NANOSCALE IMAGE PROCESSING - Bio-Inspired and Nanoscale Integrated Computing - page 515

19
NANOSCALE IMAGE
PROCESSING
Mary Mehrnoosh Eshaghian-Wilner and Shiva Navab
This chapter has two main sections. The first section concentrates on image
processing applications of nanoscale spin-wave architectures. More specifically,
how three image processing problems, namely, labeling, finding the convex hull,
and finding the nearest neighbor problem, can be solved on a spin-wave
reconfigurable mesh. The second section explains how to implement one of the
most widely used operations in image processing: discrete Fourier transform
(DFT). How the DFT spin-wave module can be optimized to implement a spin-
wave fast Fourier transform (FFT) module is also demonstrated.
19.1. IMAGE PROCESSING ALGORITHMS
Over the past few decades, several mesh-based parallel architectures such as mesh-
connected computer, mesh-of-trees, pyramid, mesh with multiple buses, reconfi-
gurable meshes, systolic meshes, and optical meshes have been considered for
performing low and intermediate-level computer vision tasks [1-20]. This is due to
the fact that a two dimensional image can be mapped in a straightforward fashion
onto a two-dimensional mesh. In particular, reconfigurable meshes have been
shown to be attractive computational engines due to the flexibility that the
reconfigurable bus offers. Here, we use the nanoscale reconfigurable mesh
architecture that is interconnected with ferromagnetic spin-wave buses, presented
in Chapter 7. As described in that chapter, the nanoscale spin-wave-based
reconfigurable mesh, while requiring the same number of switches as standard
reconfigurable meshes, is capable of simultaneously transmitting N waves on each
The authors of this chapter are listed alphabetically.