【摘要】 利用自适应小波变换(AWT)能融合不同母小波的优点,采用盖伯母小波和墨西哥帽母小波构成自适应小波。以光学人脸识别中降噪问题为应用背景,使用神经网络法对小波参数和组合系数进行优化,将生成的多母小波自适应小波滤波器用作人脸特征提取器。对噪声图像做特征提取,进行相关识别,采用3个指标定量分析识别结果。同盖伯小波和墨西哥帽小波识别结果的比较表明,多母小波自适应小波具有不同母小波的优点,并有良好的降噪性能。更多还原

【Abstract】 Considering the advantage for combining different mother wavelets of adaptive wavelet transform(AWT),six wavelets selected from Gabor wavelets and Mexican-hat wavelet are chosen to construct a multi-mother adaptive wavelet.For the purpose of de-noising in optical face pattern recognition,using artificial neural networks,the parameters of the wavelets are optimized.Several criterions are used for evaluating the correlation recognition performance.After comparing the criterions of Gabor wavelets,Mexican-hat wavelet and the new adaptive wavelet,the evaluation shows that the new adaptive wavelet combines advantages of these two mother wavelets and has good de-noising ability.Moreover,the optical filter of this adaptive wavelet can be applied in optical pattern recognition.更多还原