OPERATORS and THEIR APPLICATIONS

Asynchronous nonlinear fractal operators were considered in order to implement fractal applications by massive parallel asynchronous computing system as recurrent neural networks. Assuming the convergence of the synchronous fractal operator to an element ~ f which is an approximation of the original image f , it is proved that any asynchronous deterministic realization of local fractal operators is convergent to ~ f and the stochastic realization converges to ~ f with probability one. Beside compression, applications of such stochastic fractal operators for object recognition and image association are described. This new approach to the concept of associative memory, exhibits correct associative recalls for highly noised images and also for images with small 3D distortions (on Olivietti face database it results in 100% rate of face recognition).