Local manifold distance based on neighborhood graph reordering
作者：
Highlights：
• Pair-wise dissimilarity between signals is addressed as a manifold-to-manifold distance.
• Diversity between two manifolds is estimated using a measure of reordering efficiency of the corresponding neighborhood graphs.
• Multi-dimensional ordering of nodes proves to be even more discriminative.
• We demonstrate the efficiency of the proposed measure in a variety of 1D and 2D signals.
• State-of-the-art performance is achieved in face recognition under challenging scenarios.
摘要
•Pair-wise dissimilarity between signals is addressed as a manifold-to-manifold distance.•Diversity between two manifolds is estimated using a measure of reordering efficiency of the corresponding neighborhood graphs.•Multi-dimensional ordering of nodes proves to be even more discriminative.•We demonstrate the efficiency of the proposed measure in a variety of 1D and 2D signals.•State-of-the-art performance is achieved in face recognition under challenging scenarios.
论文关键词：Face recognition,Graph reordering,Local manifold,Manifold–manifold distance,Minimal spanning tree,Range-dependent graphs,Spectral reordering
论文评审过程：Received 11 May 2015, Revised 4 November 2015, Accepted 9 December 2015, Available online 19 December 2015, Version of Record 8 February 2016.
论文官网地址：https://doi.org/10.1016/j.patcog.2015.12.006
