Page 25 - 3DF Zephyr

 Output Type: raw method returns only 3D points with real 2D matches in the images, while the other two optimizations exploit some filters in post-production to make the point cloud cleaner and denser. These filters are designed to keep sharp edges (that is quite helpful when you have to deal with buildings or interiors for instance) or smooth edges (terrains, human bodies, etc.)  Number of nearest cameras: the amount of cameras used to create depth maps that is also linked to the amount of pictures taken. You'd better increase that value if you guess that there are N photos, for each camera, those are aiming at the same point of the scene. Increase the number of cameras may increase the processing time.  Resolution: it defines the resolution used for the pictures. A higher resolution keeps more details, but it may produce noisy results if the photos quality is low. The time required for the dense point cloud generation is strictly related to this parameter.  Noise filtering: low values will produce much denser clouds with noisy areas though, whereas high values will produce cleaner clouds with less endpoints.  Discretization Level: allows to Control the level of discretization of the bounding volume. A low value will produce a less dense point cloud even if it will make the