A.L. Reznik, V.M. Efimov, and S.T. Vas'kov (Russia)

Fast program algorithm, probability, digital imageprocessing, rectangular lattice, pulse-noisy cluster.

Imagine a rectangular lattice with nodes (pixels) connected by vertical and horizontal "bonds". At a certain time moment each node is "infected" by a pulse noise with equal probability P. Then the probability for a node to be "uninfected" is 1-P (see Fig. 1). It is necessary to find probability Pr(M,P) of a noise cluster occurrence occupying "area" of M elements, i.e. the probability of that each taken node (let us consider it as a center of Cartesian coordinates (0,0)) gets into the noise cluster consisting of exactly M "infected" elements (including itself), connected with vertical or horizontal bonds with neighbor nodes. Fast program algorithms are proposed to estimate the probability of definite size noisy clusters occurrence when digital images to be processed are distorted by non correlated pulse noise.

Important Links:

Go Back