3D Facial Similarity Measure Based on Deformation Field
Jiajun Ma, Guoyuan Liang, Yu Liang, Xinyu Wu · 2021
3D facial similarity measure is a critical research direction in fields such as 3D face recognition, craniofacial reconstruction, and heritability estimation, etc. and represents a challenging and meaningful task. In this paper, we propose a novel method based on deformation field to establish the similarity relationship between 3D faces. The main idea is divided into two steps as follows: (1) The landmarks are semi-automatically marked on all faces to build correspondences between landmarks (2) According to the distance field and the dense corresponding relationship of faces, a deformation field is generated between 3D faces that deform the source model to the target model. The path of deformation is computed to measure the similarity between two 3D faces. Comprehensive experiments conducted on the BU3DFE database illustrate that the similarity measure method is consistent with the subjective evaluation and provides high discriminability for different individuals.