We present an improved Locality Preserving Projections (LPP) method, namedGloablity-Locality Preserving Projections (GLPP), to preserve both the globaland local geometric structures of data. In our approach, an additionalconstraint of the geometry of classes is imposed to the objective function ofconventional LPP for respecting some more global manifold structures. Moreover,we formulate a two-dimensional extension of GLPP (2D-GLPP) as an example toshow how to extend GLPP with some other statistical techniques. We apply ourworks to face recognition on four popular face databases, namely ORL, Yale,FERET and LFW-A databases, and extensive experimental results demonstrate thatthe considered global manifold information can significantly improve theperformance of LPP and the proposed face recognition methods outperform thestate-of-the-arts.
展开▼
