Facial expression recognition with local prominent directional pattern
作者：
Highlights：
• A novel feature extraction method using a local histogram of directional variations is proposed.
• The feature encoding process is discriminative, and stable against local distortions.
• An adaptive thresholding process for avoiding feature extraction at featureless flat/smooth regions is introduced.
• A feature selection mechanism that retains expression-related micro-level features is proposed.
• Extensive experiments on various datasets including posed, spontaneous, noisy and position-varied facial expression images are conducted.
摘要
•A novel feature extraction method using a local histogram of directional variations is proposed.•The feature encoding process is discriminative, and stable against local distortions.•An adaptive thresholding process for avoiding feature extraction at featureless flat/smooth regions is introduced.•A feature selection mechanism that retains expression-related micro-level features is proposed.•Extensive experiments on various datasets including posed, spontaneous, noisy and position-varied facial expression images are conducted.
论文关键词：LDPD,Histogram of directional variations,Prominent directions,Shape pattern,Facial expression recognition
论文评审过程：Received 26 June 2018, Revised 5 January 2019, Accepted 5 January 2019, Available online 11 January 2019, Version of Record 21 January 2019.
论文官网地址：https://doi.org/10.1016/j.image.2019.01.002
