Image compression tries to make an image simpler, and using fewer pixels, but without losing the meaning of the original picture. One of the many ways used to compress these images is through fractals.
Fractal geometry is based on the idea that certain systems can not be easily modeled by linear methods. Their shape is both too irregular and yet self-similar. For example, when modeling a cloud, there at first seems to be too much complexity to make a detailed description of its shape. Throughout the cloud, sections are very similar to one another. A tiny corner of the cloud has much the same shape as the cloud as a whole. It is as if on each cloud, there are hundreds of other clouds, each with their own hundred clouds. This series repeats forever, giving a clearer and clearer model the more iterations are computed.
This self-similarity appears everywhere in nature, making fractals a good model for almost everything in the natural world. Fractal image compression attempts to work backward, from a picture of a complex and repeating world to a smaller, simpler picture that seems to show the same information. The program looks for sections of the image that look similar, regardless of their respective size. It then works to split those into sections of color, representing the original image.
Unfortunately, fractal image compression still needs some work. It still takes too long to encode an image for something like a video. Most pictures do not have enough similarity to make this system efficient. If you are willing to wait a few more seconds, however, fractals seem to be the answer to optimal image compression in the future.