Frame grouping measure for factorization-based projective reconstruction

The factorization-based method generally suffers less from drift and error accumulation than the merging. However, the factorization method assumes that all correspondences must remain in all frames. In order to overcome the limitation, we present a new factorization-based projective reconstruction from un-calibrated image sequences. The proposed method breaks the full sequence into sub-sequences based on a quantitative measure considering the number of matching points between frames, the homography error, and the distribution of matching points in the image. All of projective reconstructions in sub-sequences are registered into the same coordinate frame for a complete description of the scene. Experimental results showed our algorithm could recover more precise 3D structure than the merging method.