2006. 11. 3 (Fri) Young Ki Baik, Computer Vision Lab. . 3D Surface Reconstruction from 2D Images (Survey). 3D Surface Reconstruction from 2D Images. References A Comparison and Evaluation of Multi-View Stereo Reconstruction Algorithms Steven M. Seitz, Richard Szeliski et. al. (CVPR 2006)
→3D medical imaging [Greenleaf 70]
→practical alternatives to surface based geometrical representation for many applications in computer graphics and scientific visualization
Camera system for obtaining images
3D reconstruction system to make 3D object
- Back-project each silhouette along the ray
- Obtain 3D volumetric data from intersecting back-projected volume
- Calibrated cameras and object
- Set initial 3D volumetric region including object
→ commonly used as the effective initial boundary
→ specific color is used as the background
- Place the object to the fixed location.
- Set the camera on the fixed location.
- Set up voxel region covering object.
- Iterate this algorithm about all voxel in the region.
- Select a voxel and project onto the each image.
- Judge opaqueness by thresholding variance of colors.
→reconstruct object in small area
→high computational cost
Left Camera3D Surface Reconstruction from 2D Images
(fewer candidate positions)
- Carlos Hernandez Esteban and Francis Schmitt (CVIU 2004)
# SFS +multi +stereo correlation voting +Gradient vector flow +Snake
- Yasutaka Furukawa, Jean Ponce (CVR-TR-2006)
# SFS + wide baseline matching + propagation + Energy minimization
- Michael Goesele et. al. (CVPR2006)
# SFS +Stereo matching + volumetric method (range data + Level Set)
- Lingyoun Liu et. Al. (CVPR2006)
# SFS +Stereo matching + volumetric method (range data)
(SFS + Stereo + Level set + …)