The opencv_sfm module contains algorithms to perform 3d reconstruction from 2d images.
The core of the module is based on a light version of Libmv originally developed by Sameer Agarwal and Keir Mierle.
Whats is libmv?
libmv, also known as the Library for Multiview Reconstruction (or LMV), is the computer vision backend for Blender's motion tracking abilities. Unlike other vision libraries with general ambitions, libmv is focused on algorithms for match moving, specifically targeting Blender as the primary customer. Dense reconstruction, reconstruction from unorganized photo collections, image recognition, and other tasks are not a focus of libmv.
libmv is officially under the Blender umbrella, and so is developed on developer.blender.org. The source repository can get checked out independently from Blender.
This module has been originally developed as a project for Google Summer of Code 2012-2015.