We introduce new techniques to generate real-world images. Many of the
natural images exhibit a hierarchical structure and a balanced combin
ation of both order and chaos. We show how a controlled use of determi
nistic chaos yields a powerful method to concisely describe and to eff
iciently generate complex realistic images. The basic technique is the
use of Mutually Recursive Function Systems (MRFS) possibly with addit
ional control on the order of computations. Then we show that, surpris
ingly, regular sets of control sequences are convenient but not essent
ial. Some examples are presented illustrating the power of the techniq
ue.