基于小波分析与神经网络的人脸检测方法

本文设计了一种结合小波分析与神经网络优点的人脸检测方法.该方法主要包括两个阶段.在训练阶段首先借鉴图形学中的背景生成方法对AR人脸数据库中的人脸样本进行背景叠加;然后对经过预处理的训练样本进行小波分解,并将得到的小波系数输入精心设计的神经网络进行训练.在人脸检测阶段,通过将缩放后的图像的各个区域输入神经网络,由神经网络判断输入区域中是否包含人脸:在得到检测结果后,本文通过基于规则的方法以及基于SUSAN的方法进行人脸区域验证取舍与合并;最后通过实验结果的分析比较可以发现本文的方法不仅可以实现人脸区域的快速检测,而且不受光照变化以及噪声的影响.

A neural network and wavelet based face detection method

A novel face detection method, which combined the merit of neural network and wavelet, is presented in this paper. Two stages are involved in this method. In the training phase, some background masks are generated and added to the background of the AR face database examples. Then the pretreated face images are decomposed by the wavelet transformation and the corresponding wavelet coefficients are set as the input samples for a well-designed neural network. In the face detection phase, the neural network is used to judge whether the input region of the resized images contained a face. The detection results are validated and combined by the rule-based method and SUSAN-based method. The experimental results illustrate the efficiency and robustness of the proposed approach.