Sparse tree structured representation for re-identification

作者：

Highlights：

• We model pairwise image similarity for person re-identification.

• We learn the sparse tree structured representation in an unsupervised fashion from various color spaces.

• The feature vectors are extracted from nested patch trees.

• Deeper the tree and more color types will increase the performance.

摘要

Highlights•We model pairwise image similarity for person re-identification.•We learn the sparse tree structured representation in an unsupervised fashion from various color spaces.•The feature vectors are extracted from nested patch trees.•Deeper the tree and more color types will increase the performance.

论文关键词：Appearance model,Dictionary learning,Image similarity measure,Re-identification,Sparse coding,Tree similarity,Unsupervised feature learning

论文评审过程：Received 1 May 2015, Revised 24 April 2016, Accepted 11 May 2016, Available online 24 May 2016, Version of Record 15 June 2016.

论文官网地址：https://doi.org/10.1016/j.patcog.2016.05.003