Laplace平滑变换在人脸识别应用中的问题及相关算法

A Question in Laplacian Smoothing Transform for Face Recognition and Related Algorithms

本文通过根据定义直接计算的方法,得到"Laplace平滑变换及其在人脸识别中的应用"一文的Laplace矩阵,指出该文献的个别结果与其定义并不一致,其理论基础值得商榷.然而,该文对ORL和Yale人脸图像库的高识别率仍然的没有问题.同时也指出类似的现象也出现在由二阶差分导出的矩阵及相关的变换中.仿真实验表明,由此导出的人脸识别算法也具有非常高的识别率,并且具有更高的速度.

In this paper, the matrix of Laplacian transform in the paper ＂Laplacian smoothing transform for face recognition＂ is computed by its definition directly and is poined out that a result is not consistent with its definition, hence its theoretical foundation is debatable. However, the rate of recognition for face image database ORL and Yale is still true. It is also pointed out that similar phenomenon also appears in the matrix and related transform which derived by second order difference. Simulation results show that the derived a face recognition algorithm also has a very high recognition rate and a higher speed.