Dense Pointcloud Generation

As discussed under Structure from Motion, the initial result of an SfM run consists of an estimation of the original camera positions and orientations, rectified versions of the original input images, and a sparse point cloud. The latter can already be useful for a number of purposes, such as aligning the original images in 3D space (Photosynth style) or as base data for terrain models.

A realistic and detailed reconstruction of the scene, however, can only be based on a much denser 3D pointcloud.