C6-P-09Image enhancement of electron microscope images with Gabor wavelets-based filtering

Abstract

In general, an electron microscopic image is processed with some enhancement tools, such as integration types, various kinds of filtering and so on, because of the low sn ratio and low image contrast. Since, it is essentially desirable to suppress the electron dose, such an image enhancement method is significant. The Fourier filtering especially for enhancement of a crystal lattice image is well known as one among those methods. In this type of filtering (conventional case), diffraction spots and streaks appearing in a pattern of Fourier transform are extracted and the other spectra are filtered. However, because Fourier transform of an image reflects all localized lattice information, artificial lattice images are mixed in the filtered image. Fourier filtering cannot enhance local lattice images only. On the other hand, Gabor wavelets-based filtering is an alternative method to resolve the problem. It is well-known that the response characteristics of the cortical neurons are best modeled with a family of self-similar 2D Gabor wavelets [ 1 ]. We think that 2D Gabor wavelets are suitable to analyse the frequency characteristics of local lattice images. It seems that there are few applications of the Gabor wavelets filtering to electron microscopic images. Then, we have applied the Gabor wavelets filtering to some high-resolution transmission electron microscope images of locally crystalized specimens. As a result, local lattice image enhancements were achieved avoiding the artificial generation of the Fourier spectral lattice images.