Robust face recognition with structural binary gradient patterns

作者：

Highlights：

• A compact and efficient face image descriptor, termed Binary Gradient Patterns (BGP), is proposed.

• Structural BGP extracts structural gradient patterns as binary strings in local regions for effective representation.

• BGP and its enhanced BGPM are shown robust to various variations, esp. illumination.

• Experimental results demonstrate significant performance improvements over the existing methods, including with deep networks.

摘要

•A compact and efficient face image descriptor, termed Binary Gradient Patterns (BGP), is proposed.•Structural BGP extracts structural gradient patterns as binary strings in local regions for effective representation.•BGP and its enhanced BGPM are shown robust to various variations, esp. illumination.•Experimental results demonstrate significant performance improvements over the existing methods, including with deep networks.

论文关键词：Face representation,Invariant face recognition,Gradient domain,Binary features and patterns

论文评审过程：Received 27 October 2016, Revised 13 February 2017, Accepted 5 March 2017, Available online 6 March 2017, Version of Record 15 March 2017.

论文官网地址：https://doi.org/10.1016/j.patcog.2017.03.010