One of the confusions that surround scanning and printing is the difference between DPI (Dots Per Inch) and PPI (Pixels Per Inch). The two terms are often used interchangeably, which they can be in pixel-based systems such as scanners, laser imagers, film writers & image processing software. However, in an ink-jet printer, a Dot is a fixed point of colour that can only be present or not, the colour of which can only be changed by using a different colour ink and whose size can only be varied if the print head allows. A Pixel is also a fixed point of colour, but its colour is variable depending on the colour depth in use. So where as a dot in an ink-jet printer may only be the colour of the four (or in some printers, six) inks in use, a pixel may be any shade and colour in the colour space in use. So in an 8 bit per channel (24-bit) colour system each pixel may be any one of 16,777,216 colours, against the four or six colours available in an ink-jet printer.
So unless a file is being printed on a pixel based printer (such as a film writer or a laser imager) any file has to be converted, or mapped, by the printer or its driver software into a dot pattern using a screening technique such as frequency modulation, error diffusion or Stochastic. Exactly what happens will depend on the ratio between the PPI of the file and the DPI of the printer, the colour & density of adjacent pixels and the screening method used. Therefore, except for the pure ink colours, this mapping ratio can never be 1:1; one file pixel can never be printed as one ink dot. This has the effect that once a certain PPI to DPI ratio is reached, any further increase will not improve either the detail or tonal range present in the print. Indeed, increasing the file size too far can cause a reduction in quality as the printer driver struggles to ‘boil-down’ the file to a sensible size, often accompanied by a reduction in printer speed. This situation is not changed significantly by six colour ink-jet printers because the two additional inks, light cyan and light magenta, are used to increase the number of dots of ink in light coloured areas. This is due to any ink-jet printer having to use the white of the paper to dilute ink colour strength and the only way that this is possible is by increasing the spacing between each dot of ink, which both reduces detail and leads to a coarse, dotty appearance.
Because manufacturers do not publish mapping ratios for their products, these have to be found by trial and error and it is generally accepted that the DPI of the printer can be halved or even quartered when setting the PPI of the file. Even this has a significant effect on file sizes since halving the PPI of a file reduces the file size in MB by a quarter. In tests on an Epson 1520 1,440 DPI printer it has been shown that the mapping ratio is approximately 1:10, allowing a file of 144 PPI at ‘same size’ reproduction to be used. So for example, an A3 print at 1,440 PPI gives a file size of 346.2MB, at 720 PPI gives a file size of 211.6MB and at 144 PPI a file size of 8.5MB. Whilst it is reasonable to use a bit larger file size "just to make sure" (particularly since exactly how an image is mapped will depend on its brightness and contrast overall, the contrast of detail within it), it makes no sense whatsoever to incur the costs of scanning, processing, storing and printing files that are much larger than required. As in conventional photography, the secret is to carry out your own tests to see what works for the combination of equipment and type of images you are using. Where there is the possibility of the image being printed on different devices which require a different file PPI, for example an ink-jet printer and a laser imager, you will always get better results by creating and working on the file at the larger file size, then reducing it for the smaller.
Resolution
The figure that is most often seen as reflecting the quality of a scanner is its resolution. This is divided into optical and interpolated resolution. Unfortunately, it is quite easy to produce a scanner that has a moderately high optical resolution, since this can be measured not by the true optical resolving power of the scanner, but by the image to CCD ratio. So a scanner can claim to be, for example, ‘1200 DPI’ because the CCD will produce this without interpolation (a process where additional pixels are created based on a calculation performed on adjacent pixels), but the images produced can be inferior to a 600 PPI scanner because the lens and/or mirror system is of too low a quality. This is because CCDs are much cheaper to produce than high quality optics. Interpolated resolution only produces a larger file size without any increase in detail. Scanning at PPIs which are not a multiple of the optical resolution of the scanner can also lead to a degraded image, due to the file having to be interpolated to the intermediate resolution. It should be noted that, the larger the original is the lower the resolution required to produce a file of a certain size. For example, although a 35mm transparency would need to be scanned at 4,800 PPI to produce a file size of some 80MB, a 5x4 would only need 1,200 PPI. It should also be noted that there is little point in scanning an original to a resolution higher than its inherent resolution. In the previous case, depending on film type, a scan of a 35mm transparency at 4,800 PPI may not reveal any more detail or tonal range than one at 2,400 PPI since all that is being scanned is the film grain. Indeed, the higher resolution scan may look worse than the lower, particularly when filters such as unsharp masking are applied, since it is the image of the grain that is being filtered and not the image itself. It has long been known by conventional photographic printers that when making very high magnification enlargements from any film format, the look and tonal smoothness of the print can be improved by slightly defocusing the image to suppress the grain.
Density range
The density range of a scanner is the maximum difference between the lightest and darkest areas in an original that it can capture. In photographic terms it is the contrast latitude of the scanner. It affects its ability to capture variations in image tone and density in both the highlights and shadows. By itself it does not indicate the quality of a scanner, for its affect on scan quality depends on the density range of the original being scanned. A machine that is to be used only for scanning prints or colour negatives does not need a wide density range because prints and colour negatives do not have wide ranges. However, transparencies do tend to have a wide density range and B&W negatives can fall just about anywhere since their density depends on factors such as the film used and the development received. Providing a high-density range is quite a costly thing to do since it needs both quality optics low in flare and quality CCDs that have both a wide dynamic range and low dark noise. As with resolution, there is nothing to be gained from using a scanner that has a range wider than the originals being scanned.
Scanner magnification
Magnification on scanners falls into two categories, optical (true magnification) and image size (scaling). Optical magnification requires that the scanner has an optical system equipped with either a zoom lens or a lens that can be moved further away from the CCD and refocused to provide magnification of the image falling on the CCD. This is costly and very few machines provide this facility, however it is the best way of providing different resolutions since no interpolation or sensor ganging is required. Image size magnification or scaling is simply where the width and height parameters in the file header are changed to indicate a larger or smaller image size. This has no affect on the quality of the scan or the file size, but is simply a convenience for those users, such as repro houses, whose software or hardware require image size to be indicated at a fixed PPI to avoid having to waste time and effort changing it later.
Types of scanner
A huge amount of misinformation abounds regarding types of scanners. The epitome of scanning is usually taken to be the drum scanner. Flatbed scanners are often dismissed as lower quality than a drum solely on the fact that they have a sheet of glass between the original and the optics. This shows a woeful ignorance since, as anyone who has used, or even seen in use, a drum scanner will know, not only does the plastic or glass drum that the original is mounted on sit in the light path, but so does the sheet of acetate placed on top of the original needed to retain the scan oil normally used. Some drum systems overcome this by mounting the original on the inside of the drum, but the drum still is in the light path. Some film scanners adopt a glassless flatbed approach, which by removing the glass exposes the, less easy to clean, optics to the dust that would otherwise fall on the flatbed glass. The type of scanner has virtually no impact on the quality of a scan, it is how that system is implemented and used that is important. The actual quality of a drum scanner comes, not so much from the resolution or the density range it can achieve, but simply from the fact that they are normally operated by experienced people who take the time and trouble to use their experience to produce the best scan for the intended use, rather than leaving it to the hardware and software to do it for them.
Other factors affecting image quality
There are several factors that can degrade the quality of a scan regardless of the quality of the scanner. Transparencies or prints on flatbed scanners should always have a mask placed round them to reduce light spill and the resulting flare, and the scan area should be set to include only the image and not any surrounding mask or film edge. This is particularly important when allowing software that is not ‘frame edge aware’ to select the high and low endpoints, otherwise false endpoints can be set (the frame edge should never be used as an endpoint). Ideally, software should have the option for the user to select endpoints for each scan. It is always better, particularly with high bit depth scanners, to get the end points optimised before scanning rather than adjusting levels later in PhotoShop since this can never ‘recover’ information that was not captured in the first place. For reflective scans (prints etc.), it's best to have a neutral dark grey backing, such as a sheet of paper or card, and not white or black. Once again this stops the software choosing the backing as an endpoint if it's included in the scan area. The contrast of the images on some thin papers can be reduced because of the dark background 'shooting through' the paper, but this isn't normally a problem. A problem that can be caused by software is that of automatic cleaning of a scan. Although some systems can be quite efficient at dust and scratch removal, there is always the possibility that it will mistake detail in the actual image and remove it. Therefore, as in the case of auto-focusing and auto-exposure on cameras, it is a function that needs to be used wisely rather than blindly. A build-up of dust on the optical components of a scanner can have a severe consequence, slowly and unnoticeably degrading quality.
Once the scan has been made, the choice of file format can have a serious affect on quality. Use of any lossy compression system, such as JPEG will cause both fine detail and smoothness of tone to be lost and compression artefacts to be added, the degree of which depends on the file size, chosen compression method and parameters used. If compression is needed because of lack of disk space, transmission requirements or some other reason, then either TIFF with LZW, a lossless wavelet format or compression software such as Stuffit is best be used, unless the recipient is happy or prefers JPEG. Clumsy adjustments, such as colour balancing, levels, brightness and contrast, performed in image manipulation software can seriously degrade the quality of a scan. It should be remembered that a digital image file consists of a collection of numbers that describe that image to the various software and hardware it meets. If those numbers are changed, for example by increasing the contrast or brightness of an image, there is the possibility that some will be lost. This means that subtle highlight and shadow detail can be lost or colour casts can be introduced, particularly into highlights. There is also the possibility that intermediate values will be lost as a file is adjusted, coarsening the tonal range. What happens to any particular image will depend on that image and what is done to it. This can easily be demonstrated by taking an image file and wildly over-adjusting it, saving it, then reopening and trying to adjust it back to match the original.
A further factor that can affect the effective quality of a scan is the method of output. For example, in an ink-jet print, the visibility of ink dots and the smoothness of tone has a great affect. This is dependant both on the DPI of the printer and the paper used. The more a paper allows the dot of ink to spread or bleed, the less dotty the print will look and, up to the point that the ink mixes and changes colour, the smoother the tone will look. However, it can make the print appear less sharp and some ‘art’ papers will also impose their own texture on the print. Banding, usually caused by uneven paper transport and which can vary depending on the paper used, can ruin an otherwise good print.
© Barry Leighton FRPS