Group-normalized wavelet packets for target extraction

Through the analysis, remarks on the time-frequency characteristics of wavelets and wavelet packets, and a modified adaptive wavelet analysis (group-normalized wavelet packet transform (GNWPT)) are presented to correct the drawback that the magnitude of the transform coefficients cannot exactly indicate the strength of the signal components. With the time-frequency masking technique, signal extraction and detection from formidable clutter or noises can be carried out well by GNWPT. Besides the choice of linear phased nonorthogonal FIR subband filter banks, the extended l/sup p/-norm entropy presented here improves the performance of GNWPT. A similar method can be applied in image enhancement, clutter and noise suppression, optimal detection and radar imaging.