WSEAS Transactions on Mathematics
Print ISSN: 1109-2769, E-ISSN: 2224-2880
Volume 14, 2015
Feature Combination Methods Based on Graphs
Authors: ,
Abstract: In pattern recognition, graph-based feature combination techniques attract many researchers to study this field. In this paper, we construct a unified framework based on graphs (GF), and derive that FDA, PCA, LPP, DLPP, MFA and MMC are special cases of GF, and then three new algorithms are proposed for GF, which are regularized GF (RGF), GF based on null space (NGF) and GF based on singular value decomposition (GF/SVD). Experiments are made on AVIRIS remote sensing image to illustrate the efficient and effective of our algorithms. The results show that the effects of proposed algorithms are very significantly.