Multistage Segmentation of Lamellae Colonies Based on Directional Filter Bank and PCA Analysis

Abstract

Abstract. This paper presents a new approach to segment heavily 3D textured images such as the one of lamellar titatnium alloys obtained from X-ray tomography. The presented method considers NSDFB and gradient from gray-level value to recognize directionality of structure in the image. Second level of classication is needed due to the high complexity of the titanium alloys structure. During the segmentation algorithm takes into account the interaction between objects.