融合边缘直方图和局部二值模式的稳健人脸识别

Combining edge histogram and localbinary pattern for robust face recognition

提出了一种融合边缘直方图和局部二值模式的人脸识别算法具体步骤包括(1)提取人脸的边缘特征,分块计算人脸的边缘直方图特征,连接人脸边缘的各分块直方图特征向量得到边缘直方图特征向量;(2)计算人脸分块的局部二值模式直方图特征得到人脸的局部二值模式直方图特征向量;(3)融合边缘直方图特征向量和局部二值模式直方图特征向量得到人脸最终的匹配特征向量。实验结果表明,本文提出的基于融合特征的人脸识别效果优于提取单一特征的人脸识别效果。在ORL和FERET人脸库中的人脸识别实验中,验证了该方法不仅提高了识别率,而且对噪声具有较好的稳健性。

This paper presents a hybridized method that combines the edge histogram and local binary pattern to recognize the human faces. Three main steps are concluded in our algorithm： （1） Evenly dividing image into several non-overlapping square sub-blocks and extract the edge of each sub-block, calculate the histogram of each edged image, and then link all the histogram to obtain the vector of facial edge histogram feature vector; （2） Evenly dividing image into several non-overlapping square sub-blocks and calculate the histogram of each sub-blocks, and then link all the histogram to obtain the vector of LBP histogram feature vector; （3） fuse face edge histogram feature vector and local binary pattern histogram feature vector to obtain the final face feature vector. Extensive experiments on ORL and FERET face database show that our proposed algorithm can not only increase the recognition rate but also can enhance the robustness against noise.