A novel eye center localization method for multiview faces

作者：

Highlights：

• We propose a novel eye center localization method that effectively localizes the eye centers under various situations on multiview faces with large yaw-rotation angles.

• A multiview face detector is developed to flexibly detect and precisely capture the face region on large yaw angles.

• We propose a new iris-ripple filter to increase the accuracy and robustness of gradient localization.

• A new depth corresponding-points conversion method is proposed to automatically estimate rotation variables between two faces and the conversion relationship between eye centers.

摘要

•We propose a novel eye center localization method that effectively localizes the eye centers under various situations on multiview faces with large yaw-rotation angles.•A multiview face detector is developed to flexibly detect and precisely capture the face region on large yaw angles.•We propose a new iris-ripple filter to increase the accuracy and robustness of gradient localization.•A new depth corresponding-points conversion method is proposed to automatically estimate rotation variables between two faces and the conversion relationship between eye centers.

论文关键词：Eye center localization,Multiview face,Face generation,Geometric transformation

论文评审过程：Received 25 March 2020, Revised 25 March 2021, Accepted 27 May 2021, Available online 8 June 2021, Version of Record 22 June 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108078