Optimal matched wavelet construction and its application to image pattern recognition

作者: Joseph O. Chapa , Raghuveer M. Rao

DOI: 10.1117/12.205374

关键词: Fast wavelet transformSecond-generation wavelet transformPattern recognitionMathematicsMultiresolution analysisLifting schemeWavelet transformWavelet packet decompositionArtificial intelligenceDiscrete wavelet transformWavelet

摘要: Wavelet transforms applied to multiresolution analyses of images produce outputs similar in theory those matched filters. In order maximize the output at location and scale a signal interest, it is necessary for wavelet used analysis `match' interest. Current techniques match one several predefined wavelets library, which requires be designed advance. Here, we present an alternative by developing technique deriving directly from desired spectrum such way that mean squared error between their spectra minimum. Furthermore, its associated scaling function generates orthonormal analysis. The includes algorithm finding wavelet, algorithms optimal magnitude phase given input signal. Several examples are shown demonstrate performance both known arbitrary signals.© (1995) COPYRIGHT SPIE--The International Society Optical Engineering. Downloading abstract permitted personal use only.

参考文章(0)