Anisotropic Ellipsoidal Smoothing of Volume Data

ing the sharp edges between them [2, 3, 4, 5]. They are appropriate for removing noise before edge detection. However, they are not as useful for filling gaps in surface-like high density regions. We developed a simple filter specifically tailored for this purpose. Briefly, we predicted the surface normal direction, and then used oblate (flattened) ellipsoidal filters, whose short axis were oriented along the predicted normal, so as not to blur the surface thickness, and whose longer axes were oriented tangent to the surface, to blur across the gaps. This paper describes a simple image processing algorithm for identifying and smoothing cell membranes in tomographic reconstructions of electron micrographs of frozen bacteria. CR