Accurate, Dense, and Robust Multiview Stereopsis

Abstract

This paper proposes a novel algorithm for multiview stereopsis that outputs a dense set of small rectangular patches covering the surfaces visible in the images. Stereopsis is implemented as a match, expand, and filter procedure, starting from a sparse set of matched keypoints, and repeatedly expanding these before using visibility constraints to filter away false matches. The keys to the performance of the proposed algorithm are effective techniques for enforcing local photometric consistency and global visibility constraints. Simple but effective methods are also proposed to turn the resulting patch model into a mesh which can be further refined by an algorithm that enforces both photometric consistency and regularization constraints. The proposed approach automatically detects and discards outliers and obstacles and does not require any initialization in the form of a visual hull, a bounding box, or valid depth ranges. We have tested our algorithm on various data sets including objects with fine surface details, deep concavities, and thin structures, outdoor scenes observed from a restricted set of viewpoints, and "crowded" scenes where moving obstacles appear in front of a static structure of interest. A quantitative evaluation on the Middlebury benchmark shows that the proposed method outperforms all others submitted so far for four out of the six data sets.

SF3M software: 3-D photo-reconstruction for non-expert users and its application to a gully network

“…Photo-reconstruction: in SF3M v1.0 the photoreconstruction stage is carried out through a system call to VisualSFM using command line syntax to drive automated processing. The main VisualSFM commands used in our utility are extracting SIFT features, image matching, exporting match matrix, bundle adjustment and dense reconstruction (multi-view stereo PMVS2 software, Furukawa and Ponce, 2010).…”

Image Based 3D Reconstruction of Texture-less Objects for VR Contents

“…Dense reconstruction of the point cloud is performed using either Patch-based Multi-view Stereo (PMVS) [21] or Clustering Views from Multi-view Stereo which are able to reconstruct global point clouds.…”

An Enhanced Structure-from-Motion Paradigm Based on the Absolute Dual Quadric and Images of Circular Points

“…As a counting argument, it can be seen that the configuration of the image of one circular point and the (scaled) matched images of P natural points introduces 10 unknowns while providing 2P +6 equations. Hence, P = 2 natural points is a minimal theoretical data set 1 . An overdetermined set of equations can be easily solved in a linear least squares manner.…”

“…3(d), we also considered 6 views but let vary the number of views in which the CP is visible. The key result is that, compared to "classical SfM" PF[0]+SC[0], the accuracy was improved by using either PF[0]+SC [1] or PF [1]+SC[0] and best by using PF [1]+SC [1]. This means that both proposed algorithms for projective factorisation and self-calibration independently contributed to better results even if the benefits provided by additional constraints from CP images were generally more significant in self-calibration than in projective factorization, particularly when the number of views decreased and the noise increased.…”

“…(c) No consistent Euclidean reconstruction was obtained using either Bouguet's Toolbox or P OLLEFEYS' method [5,6,10]. (c)+(d) Only the proposed method, starting from the same projective reconstruction than other methods, by adding circular points in self-calibration (see Table 4), was able to provide PMVS [1] -even if the planar shape was very small-accurate Euclidean cameras. This yielded the dense reconstruction seen here from three different viewpoints.…”