Monocular Dense 3D Reconstruction Algorithm Based on Inverse Depth Filter

In this paper, an inverse depth filter based real-time monocular 3D dense reconstruction algorithm is proposed. We construct the Gaussian-uniformly mixed probability distribution model by fusing the right and wrong observations in inverse depth filter, which effectively solves the "long tail" problem in conventional depth filter. Then a fast smoothing method is adopted to reduce the effects of noises and false observations, which greatly improves the accuracy and robustness of our method. Experiments are designed to evaluate the accuracy and effectiveness of the proposed algorithm, and the results show that our algorithm has good 3D dense reconstruction performance and high computational accuracy.