3D Reconstruction of Epithelial Surfaces using Optical Flow
We provide Matlab codes and supplementary materials of the paper:
Tan-Binh Phan, Dinh-Hoan Trinh, Didier Wolf, Christian Daul, "Optical Flow-based Structure-from-Motion for the Reconstruction of Epithelial Surfaces," Pattern Recognition (accepted April, 2020).
- Medical image datasets can be downloaded here
- Matlab codes implementing the algorithms of the paper can be found in the zipped file named "Matlab_code_algorithm.7z".
- Matlab_code_algorithm.7z consists of the Matlab code of the optical flow as well as Matlab code for the determination of HP-groups (Groups of Homologous Points) between images in a given video-sequences.
- The codes tested with 64 bit MATLAB 8.X or 9.X on Windows OS.
- Run Demo_compute_OF.m to see the effectiveness of the OF algorithm under strong illumination changes.
- Demo_HP_groups_using_DOF.m uses function Generate_HP.m for demonstration the algorithm of homologous point group determination.
- Generate_HP.m is a function with the input is an images sequence, the output is the HP-groups as illustrated in the papers. From HP-groups, one can generate the 3D point cloud using the existing SfM code (it is recommended to use COLMAP).