A PARALLEL ARCHITECTURE FOR CURVE-EVOLUTION PARTIAL DIFFERENTIAL EQUATIONS

Abstract

The computation of the distance function is a crucial and limiting element in many applications of image processing. This is particularly true for the PDE-based methods, where the distance is used to compute various geometric properties of the travelling curve. Massive Marchinga is a parallel algorithm computing the distance function by propagating the solution from the sources and permitting simultaneous spreading of component labels in the infiuence zones. Its hardware implementation is conceivable as no sorted data structures are used. The feasibility is demonstrated here on a set of parallely-operating Processing Units arranged in a linear array. The text concludes by a study of the accuracy and the implementation cost.

Keywords

• distance function
• hardware for image processing
• partial differential equations
• parallel computing