Generalized Asynchronous Iterations

Asynchronous iterative methods for multiprocessors are generalized to relaxation techniques involving discrete variables. Asynchronous algorithms are more efficient than synchronized algorithms in multiprocessors because processes do not have to wait on each other at synchronization points. Sufficient conditions for the convergence of generalized asynchronous iterations are given and proved. Applications of the theory presented in this paper include asynchronous relaxation algorithms for scene labeling in image processing applications.