A nano-scale architecture for constant time image processing

Abstract

In this paper, we study a nano-scale highly parallel computing architecture for image processing: a reconfigurable mesh that is interconnected with ferromagnetic spin-wave buses. In this architecture, information is encoded into the phase of the spin-wave, and interommunications are performed without any charge transfer. This spin-wave reconfigurable mesh, while requiring the same number of switches as standard reconfigurable mesh, is capable of simultaneously transmitting multiple waves on each of the spin-wave paths using different frequencies. Moreover, in this architecture, the superposition characteristic of spin-waves allows concurrent writes if all the requesting processors write a 1. Employing these parallel features, we present three constant time image processing applications for this architecture. We show that given an N 1/2 x N 1/2 image, using a N x N spin-wave reconfigurable mesh, in constant time, all figures. can be labelled, and the nearest neighbour, and the convex hull of a single figure can be found.