Technical requirements

These technologies and installations are required to build and run the code in this chapter:

• OpenCV 4 (compiled with the sfm contrib module)
• Eigen v3.3+ (required by the sfm module)
• Ceres solver v2+ (required by the sfm module)
• CMake 3.12+
• Boost v1.66+
• OpenMVS
• CGAL v4.12+ (required by OpenMVS)

The build instructions for the components listed, as well as the code to implement the concepts in this chapter, will be provided in the accompanying code repository. Using OpenMVS is optional, and we may stop after getting the sparse reconstruction. However, the full MVS reconstruction is much more impressive and useful; for instance, for 3D printing replicas.

Any set of photos with sufficient overlap may be sufficient for 3D reconstruction. For example, we may use a set of photos I took of the Crazy Horse memorial head in South Dakota that is bundled with this chapter code. The requirement is that the images should be taken with sufficient movement between them, but enough to have significant overlap to allow for a strong pair-wise match.

In the following example from the Crazy Horse memorial dataset, we can notice a slight change in view angle between the images, with a very strong overlap. Notice how we can also see a great variation below the statue where people are walking about; this will not interfere with the 3D reconstruction of the stone face:

The code files for this book can be downloaded from https://github.com/PacktPublishing/Mastering-OpenCV-4-Third-Edition.