3D reconstruction of multiple views based on trifocal tensor.
- Author:
Chunxiao CHEN
1
;
Juan ZHANG
Author Information
1. Department of Biomedical Engineering, Nanjing University of Aeronautics & Astronautics, Nanjing 210016, China.
- Publication Type:Journal Article
- MeSH:
Algorithms;
Calibration;
Face;
anatomy & histology;
Humans;
Image Enhancement;
methods;
Imaging, Three-Dimensional;
Pattern Recognition, Automated;
methods;
Photogrammetry;
methods
- From:
Journal of Biomedical Engineering
2012;29(4):769-774
- CountryChina
- Language:Chinese
-
Abstract:
Reconstruction of 3D structure of an object from 2D views plays an important role in plastic surgery and orthopedics. This method doesn't need camera to do specific movements, such as translation or rotation independently. It only needs a hand-hold camera arbitrarily to take a few pictures, and apply the geometry relationship among the three views to obtain the projective reconstruction of the object. Then, it needs to introduce cheirality constraint in stratified reconstruction to determine the search area of the infinity plane, and finally achieve the camera intrinsic parameters calibration, and complete the metric reconstruction. This model has also been reconstructed with mouse and keyboard response coordinates to observe the model from different angles. Experiments with both pictures of object and face pictures show that the proposed method is very robust and accurate.
