A Processor Array for Computing Wavelet Energies

Abstract

Efficient remediation of sites contaminated with buried unexploded ordnance (UXO) requires highly accurate discrimination between intact ordnance and fragments or other objects. The energies at each level of the discrete wavelet transformation (DWT) of electromagnetic signal responses from subsurface objects have been shown to be characteristic of some types of ordnance. This paper investigates low-level parallel processing of energy calculations that may form part of a classification system that uses energies as characteristic features. Existing designs for the DWT are discussed as well as specific architectural requirements for the remediation application. The proposed design is described in terms of space-time diagrams, recurrence equations, scheduling and allocation functions.