Face recognition with Patch-based Local Walsh Transform
作者：
Highlights：
•
• We present novel unsupervised dense local image repr. methods called LWT and CLWT.
• The PLWT and PCLWT methods combine the adv. of both sparse and dense local repr.
• PLWT and PCLWT are robust to illum. and expression changes, occlusion and low res.
• The methods are applied to face identification and face verification problems.
• The state-of-the-art performance is achieved on the FERET and SCface databases.
• The second best unsupervised category result is achieved on the LFW database.
摘要
•We present novel unsupervised dense local image repr. methods called LWT and CLWT.•The PLWT and PCLWT methods combine the adv. of both sparse and dense local repr.•PLWT and PCLWT are robust to illum. and expression changes, occlusion and low res.•The methods are applied to face identification and face verification problems.•The state-of-the-art performance is achieved on the FERET and SCface databases.•The second best unsupervised category result is achieved on the LFW database.
论文关键词：Face recognition,Local representations,Walsh Transform
论文评审过程：Received 9 June 2017, Revised 18 November 2017, Accepted 19 November 2017, Available online 2 December 2017, Version of Record 5 December 2017.
论文官网地址：https://doi.org/10.1016/j.image.2017.11.003
