Monocular-Cues Based 3-D Reconstruction: A Comparative Review

Background

The 3-D reconstruction problem can be formulated in multiple ways (Zhang, R., Tsai, P.-S., Cryer, J., & Shah, M., 1999). A common formulation requires the calculation of depth value at each coordinate. The depths can be expressed as the distance of the surface point to the camera or from any other standard surface like the x-y plane. Other ways to express 3-D structure is by using surface normals, surface gradients or surface slant and tilt angles. In addition to these quantitative approaches, 3-D reconstruction algorithms can simply generate geometric/qualitative labels that allow the organization of the image into a 3-D hierarchy.

In this section, we provide a high level review of previous work on 3-D reconstruction both using binocular or multiple views and monocular views.