Image File Formats

There are currently many different ways in which images are regularly stored inside a computer, and for each of them there has to be a way of reading the image into some sort of viewing program to actually look at it, and the pixels have to be stored in an array in memory in order to process the image or subimage. If we wish to restrict ourselves to binary images, then a single bit can be used to store the value of a pixel at a particular location. A binary file can therefore be stored as a long list of bits, but it also needs some information to say how many there are and what the dimensions of the array are.

The graphics systems of different computers also vary a lot. This presents the writer of a text book on image processing or pattern recognition with certain problems. What machine does one suppose the reader has access to? What programming languages does one assume he can use, and what make of scanner or camera is he going to be able to get his hands on?

These problems are similar to the problems faced by a web based system too. In designing this course we had to make choices about what kind of hardware you would have, and what kind of operating system it would be running. The decision was to go with the most common systems used in industry as far as the hardware was concerned, and to discuss the software that could run on these systems. So we are going with the Intel Pentium based machines despite the many virtues of the alternatives.

It is possible to get around these matters by writing at a level of abstraction so high that the reader will not actually connect up what he reads with actual images or patterns, but I find this unappealing. I have agonised over these issues and concluded that there is not much use tackling serious pattern recognition without a UNIX workstation, and that X-Windows is currently the de facto standard for graphics. At the same time, PC's are widely available and many graphics acquisition cards are available for these machines at low price. Rather than choose between these competing alternatives, I have opted for both of them. I shall start off by supposing that the scanner is attached to a PC and that images can be stored in the common .tif files.

TIFF, short for Tag Image File Format, is a convention for storing a variety of images, and since it is a common one and most scanners can read and write such files, I shall refer only to such files. Other standards such as GIF, JPG, MPG and the page description languages can generally be converted, for a particular image, and I do not want to have to write programs for them all.

One of the many programs available with this splendid book, at no extra cost to enrolled students, is a TIF reader, which will take a small binary TIF file and display it on the screen of a PC. It stores the image internally in an array, which means that the enthusiast can monkey about with it to his hearts content. At a later stage in the book, the reader will be introduced to a somewhat beefier X-Windows variant of this program for reading larger images in many colours into a workstation. The reader should certainly acquire the art of scanning documents; it is easy and gives a satisfying sense of power for very low expenditure of effort.

Unless you live far from the madding crowd, there are almost certainly print shops near you which will do scanning, writing the tif file onto a floppy disk. If you do live far from the madding crowd, and don't already own one, scanners may be purchased via the Master's Course at this University at reasonable cost: email your tutor. Failing this, you can post us an image you want scanned and we can put it on a site for you to ftp back. There is a small charge for this service.

I shall suppose then that the images that follow can be obtained as .tif files output from a scanner, and that the transformation to an array of pixels is accomplished.