Neighborhood preserving embedding on Grassmann manifold for image-set analysis

作者：

Highlights：

• A novel unsupervised DR approach on Grassmann manifold for image set classification and clustering tasks is proposed.

• We propose a Grassmannian unified learning framework to fulfil the gap between similarity learning and projection learning.

• An efficient algorithm is proposed to solve the optimization problems which only involve the basic eigenvalue problem.

摘要

•A novel unsupervised DR approach on Grassmann manifold for image set classification and clustering tasks is proposed.•We propose a Grassmannian unified learning framework to fulfil the gap between similarity learning and projection learning.•An efficient algorithm is proposed to solve the optimization problems which only involve the basic eigenvalue problem.

论文关键词：Neighborhood preserving embedding,Dimensionality reduction,Grassmann manifold,Twin learning

论文评审过程：Received 20 February 2021, Revised 16 June 2021, Accepted 17 September 2021, Available online 20 September 2021, Version of Record 26 September 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108335