基于改进核范数的2DPCA人脸识别算法研究

传统的二维主成分分析法广泛应用于图像特征提取,为了使此算法更加有效,提出了一种结构化二维算法,即核范数2DPCA算法(N-2-DPCA).该算法基于核范数重构误差准则,将核范数最优化问题转化为基于F范数的最优化问题,然后通过采用迭代方法寻找到最佳投影矩阵,最后运用最小欧氏距离规则识别出待识别人脸的身份.在此基础之上,将N-2-DPCA扩展到基于双边投影的算法(N-B2-DPCA),采用曲线搜索算法寻找到双边投影矩阵,继而进行识别.最后将提出的算法在FERET和Yale B人脸数据库中进行人脸识别评估,实验结果表明所提出的算法与L1-2DPCA相比,重建误差降低了2.19％,识别率提高了2.03％,性能更好.

Human Face Recognition Based on Improved Nuclear Norm 2DPCA Algorithm

The traditional two dimensional component analysis is widely used for image feature extraction. In order to make the algorithm more effective,this paper presents a structural 2-D method, which is based on nuclear norm reconstruction error criterion, namely, nuclear norm-based 2-DPCA(N-2-DPCA).The algorithm can provide a structured 2-D characterization for the reconstruction error image, by means of converting the nuclear norm-based optimization problem into a series of F-norm-based optimization problems, the reconstruction error criterion is minimized. In addition, the N-2-DPCA is extended to a algorithm based on bilateral projection(N-B2-DPCA), as an image can be represented with fewer coefficients, so the performance is better than N-2-DPCA. Finally, the proposed algorithm is evaluated in FERET and Yale B face database, the experimental results show that the proposed method has lower reconstruction error, higher recognition rate and it performs better.