Image-editing is probably the application which most people will associate with desktop scanners and desktop computing. Anyone who has used Windows or one of the other graphic environments will probably be familiar with one of the many 'paintbrush' utilities and what it can do. However, its facilities will probably be limited to 'drawing' and 'painting' with a small number of solid colours and saving files in a limited number of formats. A dedicated image-editing package will expand on these facilities considerably, incorporating features such as:
the ability to acquire images from a scanner or other device
handling of full colour, palette colour and greyscale images (see the section on File formats), and conversion between them
masking tools for isolating areas of the image to be worked on
a wide range of drawing and painting tools in addition to the conventional paintbrush
transformation tools for rotating, mirroring, resizing and distorting areas of the image
{2}colour manipulation features for adjusting colour saturation, contrast, brightness, and other aspects of the image
numerous 'filters' for sharpening, blurring, pattern or noise removal, and other special effects
In addition, packages at the 'professional' end of the scale will be able to separate colour images into their cyan, magenta, yellow and black components (print 'progressives') for output to proofing and printing devices (see the sections on Colour theory and Output devices for more details).
At a basic level, these packages will scan an image and save it to disk for incorporation into a document prepared using a desktop publishing package. However, the manipulation facilities allow the image to be optimised beforehand. Blemishes on an original photograph - even creases or tears - can be eliminated completely without the need for manual retouching. Colours which are faint can be corrected or even changed - a car's paintwork could be changed from red to blue while preserving the shadows and highlights in the original picture. The possibilities for creativity (and for distortion of the truth) are enormous.
{3}Image-editing packages consume large amounts of memory: an image of 800 pixels wide by 600 pixels high in full 24-bit colour, will occupy approximately 1.4 MB of RAM. To perform any major manipulation of the image will require between two and three times this amount of memory in addition to the overhead of the application program and operating system. Therefore, about 8 MB of memory is a minimum requirement but the more memory available the better. Some of the better known image-editing and manipulation products are:
Picture Publisher (Micrografx)
Photoshop (Adobe)
PhotoStyler (Aldus)
Publisher's Paintbrush (Zsoft)
Image Assistant (Caere)
{4}Optical Character Recognition
The basic concept of OCR is straightforward - to convert pages of printed text into text files or word processing documents which can be stored on a PC and re-edited if necessary. This is a two-step process - first the page is scanned in as a graphic image, then pattern recognition algorithms are used to identify the letters on the page.
The scanning requirement for OCR is straightforward - usually a full page scan, at about 300 dpi, in black and white only. Some OCR packages have the ability to scan using greyscale mode - usually this is in the form of an 'auto brightness' or 'auto contrast' option. The extra information this provides can be used to compensate for coloured text or shaded backgrounds, or for recognising pictures embedded in the text. However, scanning in greyscale produces a much larger amount of data, which means that both the scanning and recognition processes will take longer.
{5}The OCR pattern recognition process is far more complex - the text on the page could be in a variety of different fonts, typestyles and sizes. It could also have been produced on a fax machine, in which case the letters may be badly formed with a messy background. While all of this is easy for the human eye and brain to decipher, it is a daunting task for a PC.
It would be fairly straightforward to write a routine to recognise a single typeface in a single size, but to produce a program flexible enough to cope with the needs of many different users, and to still do the job in a sensible timescale, requires very sophisticated and adaptable software. So-called 'neural network' systems are used, capable of 'learning' from what they find in the image, so that once the first few characters and lines have been interpreted the rest of the page can be handled more quickly, so long as the text is in the same style.
{6}In addition to the recognition of individual characters, the more advanced packages will also recognise page layout features. When the finished document is saved in the chosen word processor format, these features will be reproduced using the correct control codes for the application.
Most packages include options for pre-specifying a basic page layout - for example, the number of columns of text or whether the page contains a mixture of text and graphics. If this information is known in advance, the job of the package is made a lot easier. Most packages also support the option of multiple page scanning using a sheet feeder with the scanner.
A spell-checker will probably be included since, unless the original text is a clean printed page, it is unlikely that the recognition will always be perfect. Conventional word-processing spell-checkers are designed to find mistakes and suggest replacement words based on likely text-entry errors - like a finger mistyping an 'r' instead of a 't'. Spell-checkers provided with good OCR packages create replacement lists based on likely text-recognition errors - like a misprinted 'rn' recognised as an 'm'.
{7}Some popular OCR packages for the desktop environment (PC and Macintosh) are:
OmniPage (Caere)
WordScan (Calera)
TypeReader (ExperVision)
TextBridge (Xerox)
Recognita (SZKI)
Document Image Processing
Document Image Processing (DIP) has up to now been restricted to large corporate systems but is now becoming available on the desktop. Conceptually, it is the electronic filing and retrieval of paper-based information.
Pages are scanned in greyscale mode, usually at about 300 dpi. Then, powerful data compression techniques are used to reduce the images - each a complete rendition of a page - to a fraction of their original size for storage. Each document is then indexed by keywords and categories and stored in a database.
Usually a DIP database will be networked to allow many users to read documents stored on the system, and the ability to contribute new items either directly or indirectly. The database can grow to a large size very quickly, and often WORM (Write Once Read Many) optical media is used to contain it.
{8}The database can then be accessed using indexed search and the pages reproduced on screen in their original form (though usually in monochrome) for reading or reprinting.
Many DIP packages incorporate an OCR capability. This may be limited to picking out the most commonly used words and phrases as a guide to indexing, or it may be possible to perform OCR on the whole page and save the contents as a text file for incorporation into a report or spreadsheet. Several scanner manufacturers make monochrome desktop scanners specifically for DIP use, with high speed sheet feeders. Examples of popular DIP packages for desktop PC use are:
Pagekeeper (Caere)
KeyFile (Boston Software)
FilePlus (Greengage)
FileMagic (Westbrook)
Archis File (Softcore)
{9}Desktop Publishing (DTP)
Desktop Publishing (DTP) is now an established application for desktop computing. The ability to scan images on a desktop scanner incorporated into the DTP system brings text and graphics integration one step closer - previously, image scanning had to be sent out to specialist agencies, adding both time and cost penalties. With the emergence of the TWAIN standard for the control of imaging devices - including scanners - it is now possible to scan images directly into a DTP environment. This is only recommended where the image is a straightforward one which does not require any amendment before printing. Examples of widely-used DTP packages are:
