Development of Real-Time Adaptive Wavelet Transform Algorithms and Hardware Feasibility Study for 1-D and 2-D Pattern Recongition

Just as Fourier transform introduced spectral methods to process signal/image data for the last few decades, wavelet transform's multiresolution and real-time capabilities provide researchers with new tools for signal/image processing and related fields. In Phase I, we propose to develop an innovative and efficient real-time WT algorithms based on real-time quasi-interpolation and boundary spline wavelets for a Silicon Graphics workstation to process 1-D acoustic signal and 2-D visual/thermal imagery of Army ground targets, extracting or segmenting invariant cues and recognizable patterns (or textures). The real-time adaptive wavelets will be constructed from localized signals and patterns of interest for building these WT algorithms. Theoretical basis for such wavelets will be provided. We will also perform a thorough technical and economic feasibility study of hardware implementation of our adaptive WT algorithms using an off-the-shelf graphics workstation or microprocessor. Our research team includes highly trained engineers and internationally renowned experts in real-time algorithms/systems and wavelet theory/applications, ensuring a successful Phase I mission, leading to phase II. The Phase I effort has high dual-use potentials in the medical, environmental, acoustic, seismic, communications, and radar disciplines.