Since scanned images are just streams of pixels, it should, in theory, be easy to transfer them between different types of system. In reality, unfortunately, there are numerous 'standard' formats for storing and transferring images, each with their own advantages and disadvantages. Some of the more important formats are covered here.
Image types
Full colour images
'True colour' images are typically stored as RGB (red-green-blue) data, with 24 bits per pixel - 8 bits for each of the three primary colours, giving 256 possible shades of each or 16.7 million shades in total. (Colour theory is discussed in some detail later in this booklet). This data format is basically the same as that provided by the scanner. True colour files are versatile since all the data that was originally provided by the scanner is still present. However, the files can take up a lot of storage space, particularly at high resolutions. For example, an A4 page at 300 dpi will occupy 25 MB of disk space.
{2}Some applications provide the ability to convert files from RGB to CMYK (Cyan, Magenta, Yellow and Black), and to edit and save images in that format. This may be useful if the saved image is going to be used directly in a colour printing process, since the CMYK colours are the ones used in colour printing.
However, the data files created are even larger than RGB files because an additional 8 bits are used for the Black channel, making 32 bits per pixel in all. The RGB and CMYK colour models are discussed in more detail in the Colour theory section later in this booklet.
Palette colour images
In a desktop computing environment, '256-colour' images are often used instead of true colour images both to save space and because they are easy to support on inexpensive PC video displays.
{3}Eight bits per pixel are used, but instead of splitting the eight bits into red, green and blue components, each of the 256 possible combinations is assigned a colour shade selected from within the available range of colours.
This range may depend on the hardware in use but is typically over 250,000 shades. In this way a 'palette' of 256 colours is created.
The palette map, identifying which colour shade is to be used for each of the available 256 colours, is stored along with the image data when a 256-colour file is saved.
These files are a compromise for several reasons. Firstly, for most 'photographic' type images, 256 colours are not enough by themselves to create a convincing representation (let alone 16, as is common on many PC screens). This means that newspaper-style dithering techniques must be used which, while satisfactory up to a point, are clearly visible on close inspection. Secondly, the quality of the image is dependent on the ability of the package which is creating it to analyse the original correctly, pick the most appropriate combination of colours and make the best use of dithering. Software packages vary widely in their performance in this area and you may find that even the most sophisticated packages have problems making skin tones look convincing.
{4}Greyscale images
Monochrome images tend to be smaller and simpler to handle. Greyscale images usually contain 8 bits per pixel, giving 256 levels of grey between solid white and solid black, or 4 bits per pixel giving 16 grey levels. 256 levels of grey are enough to give a good representation of source material in monochrome.
Line art images
'Line art' images are stored using just 1 bit per pixel, which is usually either black or white. This is perfectly adequate for images such as pages of text, simple cartoons and line drawings.
{5}Data compression techniques
An image scanned in 'full colour' needs 24 bits (3 bytes) for every pixel and is three times as large as an image scanned in 256 shades of grey. An 800 x 600 screen-sized full colour image will use 1.4 MB, and an A4-size image in full colour at 300 dpi will use nearly 25 MB.
There are a number of schemes in use to reduce the storage overhead of these files. They basically fall into two categories; 'lossless' compression which preserves every pixel in the image intact, and 'lossy' compression which makes compromises with the data to improve the compression ratio.
Non-lossy compression
The most common techniques use established data compression algorithms such as LZW (Lempel-Ziv/Welch) or RLE (Run Length Extended). Depending on the nature of the image the compression ratio will typically be 2:1 or 3:1. This is usually acceptable for day to day work but may be unsuitable for archiving large numbers of images. In addition, a number of companies modified the algorithms slightly when implementing these techniques, with the result that an LZW compressed file saved on one system may not always be readable on another.
{6}Lossy compression
Lossy compression utilises the fact that there is a lot of low-level 'background' information in a typical image. This makes lossless compression less efficient, is actually not necessary for a convincing reproduction of the image and, in some cases, may not be visible at all. A mathematical process called Discrete Cosine Transform (DCT) is used which allows the amount of compression to be chosen. The more compression used, the more the data in the original image will be compromised. However, a typical compression ratio of around 15:1 will still give good image quality while reducing the amount of storage space required dramatically.
The Joint Photographic Experts Group (JPEG) was formed to devise an ISO/CCITT standard for image compression techniques encompassing both lossy and non-lossy algorithms. However, the JPEG acronym is most often taken to refer to the standard for lossy compression which is now supported by an increasing number of software products.
Fractal compression
Fractal compression is a means of analysing an image to determine how it can be represented as a sequence of complex mathematical expressions, instead of as bitmap data. The fern is an ideal image to explain the concept of fractal compression.
{7}If you look at the detailed relationships of a single frond of a fern - first with its leaf and stem, and then with other fronds on other leaves and stems - you can quite quickly build up pattern information which can be used to describe the structure of the whole plant.
Fractal compression is more a 'conversion' process than a simple compression process and requires a great deal of computer processing power to analyse the image and perform the conversion in a reasonable time.
{8}To carry out fractal compression on a PC, a dedicated co-processor board with a high-speed RISC processor is generally used. The advantage of fractal compression is a very high 'compression' ratio, typically as much as 40:1, with minimal loss of image information (less than with JPEG). In addition, the transformed image can be 'decompressed' fairly quickly using a software routine without the expensive hardware used to create it.
Common file formats
Choosing the correct file format is fundamental to transferring images successfully between different applications. Some of the more established standards and their main characteristics are detailed in the table opposite.
In a PC environment, the variety of file types and formats can be a nuisance. Some software packages can create some file formats but not read them, and vice versa. Some packages implement non-standard versions of the compression algorithms, which can prevent other packages from reading the files they have created.
Not all packages will handle all the variations catered for by a particular format - for example, some will handle RGB JPEG but not CMYK.
{9}The TIFF format has become an industry standard, but even this is a moving standard. TIFF revision 6 supports JPEG-type compression as part of its specification, but a file so- created will not be readable by software only supporting TIFF revision 5.
Key to image types:
R- RGB 24-bit full colour
C- CMYK 32-bit full colour
5- RGB16-bit colour (5 bits/pixel)
P- RGB Palette colour
G- Greyscale
B- Black and white line art
If image files are to be shared between different systems, or copies distributed electronically or by post, it is best to stick to one of the more common formats and to avoid CMYK files or unusual compression options.
{10}Vector Files
Not all types of graphics files contain 'bitmap' data. CAD software and tracing packages, which work with lines and shapes, usually store their drawing files in vector format. Instead of a mass of individual pixels, each element of the picture is represented by a start point and an end point (for a line) or a set of location coordinates (for an object). This enables individual parts of the drawing to be altered, moved around relative to each other or deleted without affecting the others. It also keeps the file sizes down to manageable proportions.
However, image scanners scan a line at a time before sending data back to the computer and are not able to analyse the pages they have scanned. This means that it is necessary to use special tracing or vectorising software to create a vector file from a scanned image.
{11}Storage requirements and media
As described above, high definition images, particularly colour images, take up a lot of space. The examples below show the typical amounts of data involved before data compression is applied:
Until recently, the storage of large numbers of images, each of several megabytes in size, was a problem particularly when it came to distribution. However, the rapid development of mass storage media has simplified matters considerably.
CD-ROM - Compact Disc Read Only Memory
Taking its name from the ROM memory devices used in computers, CD-ROM is a development of the audio CD technology pioneered by Philips and Sony which is finally being accepted in the mass market. Around 650 MB of data can be stored on a single CD and the discs can be mass-produced in the same way as audio CDs. There are several standards for the data format on a CD-ROM, one of which is Photo CD, developed by Kodak specifically for image storage.
{12}Compact Disc Recordable CD-R
One of several kinds of WORM storage (Write Once Read Many), CD-R is exactly as its name implies and it enables an organisation to produce its own CD-ROM discs for archiving or distribution. The discs can be written in any of the standard formats for CD-ROM and it is also possible to generate audio discs from a digital audio source (such as another CD, or a Digital Audio Tape (DAT) player). The discs produced are completely compatible with conventional mass-produced CDs. However, the blank discs are more expensive than mass-produced CDs and the special CD recorder needed to produce the discs is many times the cost of a conventional CD player.
Rewritable Optical Disk
Developed by Sony, and sometimes referred to as 'floptical' disks, these come in 3.5" and 5.25" sizes which can contain over 120 MB and over 600 MB of data respectively. They are similar to a 3.5" floppy disks in both construction and appearance.
{13}Hard Disks
Hard disks are a well-established storage media for large files. Many removable hard disks are available which use special cartridges on which you can store up to 88 MB of data. Although their capacity is lower than that of optical disks, they are less expensive and their speed of access is higher.
