Open Access

Table 3

Advantages and disadvantages of common visual 3D reconstruction algorithms.

Common visual 3D reconstruction algorithms References Advantages Disadvantages
Incremental SFM Schönberger and Frahm [82], Yin and Yu [84] The performance is robust and the reconstruction precision is high Affected by the initial image on the selection and camera add order, the cumulative error is large and the efficiency is not high in the reconstruction of large scenes
Global SFM Moulon et al. [80] Not affected by the initial image pair and the order of camera addition, the cumulative error is small, and the reconstruction efficiency is high The robustness is not good, and the completeness of scene reconstruction is insufficient
Hybrid SFM Cui et al. [83] The cumulative error is small and the robustness is good The efficiency is not high
Voxel based MVS Sinha et al. [87] The generated point cloud is regular and mesh is easy to extract Reconstruction accuracy is related to voxel particle size, and it is difficult to deal with large scenes
Feature point growing based MVS Lin et al. [88] The point cloud has high precision and uniform distribution Areas with weak textures are prone to holes and require reading all images at once
Depth-map merging based MVS Seitz et al. [86], Lindenberger et al. [89], Zhou et al. [90] It can be used in parallel computation for 3D reconstruction of large scenes, and the number of point clouds obtained is large Too dependent on neighborhood image group selection

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.