Published online by Cambridge University Press: 22 July 2014
Some image analysis techniques are developed for simplifying lattice fringe images of deposited pyrocarbon in carbon/carbon composites by chemical vapor infiltration. They are mainly the object counting method for detecting the optimum threshold, the self-adaptive morphological filtering, the node-separation technique for breaking the aggregate fringes, and some post processing algorithms for reconstructing the fringes. The simplified fringes are the foundation for defining and extracting quantitative nanostructure parameters of pyrocarbon. The frequency filter window of a Fourier transform is defined as the circular band that retains only those fringes with interlayer distance between 0.3 and 0.45 nm. Some judge criteria are set to define topological relation between fringes. For example, the aspect ratio and area of fringes are employed to detect aggregate fringes. Fringe coaxality and distance between endpoints are used to judge the disconnected fringes. The optimum values are determined by using the iterative correction techniques. The best cut-off value for the short fringes is chosen only when there is a reasonable match between the mean fringe length and the value measured by X-ray diffraction. The adopted techniques have been verified to be feasible and to have the potential to convert the complex lattice fringe image to a set of distinct fringe structures.