Section: New Results

Multiple camera reconstruction and motion segmentation

Robust perspective factorization.

One of our major research topics is the problem of recovering structure and motion from a large number of intrinsically calibrated perspective cameras. We describe a method that combines (1) weak-perspective reconstruction in the presence of noisy and missing data and (2) an algorithm that updates weak-perspective reconstruction to perspective reconstruction by incrementally estimating the projective depths. The method also solves for the reversal ambiguity associated with affine factorization techniques. The method has been successfully applied to the problem of calibrating the external parameters (position and orientation) of several multiple-camera setups. Results obtained with synthetic and experimental data compare favourably with results obtained with non-linear minimization such as bundle adjustment.

3-D motion segmentation.

We investigated the problem of motion segmentation within the context of spectral clustering. The input data consisted in sparse scene flow. The latter consists in a set of trajectories associated with the motion of 3-D points. We addressed the problem of segmenting these trajectories based on the rigidity constraints. Since it appeared difficult to represent these trajectories in some parameter space, we decided to use spectral methods. We proposed a similarity measure between two trajectories and we showed how it can be plugged into a spectral method. In practice we implemented and tested two algorithms: the first one seeks degrees of correlation within the data, the second one uses the Multiway Normalized Cut method.