Fractal image compression?

Question

Programming

Hello!
An urgent question about this image compression algorithm.
I read a lot of articles today, including the most simple one on Habré: habrahabr.ru/post/126653/
Feeling that all the articles are not written to the end.
Two questions:
-What if the ranked block is not similar to any domain block?
-What is in the compressed file? One black pixel, which turns into a domain block through transformations, and, recursively, new domain blocks are built, and so on up to the level of ranked blocks?
Help me understand please. Thank you.

Answer 1

Most likely an ordinary picture, where in place of some blocks there are links to other blocks from which they are obtained (well, it is written exactly how). And those that do not work out - they are - on their own.

Answer 2

Well, everything is simple here. When dividing the image into domain and ranking ones, we need to leave as little as possible, in fact, domain ones. Modern algorithms look something like this - there are no blocks on the images, we take two random blocks and try to choose a rank block from them. It turned out cool. No, bad. This is suitable for any images, on some special ones you can try to find the structure and somehow proceed from it. In fact, the algorithm may even work for a very long time, it stops again on a whim. For example, when the compression suddenly became a thousandfold, in general, you can not continue.
In general, the topic is very complex and sometimes incredibly useless. It is still impossible to compress stronger than entropy, and modern compression methods already show themselves very well. Moreover, it would be lossless, but the errors are not encouraging at all. Increasing the accuracy - you increase the size. Vicious circle. Well, there are still problems with the literature - there are few methods, there are many affine transformations, and the complexity is high. So walk around here, train your miserable semblance of reason, hoping for enlightenment... Oh... What if you calculate the correlation by the method... No, it's nonsense. That's how we live.