基于稀疏编码的提取人脸整体特征算法

为了提高特征的分类效果 ,提出了一种新的基于稀疏编码的人脸特征提取方法。稀疏编码是去除图像冗余的一种有效方法 ,适于描述具有 non- Gaussian分布的图像集合。和基于主成分分析 (PCA )的传统方法相比 ,利用稀疏编码提取的特征具有更好的分类特性。根据稀疏编码的聚类特性 ,利用模糊 C均值聚类对稀疏编码基函数进行初始化以进一步提高特征的可分性。人脸识别的试验结果表明 ,该方法明显优于传统的“特征脸”方法 ,是一种有效的图像整体特征提取方法

Sparse coding-based global face feature extraction algorithm

A new sparse coding based global face features extraction method is proposed to improve the feature classification ability. Sparse coding is an effective method for removing image redundancy that is suitable for describing images with non Guassian distributions. Feature extracted by sparse coding has better discrimination than the traditional PCA based method. A Fuzzy C mean clustering method based on the clustering property of sparse coding is used to initialize the basis function of sparse coding to further improve the sparse feature recognition performance. Experimental results on facial image recognition proved that our method is much better than the traditional eigenface method. The method can also be used as an effective global image features extraction method for other image recognition tasks.