【作者】 蒋建国； 卢晓红； 齐美彬； 詹曙；

【Author】 Jiangjian-guo Lu xiao-hong Qi mei-bin Zhan shu(School of Computer and Information,Hefei University of Technology,Hefei 230009)

【机构】 合肥工业大学计算机与信息学院；

【摘要】 提出一种基于提升小波的人脸识别算法。该算法利用 D9/7提升小波对归一化人脸图像进行变换,分别提取低频平滑,水平细节和垂直细节三个子图的小波特征,将提取的小波特征利用新的方法聚类为观察向量,使用一维离散隐马尔可夫模型进行训练,并把优化后的模型参数用于人脸识别。该算法充分地利用人脸小波子图的判别信息,对于光照、姿态和表情均具有良好的鲁棒性。在 ORL 人脸库中的实验结果表明本文算法运算复杂度低,识别效率高。更多还原

【Abstract】 A new approach based on lifting wavelet transform for face recognition is presented.The Daubechies 9/7 Lfting Wavelet Transform(D9/7 LWT)is employed to preprocess the normalized faces and three groups of wavelet features: approximation sub-image,horizontal detail sub-image and vertical detail sub-image,are obtained respectively.New way of extracting and using observation sequence from the three wavelet sub-images for recognition is introduced.The observation sequence is then modeled by using 1D Discrete Hidden Markov Model(1D-DHMM).The experimental results on the Olivetti Research Laboratory(ORL)face database show that the proposed method has a high recognition rate with relatively low complexity,and it is also fairly robust to the variety of different illumination condition,face pose and expression.更多还原