Surface is one of the important features in industrial CT (ICT) cubic data. In actual, most of the surface features do not exist alone, but exist in the form of different gray regions. 3D-Wedgelet, one of the multiscale geometric analysis methods, provides a powerful competence to describe this structure. After analyzing the component and relation of Planelet in mono-scale, a fast 3D-Wedgelet decomposition method is gained. On the basis of it, two ICT cubic datas surface feature extraction methods are proposed. The first one is based on multiscale 3D-Wedgelet decomposition. The second one is composed of several parts as follows, first dividing the ICT cubic data into 3 slice group in 3 perpendicular directions each other, then extracting linear features based on multiscale Wedgelet decomposition, finally fusing the 3 slice groups linear features to gain surface features. Numerical experimental results show that the methods can extract the surface features of ICT cubic data effectively.