Deep learning-based single image face depth data enhancement

作者：

Highlights：

• Pure depth image enhancement using deep learning is effective for facial biometrics.

• Synthesis of realistic low detail face depth enhancer training data is viable.

• Comparisons with more general enhancers favor the face-specific model.

• Depth is not overly falsified for non-face input during enhancement.

• Face depth enhancement can be used to aid real-time presentation attack detection.

摘要

•Pure depth image enhancement using deep learning is effective for facial biometrics.•Synthesis of realistic low detail face depth enhancer training data is viable.•Comparisons with more general enhancers favor the face-specific model.•Depth is not overly falsified for non-face input during enhancement.•Face depth enhancement can be used to aid real-time presentation attack detection.

论文关键词：3D face depth,Deep learning,Image enhancement,Face depth synthesis,Face recognition,Presentation attack detection

论文评审过程：Received 4 March 2020, Revised 2 July 2021, Accepted 15 July 2021, Available online 19 July 2021, Version of Record 23 July 2021.

论文官网地址：https://doi.org/10.1016/j.cviu.2021.103247