John Henshall gets into scanners
Now that you are aboard the Digital Express
it's time to look in more detail at the components of a digital imaging
system: acquisition, manipulation and output.
First we look at the acquisition of digital images - how to capture or translate your images into a form suitable for the computer.
The way of the future - already of the present for many photojournalists
and catalogue photographers - is direct acquisition using a digital camera.
Digital cameras are undoubtedly at the leading edge of technological development,
with a price tag to match. In effect these are miniaturised computers with
a lens and, although they have come a long way in the last five years, they
remain infant technology, not yet suitable for all applications.
Look again at the double page spread produced by a Dicomed scan back on a Sinar, reproduced in the June 1995 issue, for example. The quality is superb but the drawback is that such scan backs can only be used for subjects which do not move. Compare this with our April 1996 cover picture, using a Dicomed BigShot.
Here the drawback is that it is presently only available in black and
white. Even that is better than Henry Ford's 'Model T' - available in black
only. How far the automobile has come since then. For many applications
the quality of digital cameras is already more than acceptable and in some
ways they are already better than film.
LOW COST DIGITAL CAMERAS
In the past two years the availability of low cost digital colour cameras
has increased from just one - the Apple QuickTake 100 - to many, by Apple,
Canon, Casio, Chinon, Dycam, EPix, Epson, Fuji, Kodak, Kyocera-Yashica,
Logitech, Pixera, Ricoh, Ritz, Sanyo, Sony and others. New cameras seem
to appear almost every week, prices have come right down, the new Kodak
DC20 having a street price of only only £250 plus VAT.
Yet none of these cameras produce images as good as a 35mm compact -
even though some of them resemble compacts. But these digital cameras are
only the start and we are likely to see an immense amount of competition
in this field in the next few years. Someday low resolution cameras will
be as inexpensive as pocket calculators.
Make no mistake, low resolution digital cameras do have professional imaging uses - for example for reports, insurance claims, identity passes and driving licenses. Although they do not have the resolution to produce a professional size print, they do help you to learn about digital capture in a fun way. What ensures the future of digital cameras is the 'free' per shot cost once you have bought the camera.
Film has enjoyed a hundred and sixty years to mature to near perfection.
It crams the equivalent of a massive amount of digital storage onto a flexible,
thin piece of plastic of modest area. A transparency is unique in that the
same piece of plastic acquires, stores and outputs the image, requiring
no special equipment to view it.
You already have a wide range of camera equipment and the best way to make digital imaging work for you is to integrate it into your established working pattern, bringing you new creative tools.
Turning film originals into digital image computer files is a hybrid
technique which can offer the best of both worlds.
To get the image into a form understandable by a computer imaging program it must be analysed in terms of detail, tone and colour. It's like looking at it through a grid of squares and noting down the average brightness and colour of each square, then plotting them onto a piece of graph paper. The larger the number of squares, the greater the amount of detail - up to the limit of resolution of the film or print. Each of these squares is called a picture element or "pixel".
When this information is reassembled and displayed in grid form on a
computer monitor we have a digital representation of the image.
Kodak's Photo CD is a great way to get film digitised at reasonable cost.
The professional version will digitise up to 4x5 but there is a great deal
of variation in quality between bureaus and operators. Some would have you
believe that a Pro Photo CD scan is akin to a hand-made print but comparison
with a good machine print would be more accurate as the scanner cannot deal
selectively with different parts of the picture.
If you took our advice and bought a workstation with a CD ROM, Photo CD is a very cost-effective way of getting images digitised without purchasing any additional hardware at all, though you do have to send the work out.
The term 'scanner' has been used since television pioneer John Logie
Baird used a 30-mirror drum mounted in a wooden caravan to 'scan' the Epsom
Derby in 1931. Since then, television outside broadcast vehicles have always
been called 'scanners'.
The hardware which analyses a piece of film or print is also referred to as a scanner - a device which analyses and resolves all the parts of a picture into a prearranged sequence.
In digital imaging there are two main types of scanner: transmissive and flatbed.
Transmission scanners analyse pieces of film - negative or positive, black and white or colour - but cannot scan prints or artwork. They can turn negative direct to positive, without the need for an intermediate stage. They have a light source and an optical system, used to focus the image of the film onto a CCD (Charge Coupled Device) sensor which converts the light into tiny electrical signals.
Flatbed scanners are primarily used for scanning prints and artwork. The print is placed face downwards on a glass platen and a long line of CCD elements, as wide as the platen, then chugs off down the length of the platen, stopping every fraction of an inch to read off each row of signals reflected from the print. These signals are then transferred into the computer's memory.
Add a transparency hood, an optional replacement for the hinged lid, and you can also scan film. These units have a further light source which shines through the film. The light travels along inside the hood, in step with the CCD sensors below the platen. These hoods can double the price of less expensive scanners though, when it comes to scanning film, flatbeds never perform as well as a dedicated transmission scanner. One limitation is that there is no optical system to focus the long line of CCD elements so that can all be 'seen' by the transparency.
The number of pixels needed to analyse an image is one of the least understood
aspects of digital imaging. Basically, the more pixels in an image the more
detail we resolve. So we need the maximum number of pixels possible? No.
We need 'sufficient' for the image's intended purpose.
This is a particularly difficult concept for photographers to grasp. We are used to having masses of 'redundant' detail on our hands in film, which is usually enlarged or reproduced far below its resolution limit. The unused detail stays there on the film - in case it is needed later for some other purpose.
In digital imaging, storing spare resolution takes on a whole new expense and time dimension. Doubling the resolution quadruples the size of the image file - the square law at work again - which then takes four times as long to move about and to process.
HOW MUCH DO WE NEED?
To find out, let's start at the end - the output.
Say we want an 8x10 inch print. A typical dye-sublimation print head has 300 elements per inch. To cover this area of print therefore needs 300x8 or 2400 elements for the width, and 300x10 or 3000 elements for the length. Multiplying 2400 by 3000 gives 7,200,000 elements. That's how many pixels need to be output to the printer.
If the original is 8x10 inches we need to analyse it at 300 pixels per inch. Any more than this and we end up with redundant information which needs to be discarded before printing. So it would be a waste to scan the original at 600 pixels per inch.
If the original is 4x5 inches, scanning at 300 pixels per inch would result in a 4x5 inch print. If we want to produce an 8x10 print from the 4x5 inch print there are two options: resample - "res" - our 300ppi scan up to 600ppi, or scan the original at 600ppi.
"Res-ing up" a 300ppi scan involves the computer adding made-up picture information between the true detail it has scanned. This involves "interpolation", which averages adjacent pixels and then inserts a new pixel of this value between the existing ones. Some scanners will do this at the time of scanning, but you can always do it later, in an imaging program such as Adobe Photoshop.
It is particularly important to bear in mind that interpolation does not and cannot add real detail to the scan. Instead, it smooths out transitions between adjacent pixels, avoiding the 'jaggies', when the intention is to blow them up larger than their inherent detail would allow. Add a little carefully controlled sharpening and - subjectively - the result can actually look sharper than the original.
If we scan a 4x5 print at 600ppi we get 600x4 (2400) pixel analysis along one side of the print and 600x5 (3000) pixels along the other. These numbers are exactly the numbers required for an 8x10 print and this is a better solution than resampling because the resolution is real.
To get real detail we need a scanner with good optical resolution. Bear in mind, however, that a photographic print only holds a maximum of 200 line pairs of resolution per inch, so a 300ppi scanner would be adequate, 600ppi more than adequate.
Having a scanner with too high a resolution for normal purposes can actually be a disadvantage. A scanner which has an optical resolution of 1000ppi but is normally used to scans prints at 300ppi will miss 70% of the image information, performing its scan in narrow strips - or it will take longer resampling the scan down.
Devote the smallest "bit" of computer power to describing each
pixel of an image and that bit can either be 'on' or 'off'. These two states
can be made to represent black and white. Devote two bits to describe each
pixel and we get 2x2 combinations, or four tonal levels. The number of tonal
levels doubles for each added bit, so eight bits gives 256 levels per pixel.
This amount of descriptive power is required for each of red, green and
blue, making 24 bits per pixel 'bit depth' in total. This is the minimum
Some scanners analyse the image in more than 256 steps, to preserve the
subtle highlight and shadow tones in originals with large dynamic ranges.
For example, a 30 bit scanner increases the number of steps per colour from
256 to 1024 - from 8 f-stops to 10 f-stops.
Dynamic range is usually indicated as "D" and a number. Each D0.30 represents one -stop of range - a doubling of contrast ratio. Thus D1.80 is 6 f-stops or a 64 to 1 contrast ratio, D2.4 is 8 f-stops or 256 to 1 and D3.0 is 10 f-stops or 1024 to 1.
Go for as high a 'D' number as possible, especially if you want to get the best from transparencies. This information is sometimes difficult to find and you might have to look very closely at manufacturers' specifications to get it.
CHOOSING A SCANNER
The choice is huge, with flatbed scanners starting at less than £200.
At the low cost end, scanners all look quite similar - a low rectangular
box three or four inches high, a footprint slightly bigger than A4 and a
hinged lid. Names to look for include Agfa, Hewlett Packard, Microtek, Nikon
and Umax. Look for a bundled copy of Adobe Photoshop - the 'full' version
if possible. Scanners invariably come with their own software, usually a
'Plug-In' which provides a 'bridge' to Photoshop and 'drives' the scanner.
At its most basic, this software provides control over exposure and cropping.
Flatbed scanners have been around a long time, are versatile and there are plenty to choose from. Although they have some limitations, they have one overriding advantage: they are unbelievably inexpensive. You should therefore place a flatbed scanner right at the top of your digital shopping list.