You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
If using surfel densification, especially if setting the minimum observation count to 1, there can be quite some noise in surfaces. This is likely because of surfels with roughly correct position but a wrong normal vector, which won't be affected by the free-space outlier removal.
If they were included in that outlier removal, then it might possibly cause issues for the reconstruction of thin surfaces, or generally object edges, however, since in these cases, surfels with strongly differing normals are close together but should not cancel each other out.
An alternative way to reduce the apparent noise might be to store the surfel normal in the vertex data used for rendering, and hide a surfel if its normal points away from the camera (i.e., backface culling for surfels).
The text was updated successfully, but these errors were encountered:
If using surfel densification, especially if setting the minimum observation count to 1, there can be quite some noise in surfaces. This is likely because of surfels with roughly correct position but a wrong normal vector, which won't be affected by the free-space outlier removal.
If they were included in that outlier removal, then it might possibly cause issues for the reconstruction of thin surfaces, or generally object edges, however, since in these cases, surfels with strongly differing normals are close together but should not cancel each other out.
An alternative way to reduce the apparent noise might be to store the surfel normal in the vertex data used for rendering, and hide a surfel if its normal points away from the camera (i.e., backface culling for surfels).
The text was updated successfully, but these errors were encountered: