Fig 1. XYZ view of the final reconstruction. |
A triple tilt series of a Caulobacter Crescentus specimen was acquired with rotating the sample around the beam axis at 0, 45, and 90 degrees, and collecting for each of those situations a classical single tilt series. For each series, 61 micrographs were taken by moving the sample around the same tilt axis (perpendicular to the optic axis) with a 2 degree increment from -60 to +60 degrees.
This technique allows to reduce quantitatively the reconstruction artifacts associated with the missing wedge problem. At the same time, because three time as much data as in a regular single tilt-series is now available, the background noise becomes also pretty low.
As a drawback of this technique, the sample should be able to withstand more electron beam damage.
A bin version of the dataset (500‚òì500 pixels) is available here.
When running the TxBR bundle adjustment with a projective model (n2=1), the resulting mean reprojection distance error is 0.2 pixel. In figure 2, the error by tilt, by patch and by tilt and patch are displayed.
Fig 2. Reprojection error result from the bundle adjustment.
On figure 1 and figure 3, views from the 3D reconstruction are shown.
Fig 3. XYZ view of the final reconstruction.