3D Registration Method Based On MVFNet And Super4PCS For Nonhomologous Face Pointclouds
Chenbo Dong, Ya Zhou, Jiacheng Han · 2022
To find an economical and accurate reconstruction method for face efficacy evaluation in the medical field, especially for clinics environment can be a challenge because the comparison before and after operation is always involved. In this paper, a MVFNet-based 3D face reconstruction for efficacy evaluation is proposed. However, it is not practical yet since the MVFNet reconstructed pointcloud is far different from accurate pointcloud reconstructed by professional equipment. Therefore, the Super4PCS algorithm is used, which can be used to register nonhomologous pointclouds with high difference. The registration experiments are designed and conducted to evaluate the performance of Super4PCS, and we analyze the accuracy of MVFNet reconstructed pointcloud. The results show that MVFNet reconstructed pointcloud is accurate enough and Super4PCS can register two pointclouds well.