Block-wise 2D kernel PCA/LDA for face recognition-笔记

In the present work, we propose a framework for kernel-based 2D feature extraction algorithms tailored to face recognition .     
extending 2D-PCA/LDA in the following two aspects: 
(1)kernel technique is incorporated to capture the higher order statistical dependencies among the rows of input images. Thus, face recognition results are considerably improved 
(2)A block-wise model for face recognition is introduced and proved to be reliable through our experiments.

2D-KPCA所带来的问题：

（1）Computational complexity: unlike direct extension of 2D-PCA/LDA to kernel-induced feature space which is computationally intractable, proposed B2D-KPCA/GDA perform subspace projection inside each block manifold, significantly alleviating computational cost 
（2）Locality: as widely known, distribution of face images is both multimodal and highly nonlinear [7,17]. Direct extension of 2D-PCA/LDA to kernel-induced feature space, overcomes the shortcomings of 2D-PCA/LDA in extracting global nonlinear structure of input data, but fails to learn local characteristics of input data.