基于改进LBP的单样本人脸识别算法

A SINGLE SAMPLE FACE RECOGNITION ALGORITHM BASED ON IMPROVED LBP

针对人脸识别难题,提出一种基于改进LBP(Local Binary Patterns)算子的单样本人脸识别算法。采用Bernser算法与LBP算子结合的BLBP算子,最后利用Chi平方统计方法计算直方图的相似度。在识别时,采用的是核实式的一对一匹配,根据训练的阈值,判断两张比对的人脸图像是否为同一个人。所提出的算法在实际人脸图像和FERRET人脸数据库下的人脸识别中,与原LBP算法相比识别率有所提高。实验结果表明,改进后的LBP算子有较好的去噪能力,在实际的人脸识别中能获得更好的识别率。

For the single sample face recognition problem, this paper proposes a face recognition algorithm based on improved LBP operator. We used BLBP operator combined with Bernser algorithm and LBP operator; we calculated the similarity of histogram by using Chi square statistical method. During the recognition process, we used the verification type of one to one match. According to the training threshold, our algorithm could identify determine whether the two face images were the same person. Compared with the original LBP algorithm, the proposed algorithm could improve the recognition rate of face recognition in real face image and FERRET face database. The experimental results show that the improved LBP operator has a better denoising ability and can achieve better recognition in the actual face recognition.