基于连续均值量化变换的人脸检测算法

针对人脸检测中对旋转、光照、遮挡等检测的准确性与稳健性问题,提出将连续均值变换用于人脸检测的方法.首先利用连续均值变换提取候选区域人脸的特征,用提取的特征训练SNoW分类器,通过分类器对人脸与非人脸样本进行分类,达到准确确定人脸位置的目的.实验结果证明,与人工神经网络、支持向量机和朴素贝叶斯相比,该方法在复杂背景和光照和遮挡以及多人脸等情况下仍然具有很好的准确性和稳健性.

Face Detection Based Successive Mean Quantization Transform

To improve the problems of accuracy and robustness under the circumstances of rotate, covering and illumination in face detection, the paper represents a method which uses the Successive Mean Quantization Transform (SMQT). Firstly, extract local face feature of candidate regional using SMQT. Then learn the face feature to train the SNoW classifier. Finally, achieve the purpose of positioning faces accurately through classify face and face samples with the SNoW. Empirical results show the method outperforms than neural networks, support vector machines and Bayesian methods in accuracy and robustness even though with complex background, more than one faces, covering and illumination effects situations.