m23d : Manchester's Library for Generating 3D models from 2D data

Various functions for generating linear models of 3D point positions and their variation, given multiple 2D views.

Initial classes assume scaled orthographic projection models, and that all points are visible in all frames. Note that the Oxford libraries contain more well developed tools for dealing with tracking of rigid objects under projective projections.

Key classes:

• m23d_ortho_rigid_builder: Reconstruct 3D shape of rigid object
• m23d_ortho_flexible_builder: Reconstruct 3D shape basis of deformable object

Example usage:

*
* vnl_matrix<double> W;  // Measurements ( rows are 2D shapes, (x1,y1,x2,y2...))
* // Fill in measurement matrix W
* ...
*
* m23d_ortho_flexible_builder builder;
* builder.reconstruct(W,n_modes);
*
* // Matrix containing mean shape (3 x n_points, each column is one point)
* vnl_matrix<double> M = builder.mean_shape();
* // Matrix containing first shape mode (3 x n_points, each column is one point)
* vnl_matrix<double> S0= builder.shape(0);
* // etc
*
*