Abstract

This paper proposes a quasi-dense reconstruction from un- calibrated sequence. The main innovation is that all geometry is com- puted based on re-sampled quasi-dense correspondences rather than the standard sparse points of interest. It not only produces more accurate and robust reconstruction due to highly redundant and well spread input data, but also fflls the gap of insuffciency of sparse reconstruction for visualization application. The computational engine is the quasi-dense 2-view and the quasi-dense 3-view algorithms developed in this paper. Experiments on real sequences demonstrate the superior performance of quasi-dense w.r.t. sparse reconstruction both in accuracy and robustness.