum range that can potentially be contained within the scanned image rather than the dynamic range that can be captured by the scanner's CCD|
|Media density ranges|
|Source media||Typical dynamic range|
|Images on newsprint||0.9|
|Images on coated stock||1.5 - 1.9|
|Photographic Prints - normal (C-type)||1.6 - 2.0|
|Photographic Prints - High contrast (R-type, cibachrome)||2.0 - 2.3|
|Negative film||2.4 - 2.8|
|Colour transparency||2.8 - 3.2|
|High quality transparency and dupe film||3.4 - 3.8|
Optical density is measured in terms of image brightness with an optical densitometer. It is a logarithmic scale like that used by the Richter Scale for earthquakes. The density of 2.0 is ten times more than 1.0; An intensity ratio of 100:1 is 2.0 and 1000:1 is 3.0. An intensity range of 4.0 is mathematically possible but very rare to find in the real world, needing a intensity ratio of 10,000:1
Noise and crosstalk
'Noise' is defined as any unwanted energy within the signal and all electronic devices suffer to a greater or lesser extent. This unwanted energy may be caused by many faults in the scanner design including inadequate electronic shielding of the CCD and the use of cheaper components.
Noise is visible within a digital image as a grainy roughness within areas of very low signal (shadows). Visually this appears much like the graininess of photographic film. Often going unnoticed in 'routine' scanning, noise is most likely to show itself when shadow areas of an image are lightened or have their contrast range increased. If particularly bad, the noise can mask shadow detail and be visually unpleasant.
The level of this 'noise' is quantified by the signal/noise (s/n) ratio. Many manufacturers of low-end scanners prefer not to supply the s/n ratios for their products due to their revealing nature. Figures should certainly exceed 60dB for 8Bit/colour and 75dB for 12Bit/colour. Scanners based on CCD technology are particularly susceptible to noise and use various techniques to prevent it, such as multi-passing, where the image is scanned multiple times before the results are averaged to smooth out any apparent noise.
Crosstalk is a degradation of image quality where a flair appears around bright highlights in the scanned image, especially in dark areas. It is caused by the highly stimulated CCD elements (in the image highlights) distorting the readings in the elements (in shadow) immediately adjacent to them. Again this is a problem more associated with CCD technology than drum scanners. Crosstalk can be a real problem with cheaper CCD flatbed scanners but does not normally affect more high-end CCD flatbeds. Generally, crosstalk can only be discovered by testing the scanner with a range of test images with adjacent highlight/shadow borders.
Scanning times vary greatly between comparable devices let alone scanners using differing technology. Within a busy workflow, scanning speed can often be a deciding factor in scanner choice and should always be researched and considered before a choice is made. Many scanners offer a choice of differing qualities of scan which are dependent upon the number of passes and/or speed of the CCD: the more passes the CCD makes, the higher the quality and the slower the scanning speed. Some early CCD scanners were unable to scan Red, Green and Blue data in one go (one-pass) and had to make three separate scans (three-pass). This did not normally affect the quality but was very slow; luckily these units are now very rare.
Scanning technology is steadily improving and this is reflected in scanners that are becoming faster as well as providing higher quality. Of course as a general rule the faster a scanner is the more productive it can be. This is reflected in the costs of these units where faster scanners with higher productivity are considerably more expensive than slower units.
Most flatbed scanners have a nominal size of A4 but can scan an area of about 8.5" by 12-14". A3 sized scanners are available but they can take up a considerable amount of space. They are, of course, essential if you need to capture large works (over A4) although if the objects are very large or difficult to handle a digital camera might well offer a more pragmatic alternative. Hi-end A3 flatbed scanners are very popular with commercial digitisation bureaux as (with the appropriate software) they can be set up to scan a number of images at one go. This offers greatly increased efficiency and increased throughput. It should be noted however that these machines are very expensive and for maximum productivity, two machines should be alternated so that one scanner can be loaded whilst the other is scanning and vice versa.
Some flatbed scanners offer the addition of dual optics where the optical system can be switched to scan a 'sweet-zone', which offers a smaller scan area with a greatly increased resolution. This is normally of use when scanning small to medium sized transparencies within the full size of the scanner bed.
Dedicated film scanners provide a scan area appropriate for the size of slide/negative that they handle. Some additional productivity can be gained by the use of additional add-ons such as the film-strip holder and and other add-ons as mentioned below.
There are a range of optional add-on parts that can provide additional functionality and productivity for many mid-range to high-end scanners. Two of the most common options for flatbed scanners are the automatic sheet/transparency feeder (ASF/ATF) and the transparency media adapter (TMA).
Automatic sheet/transparency feeder (ASF/ATF)
An ASF or ATF is used to batch scan quantities of single sheets or transparencies. All originals must be the same size and will need to be scanned with automatic or generic settings. Although these units offer high productivity with large throughput, they will only work if the originals can be scanned without any individual attention.
It should also be remembered that the memory, storage and bandwidth requirements are always high when scanning and these feeders can create enough digital data in a small time period to require a computer, storage and network of the highest specifications.
Normally ASF/ATF are best for creating small and low quality scans, either 1-Bit Black and White images from text for later optical character recognition (OCR) or small scans for thumbnail creation.
Transparency Media Adapter (TMA)
Low-end and consumer flatbeds have been designed and optimised for scanning reflective artwork such as photos and drawings. However some can work with a TMA that provides an alternative light source within the scanner, which enables transparent artworks such as photo-slides and larger colour transparencies to be scanned. These units can give acceptable results with larger transparencies. However, in general, these cheaper flatbeds do not have the optical resolution (or quality) to create large and high-quality scans from a 35mm photo slide. If you only need small and low quality scans from slides/negatives, then a TMA may well remove the need to buy a dedicated film scanner. However, if you want a reasonable level of quality or have a steady flow of work, it will be necessary to buy a scanner dedicated to the job.
Several years ago SCSI (Small Computer Serial Interface) was the most common scanner/computer interface. However, it has largely been replaced by USB 1 and USB2 (Universal Serial Bus), developed by an IT industry consortium, and IEEE1394, known more commonly as Firewire (or iLink by Sony).
A few high-end scanners and peripherals are still using SCSI as an interface although FireWire is becoming dominant. Low-end scanners nearly all use USB, which has now become a generic standard interface on all currently supplied computers and operating systems. In addition to data transfer the six pin Firewire connection can also carry up to 45 watts of electricity allowing it to power some devices directly.
USB and FireWire both offer the advantages of being hot-swappable, allowing scanners to be connected and recognised whilst the computer is running; whereas SCSI is considered a fragile technology that can be easily damaged by incorrect or untimely connections to the computer. As a general rule, all SCSI devices should be connected and turned on before their host computer is booted up.
FireWire and USB are evolving standards that offer higher speeds each and every time a new version of the standard is released. However the speed of data transfer is rarely a limiting factor on a scanner where the bottleneck is rarely the interface and more often the scanning process itself.
The scanning process is controlled by scanner software. As well as driving the hardware that captures the image data and passes it on to the next stage of the image workflow, this software usually offers a range of image processing features. Scanner software can either be a device-specific program designed to work with one scanner, or a plug-in based on a driver interface such as TWAIN or ISIS, which can be accessed from within a host program.
The software can play an important role within a workflow in terms of productivity and quality of the scan, so it is important to consider how best to combine the work undertaken by scanning software with that done by image processing software.
In addition to setting resolution, scan area and colour/greyscale, reflective/ transmissive quality, the scanner software can also be used to control:
- ICC colour profile creation
- Colour optimisation
- Colour transformation (e.g. RGB to CMYK)
- Film type colour correction
- Descreening (removal of the matrix of dots in printed material)
- Tonal optimisation
- Automated dust/scratch removal
- Negative to positive image selection
- Scan quality control (lower speed scan for higher quality, vice versa)
- Image rotation (portrait/landscape, flip/mirror)
- Batch scanning
Using any of these facilities at the time of acquiring the image can save a lot of time in corrective manipulation later on in the workflow, but it is worth comparing the performance (both quality and speed) of these functions between the scanner software and the image processing software when deciding which is going to be more effective.
Additionally, a project's workflow will have a bearing on which is more appropriate. For example, some workflows may have a high throughput of similar image types that can all be acquired using the same scanner software settings, adjusting each image by the same criteria. Other projects may have a lower throughput, or may be scanning a wider range of source material, where it would be more beneficial to keep adjustment to a minimum at the time of capture and use image processing software to fine-tune each image on an individual basis. Where higher levels of optimisation are required, using an image manipulation program will be more versatile than relying on scanner software alone.
Colour management systems
For many users, however high the quality or advanced the specification, the scanned images are only useful, if it means that the colour gained from the scan is an honest representation of the colour in the original.
These days many scanners come with some form of colour management software. Normally this is based around 'open loop' colour management systems as proposed by the International Colour Consortium (ICC).
In this system a standard reflective colour test target (IT8.7/2) or the transparent (IT8.7/1) is scanned and the readings taken to create an exact record of the scanners colour characteristics. This ICC profile can then be used to accurately transform the colours within the original to a standard 'working colour space' that the image can be edited in.
Low-end and consumer scanners do not normally have the ability to create their own bespoke ICC colour profile but come supplied with a generic profile that has been created by the manufacturer for that model of scanner. This will not be as accurate, but should provide the chance of 'truer' colour than can be created without any profile. ICC Colour Profile creation software can also be sourced separately, allowing the creation of customised profiles for any scanner as well as colour profiles for monitor and any output device. (A fuller explanation of Colour Management and the ICC system can be found in the JISC Digital Media advice documents Colour Management and Colour and Resolution Targets).
The steady growth in digital imaging over the last ten years has lead to a vast range of professional and consumer scanners being available on the market. Quality and speed are steadily rising across the board whilst the costs fall. However, it remains true that although it is possible to buy fast low-quality scanners or slow high-quality scanners at a cheaper price, productive and high-quality scanners still tend to be very expensive.
There are three main types of scanner, although they are based on only two underlying internal technologies to convert light to a digital signal. The drum scanner uses traditional Photo-multiplier Tube (PMT) technology whilst the flatbed and transparency scanners use more recent Charge-Coupled Device (CCD) technology.
The drum scanner
Drum scanners use photo-multiplier tubes (PMT) to produce very high quality results. They typically have a density range of 3.4 - 4.0 with a 'dMax' at the top of that range. They can offer an optical resolution of up to 8000 samples per inch (spi). Drum scanners are the tool of choice for the print industry and normally only used by professional digitisation bureaux. This is due both to their expense and their complexity, requiring skilful operation to get the best from them.
Only flexible original artwork can be scanned in a drum scanner as it has to be 'mounted' on a transparent acrylic cylinder (drum) and then spun at high speeds around the photo-multipliers within the cylinder. Mounting transparencies on the drum is a slow and skilled operation and it is normal to have at least two drums in use so that one can be mounted whilst the other is being scanned.
Although the quality from these scanners is exemplary, they tend to be slow and cannot normally provide the level of productivity required from most digitisation projects. There are also some preservation issues with the standard use of a 'mounting oil' to avoid 'Newton's rings' between the transparency and the drum. If a mounting oil is used, then the transparencies must be scrupulously cleaned after scanning. This is quite possible, although many owners will not wish to not run the risk of treating their transparencies in this harsh way and will prefer to scan them on a flatbed that can provide a far safer capture process.
Drum scanners typically cost from tens to hundreds of thousands of pounds although there have recently been a few desktop drum scanners introduced at a more affordable rate. This kind of investment is only worthwhile within a highly commercial business structure, but will offer superlative quality to those that invest the money and the time to get the best out of them.
The flatbed scanner
Flatbed scanners use a linear CCD array, made up from a long line of charge-coupled device elements in a row. The CCDs themselves can only detect the presence or absence of light. To enable the scanner to capture colour, they must either make three passes with a Red, Green or Blue filter in front of the CCD or have 3 lines of CCD each with either a red, green or blue filter on top.
CCD based scanners are much cheaper to produce than PMT based scanners and also tend to be much easier to operate. They range from very cheap and low-end consumer devices (as low as £30) right up to professional quality devices with costs comparable with cheaper drum scanners (in the tens of thousands).
Recently CCD technology has improved in leaps and bounds and within all price ranges the quality has greatly improved. The most expensive professional flatbeds are still cheaper than traditional drum scanners but can now easily provide scans to a similar quality, while cheaper scanners offer easy image capture at a correspondingly lower cost.
For the digitisation project one of the main advantages of the flatbed scanner is that they are generally much simpler to use, allowing unskilled operators to effectively use them without many weeks of extensive training. The flatbed scanner is also much more acceptable in conservation terms as the original does not need to be bent around a drum (and then taped down!), nor does it need any mounting oil to avoid Newton's rings. Any artwork that can be safely placed face down flat on the scanner can be quickly and easily scanned.
Flatbeds also offer much faster scanning times than drum scanners. This, along with the lack of any required preparation (mounting transparency on drum with tape and oil) makes it much easier to establish a productive and speedy workflow.
Most high-end flatbed scanners normally offer the ability to scan both reflective and transparent originals (sometimes within a sweet-spot and sometimes over the whole scanning area), whilst mid-range devices normally need a separate transparency media adapter (TMA) to scan slides and transparencies. Most low-end flatbed scanners are only able to scan reflective originals and even if they can be made to work with transparencies are unlikely to have the necessary optical resolution to provide anything more than the most basic quality.
The optical resolution of flatbed scanners is largely dependent on how much you want to spend on them. High-end flatbeds are capable of producing 5,000 spi, almost as high as comparably costing drum scanners, whilst at the low-end of the spectrum, the available resolution of consumer devices appear to improve every month and currently stands at around 2400 spi.
The transparency scanner
Transparency and slide scanners again use CCD technology to scan original transparencies. Rather than attempting to provide reasonable quality over a comparatively large A4 bed these scanners are dedicated to just scanning a small transparency to a high quality at a much higher resolution. The smaller and cheaper transparency scanners only scan 35mm film but the larger more expensive models can scan all formats up to 5" × 4". The smaller film scanners are around the same price as a medium quality flatbed, while the larger models may cost several thousand pounds.
Transparency scanners can offer resolutions similar to high-end flatbeds with up to 5,000 spi although most low to mid-end transparency scanners typically scan at about 4,000 spi for 135mm slides.
Typically these devices offer a much higher quality scan than can be made with a low to mid-end flatbed scanner but at a price much lower than the high-end flatbed (around £500 - £1000). They are easy and relatively fast to operate with options such as Automatic Transparency Feeders enabling them to provide highly productive workflows.
Transparency scanners therefore offer a good alternative if you need to scan large number of slides/transparencies but do not have the time or budget to operate a high-end flatbed or drum scanner.
Book scanners/planetrary scanners
Books, newspapers and journals can be easily damaged in standard flatbed scanners. The book's binding or the fragile nature of the document may prevent the object from being held flat on the glass. This can result in uneven illumination and lack of sharpness, heavy books and brittle pages can also be damaged when placed face down in a scanner. The typical book scanner (also known as a planetary scanner) uses one or two high resolution digital cameras with two or more lights to provide even illumination. By using a camera to capture the image the object does not have to come into direct contact with the capture device.
Some book scanners use a single camera above the object, the book's spine and cover is supported in a frame and the photographed page is held parallel to the camera. A few book scanners use a V-shaped cradle and two cameras, one camera captures the right side, the other the left. The V-shaped cradle places a minimum of strain on the book’s binding and pages, the cameras are positioned so that they are parallel to the page they are capturing
Some of the more sophisticated book scanners offer automated operation, this might include page turning and correction for the curvature of the page. While a book scanner offering this level of functionality will cost tens, or even hundreds of thousands of pounds, a basic level system for one-off projects can be constructed quite easily with a consumer digital SLR and copystand.
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 input devices. Digital cameras offer a highly flexible alternative to the scanner and may be used for capturing large 2D objects, such as maps or for fragile items which may be damaged in the typical scanner.
Scanners range in cost and capability, from basic consumer items right through to very expensive and specialised professional devices. It is certainly advisable to undertake some rigorous research to ascertain what you require from the scanner before any decision is made. Remember the right scanner for one project is unlikely to be perfect for another, unless both happen to have an identical set of aims and a similar budget.
Whatever scanner is sourced, it is worth considering that however high its quality, it cannot be realised without a correspondingly high level of skill from the operator. In fact, as a rule of thumb the higher the quality (and cost) of the scanner, the more skill will be needed from the operator to realise the available quality. There is little point in investing a large amount of a project's budget in a high quality scanner, if there is no budget left to pay for a skilled person to operate it. The reverse is just as true: there is no point in employing a highly skilled operator if you cannot afford a scanner that will make best use of his/her skills.
Why not come to our BTEC Professional certificate in digital imaging to fnd out and get hands-on practical support.