1998
[Back to Software in Reality or Home page]
One of the pioneers of fractal image compression is Barnsley, with his firm Iterated systems. You can have a look at one of my compressed pictures (17Kbyte), and compare them with a similar gif (12 Kbyte) or wif version (4 Kbyte).(wif compression by Cengines.com).
There is also another type of compression, based on wavelets. An example of that would be a wif picture (16 Kbytes), compared to a jpg version (52 Kbytes).
An analysis of image compression algorithms reveals some underlying assumptions about the structure of the data, and in particular assumptions about the structure of the interpretational systems: the eye and the mind.
The gif format compresses by replacing long sequences of identical pixels with a special code. This assumes that the picture has a lot of very evenly colored patches. This tends to be true for line art and other diagrams.
The jpeg format compresses by taking a spacial fourier transform of the picture and representing the higher frequency image components with fewer bits. This works because the eye is not set up very well to detect absolute color differences, so that two adjacent pixels that have different colors are, well, different. On the other hand, the eye has built in edge-detectors, so that if you have a surface that gradually changes hue, will appear rippled if this smoothness is not represented in a precise way.
Wavelet compression is basically an improvement on this.
Fractal compression tries to exploit self similarities in the image data.
[Back to Software in Reality or Home page]
Last modified 21-okt-99.
